Update LOCAL_PATCHES after libsanitizer merge.
[official-gcc.git] / gcc / ChangeLog
blob820b958b585b06753af8c2ba848432557c44e8fd
1 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
4         (mulsi3_600_lib): Remove pattern.
5         (umulsi3_highpart_600_lib_le): Likewise.
6         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
7         (umulsidi3): Remove call to umulsidi3_600_lib.
8         (umulsidi3_600_lib): Remove pattern.
9         (peephole2): Remove peephole using the above deprecated patterns.
11 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
13         PR target/87374
14         * config/arm/arm.c (arm_option_check_internal): Disable the combined
15         use of -mslow-flash-data and -mword-relocations.
16         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
17         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
18         flag_pic.
19         * doc/invoke.texi (-mword-relocations): Mention conflict with
20         -mslow-flash-data.
21         (-mslow-flash-data): Reciprocally.
23 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
25         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
26         16-byte modes held in GP registers to use an even regno.
28         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
29         (aarch64_atomic_ldop_supported_p): Remove.
30         (aarch64_gen_atomic_ldop): Remove.
31         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
32         Fully expand LSE operations here.
33         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
34         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
35         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
36         and use ATOMIC_LDOP instead; use register_operand for the input;
37         drop the split and emit insns directly.
38         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
39         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
40         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
42         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
43         (aarch64_gen_atomic_ldop): Don't call it.
44         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
45         Use aarch64_reg_or_zero.
46         (aarch64_atomic_exchange<ALLI>): Likewise.
47         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
48         operand 0; use aarch64_reg_or_zero for input; merge ...
49         (@aarch64_atomic_swp<ALLI>): ... this and remove.
51         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
52         (aarch64_split_compare_and_swap): Use it.
53         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
54         test oldval against the proper predicate.
55         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
56         Use nonmemory_operand for expected.
57         (cas_short_expected_pred): New.
58         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
59         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
60         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
61         (aarch64_plushi_operand): New.
63         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
64         Force oldval into the rval register for TARGET_LSE; emit the compare
65         during initial expansion so that it may be deleted if unused.
66         (aarch64_gen_atomic_cas): Remove.
67         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
68         Change =&r to +r for operand 0; use match_dup for operand 2;
69         remove is_weak and mod_f operands as unused.  Drop the split
70         and merge with...
71         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
72         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
73         (@aarch64_atomic_cas<GPI>): Similarly.
75 2018-10-31  Richard Biener  <rguenther@suse.de>
77         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
78         using ABSU_EXPR.
80 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
82         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
83         * config/aarch64/aarch64.md: Include saphira.md
84         * config/aarch64/saphira.md: New file for pipeline description.
86 2018-10-30  Martin Sebor  <msebor@redhat.com>
88         PR middle-end/87041
89         * gimple-ssa-sprintf.c (format_directive): Use %G to include
90         inlining context.
91         (sprintf_dom_walker::compute_format_length):
92         Avoid setting POSUNDER4K here.
93         (get_destination_size): Handle null argument values.
94         (get_user_idx_format): New function.
95         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
96         functions, including user-defined with attribute format printf.
97         Use %G to include inlining context.
98         Set POSUNDER4K here.
100 2018-10-30  Jan Hubicka  <jh@suse.cz>
102         * params.def (lto-partitions): Bump from 32 to 128.
104 2018-10-30  Jan Hubicka  <jh@suse.cz>
106         * tree.c
107         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
108         head in file.
109         (free_lang_data_in_type): Forward declare.
110         (fld_type_variant_equal_p): New function.
111         (fld_type_variant): New function
112         (fld_incomplete_types): New hash.
113         (fld_incomplete_type_of): New function
114         (fld_simplfied-type): New function.
115         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
116         (free_lang_data): Allocate and free fld_incomplete_type; update call
117         of free_lang_data_in_decl.
119 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
121         * gcov.c (output_lines): Remove duplicate line.
123 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
125         * config/rs6000/rs6000.md (bswapdi2): Force address into register
126         if not in indexed or indirect form.
127         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
128         (bswapdi2_store): Ditto.
129         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
130         helper function.
131         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
132         Prototype for helper function.
134 2018-10-30  Martin Sebor  <msebor@redhat.com>
136         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
137         (target, pragma GCC optimize, pragma GCC target): Ditto.
139 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
141         * doc/extend.texi: Fix prototype and description of
142         __builtin_expect_with_probability.
144 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
146         * cgraph.h (clone_function_name_1): Replaced by new
147           clone_function_name_numbered that takes name as string; for
148           privatize_symbol_name_1 use only.
149           (clone_function_name): Renamed to
150           clone_function_name_numbered to be explicit about numbering.
151           (clone_function_name): New two-argument function that does
152           not number its output.
153           (clone_function_name): New three-argument function that
154           takes a number to append to its output.
155         * cgraphclones.c (duplicate_thunk_for_node):
156           (clone_function_name_1): Renamed.
157           (clone_function_name_numbered): Two new functions.
158           (clone_function_name): Improved documentation.
159           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
160         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
161         * final.c (final_scan_insn_1): Use the new clone_function_name
162           without numbering.
163         * multiple_target.c (create_dispatcher_calls): Ditto.
164           (create_target_clone): Ditto.
165         * omp-expand.c (grid_expand_target_grid_body): Ditto.
166         * omp-low.c (create_omp_child_function_name): Ditto.
167         * omp-simd-clone.c (simd_clone_create): Ditto.
168         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
169           new clone_function_name without numbering.
171 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
173         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
174         Assert that the allocation size is not zero.
176 2018-10-30  Richard Biener  <rguenther@suse.de>
178         PR tree-optimization/87800
179         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
180         non-induction or reduction PHIs.
182 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
184         * config/aarch64/falkor-tag-collision-avoidance.c
185         (execute_tag_collision_avoidance): Call df_note_add_problem.
187 2018-10-30  Martin Liska  <mliska@suse.cz>
189         * doc/extend.texi: Fix typo in documentation
190         of __builtin_expect_with_probability.
192 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
194         PR c++/87721
195         * input.c (get_substring_ranges_for_loc): Detect if
196         linemap_resolve_location gives us a NULL map, and reject
197         this case.
199 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
201         * config.gcc (xstormy16-*-elf): Set tm_d_file.
203 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
204             Martin Sebor  <msebor@redhat.com>
205             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
207         * doc/ux.texi (Quoting): New subsection, adapted from material at
208         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
209         MartinSebor and ManuelLopezIbanez.
210         (Fix-it hints): Note that fix-it hints shouldn't be marked for
211         translation.
213 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
215         PR middle-end/87469
216         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
217         max value.
219 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
221         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
223 2018-10-29  Paul Koning  <ni1d@arrl.net>
225         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
227 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
229         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
230         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
231         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
232         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
233         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
234         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
235         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
236         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
237         Change 'vector' to '__vector'.
238         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
239         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
240         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
241         _mm_avg_pu16): Likewise.  And, whitespace corrections.
243 2018-10-29  Richard Biener  <rguenther@suse.de>
245         PR tree-optimization/87785
246         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
247         internal defs.
249 2018-10-29  Olivier Hainque  <hainque@adacore.com>
251         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
253 2018-10-29  Olivier Hainque  <hainque@adacore.com>
255         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
256         ports configuration macro, defaults to "gnu".
257         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
258         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
260 2018-10-29  Olivier Hainque  <hainque@adacore.com>
262         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
264 2018-10-29  Richard Biener  <rguenther@suse.de>
266         PR tree-optimization/87790
267         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
268         (vect_make_slp_decision): Adjust.
269         (vect_slp_analyze_bb_1): Likewise.
270         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
271         edges.
273 2018-10-29  Richard Biener  <rguenther@suse.de>
275         PR tree-optimization/87785
276         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
277         and processing.
278         (vect_build_slp_tree): Likewise.
279         (vect_gather_slp_loads): New function.
280         (vect_analyze_slp_instance): Gather loads separately from the
281         SLP tree build.
283 2018-10-29  Martin Liska  <mliska@suse.cz>
285         * Makefile.in: Make dependency to json.o.
286         * doc/gcov.texi: Document new JSON format, remove
287         old intermediate format documentation.
288         * gcov.c (struct function_info): Come up with m_name and
289         m_demangled_name.
290         (function_info::function_info): Initialize it.
291         (function_info::~function_info): Release it.
292         (main): Rename flag_intermediate_format to flag_json_format.
293         (print_usage): Describe --json-format.
294         (process_args): Set flag_json_format.
295         (output_intermediate_line): Remove.
296         (output_intermediate_json_line): Likewise.
297         (get_gcov_intermediate_filename): Return new extension
298         ".gcov.json.gz".
299         (output_intermediate_file): Implement JSON emission.
300         (output_json_intermediate_file): Implement JSON emission.
301         (generate_results): Use ::get_name for function name.
302         Handle JSON output file.
303         (read_graph_file): Use ::get_name instead of cplus_demangle.
304         (read_count_file): Likewise.
305         (solve_flow_graph): Likewise.
306         (add_line_counts): Likewise.
307         (accumulate_line_counts): Use new flag_json_format.
308         (output_function_details): Use ::get_name instead of cplus_demangle.
309         (output_lines): Likewise.
310         * json.cc (test_writing_literals): Add new tests.
311         * json.h (class literal): Add new boolean constructor.
313 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
315         PR rtl-optimization/87701
316         PR rtl-optimization/87780
317         * combine.c (make_more_copies): Rewrite.
319 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
321         * doc/generic.texi (ABSU_EXPR): Document.
322         * match.pd (absu(x)*absu(x) -> x*x): Handle.
323         (absu(absu(X)) -> absu(X)): Likewise.
324         (absu(-X) -> absu(X)): Likewise.
325         (absu(X)  where X is nonnegative -> X): Likewise.
327 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
329         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
330         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
331         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
332         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
333         (s-tm-texi): Also check timestamp on d-target.def.
334         (generated_files): Add TM_D_H and d-target-hooks-def.h.
335         (build/genhooks.o): Also depend on D_TARGET_DEF.
336         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
337         variables.
338         * config/aarch64/aarch64-d.c: New file.
339         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
340         Define.
341         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
342         prototype.
343         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
344         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
345         * config/arm/arm-d.c: New file.
346         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
347         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
348         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
349         * config/arm/t-arm (arm-d.o): New rule.
350         * config/default-d.c: New file.
351         * config/glibc-d.c: New file.
352         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
353         * config/i386/i386-d.c: New file.
354         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
355         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
356         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
357         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
358         * config/i386/t-i386 (i386-d.o): New rule.
359         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
360         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
361         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
362         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
363         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
364         * config/mips/mips-d.c: New file.
365         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
366         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
367         * config/mips/t-mips (mips-d.o): New rule.
368         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
369         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
370         * config/powerpcspe/powerpcspe-d.c: New file.
371         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
372         New prototype.
373         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
374         Support GNU D by using 0 as the language type.
375         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
376         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
377         * config/riscv/riscv-d.c: New file.
378         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
379         prototype.
380         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
381         * config/riscv/t-riscv (riscv-d.o): New rule.
382         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
383         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
384         * config/rs6000/rs6000-d.c: New file.
385         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
386         prototype.
387         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
388         Support GNU D by using 0 as the language type.
389         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
390         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
391         * config/s390/s390-d.c: New file.
392         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
393         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
394         * config/s390/t-s390 (s390-d.o): New rule.
395         * config/sparc/sparc-d.c: New file.
396         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
397         prototype.
398         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
399         * config/sparc/t-sparc (sparc-d.o): New rule.
400         * config/t-glibc (glibc-d.o): New rule.
401         * configure: Regenerated.
402         * configure.ac (tm_d_file): New variable.
403         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
404         * doc/contrib.texi (Contributors): Add self for the D frontend.
405         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
406         * doc/install.texi (Configuration): Mention libphobos as an option for
407         --enable-shared.  Mention d as an option for --enable-languages.
408         (Testing): Mention check-d as a target.
409         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
410         name suffixes.  Mention d as a -x option.
411         * doc/sourcebuild.texi (Top Level): Mention libphobos.
412         * doc/standards.texi (Standards): Add section on D language.
413         * doc/tm.texi: Regenerated.
414         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
415         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
416         * dwarf2out.c (is_dlang): New function.
417         (gen_compile_unit_die): Use DW_LANG_D for D.
418         (declare_in_namespace): Return module die for D, instead of adding
419         extra declarations into the namespace.
420         (gen_namespace_die): Generate DW_TAG_module for D.
421         (gen_decl_die): Handle CONST_DECLSs for D.
422         (dwarf2out_decl): Likewise.
423         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
424         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
425         kinds of aggregates.
426         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
427         * genhooks.c: Include d/d-target.def.
429 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
431         PR target/85669
432         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
433         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
434         (STACK_DYNAMIC_OFFSET): Likewise.
436 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
438         PR target/80024
439         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
440         error message.
442 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
444         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
445         return type and other typos.
447 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
449         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
450         a shorter sequence with fewer branches.
451         (emit_final_str_compare_gpr): Ditto.
453 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
455         * config/rs6000/tmmintrin.h: New file.
456         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
458 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
460         * config/rs6000/mmintrin.h: Enable 32bit compilation.
461         * config/rs6000/xmmintrin.h: Likewise.
463 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
465         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
467 2018-10-26  Richard Biener  <rguenther@suse.de>
469         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
470         and wrapper.
471         (vect_mark_slp_stmts_relevant): Likewise.
472         (vect_detect_hybrid_slp_stmts): Likewise.
473         (vect_bb_slp_scalar_cost): Likewise.
474         (vect_remove_slp_scalar_calls): Likewise.
476 2018-10-26  Jan Hubicka  <jh@suse.cz>
478         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
479         (warn_types_mismatch): Fix walk of DECL_NAME.
480         (odr_types_equivalent_p): Fix overactive assert.
482 2018-10-26  Richard Biener  <rguenther@suse.de>
484         PR tree-optimization/87105
485         * tree-vectorizer.h (_slp_tree::refcnt): New member.
486         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
487         refcnt.
488         (vect_create_new_slp_node): Initialize refcnt to one.
489         (bst_traits): Move.
490         (scalar_stmts_set_t, bst_fail): Remove.
491         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
492         (vect_build_slp_tree): Add bst_map argument and lookup
493         already created SLP nodes.
494         (vect_print_slp_tree): Handle a SLP graph, print SLP node
495         addresses.
496         (vect_slp_rearrange_stmts): Handle a SLP graph.
497         (vect_analyze_slp_instance): Adjust and free SLP nodes from
498         the CSE map.  Fix indenting.
499         (vect_schedule_slp_instance): Add short-cut.
501 2018-10-26  Martin Liska  <mliska@suse.cz>
503         PR testsuite/86158
504         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
505         addr_expr and not with pointers.
507 2018-10-26  Jan Hubicka  <jh@suse.cz>
509         * tree.c (free_lang_data_in_type): Only check main variants.
510         * ipa-devirt.c (warn_odr): Make static.
511         (types_same_for_odr): Drop strict variant.
512         (types_odr_comparable): Likewise.
513         (odr_or_derived_type_p): Look for main variants.
514         (odr_name_hasher::equal): Cleanup comment.
515         (odr_subtypes_equivalent): Add warn and warned arguments; check main
516         variants.
517         (type_variants_equivalent_p): break out from ...
518         (odr_types_equivalent): ... here; go for main variants where needed.
519         (warn_odr): ... here; turn static.
520         (warn_types_mismatch): Compare mangled names of main variants.
521         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
522         (type_with_linkage_p): Sanity check that we look at main variant.
523         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
524         * tree.h (types_same_for_odr): Drop strict argument.
526 2018-10-26  Richard Biener  <rguenther@suse.de>
528         PR tree-optimization/87746
529         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
530         Simplify and fix WRT strided store groups with size not
531         equal to step in element count.
532         (vect_analyze_group_access_1): Dump the whole group.
534 2018-10-25  Carl Love  <cel@us.ibm.com>
536         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
537         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
538         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
539         precicion entry for each overloaded builtin.
540         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
541         VSCEDPUO): Rename overloaded name.
542         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
543         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
544         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
545         define_expand for xscmpexqp instruction.
546         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
548 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
549             Jinsong Ji <jji@us.ibm.com>
551         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
552         function with vec_sl.
553         (_mm_slli_epi32): Likewise.
554         (_mm_slli_epi64): Likewise.
555         (_mm_srai_epi16): Replace deprecated function with vec_sra.
556         (_mm_srai_epi32): Likewise.
557         (_mm_srli_epi16): Replace deprecated function with vec_sr.
558         (_mm_srli_epi32): Likewise.
559         (_mm_srli_epi64): Likewise.
560         (_mm_sll_epi16): Replace deprecated function with vec_sl.
561         (_mm_sll_epi32): Likewise.
562         (_mm_sll_epi64): Likewise.
563         (_mm_sra_epi16): Replace deprecated function with vec_sra.
564         (_mm_sra_epi32): Likewise.
565         (_mm_srl_epi16): Replace deprecated function with vec_sr.
566         (_mm_srl_epi32): Likewise.
567         (_mm_srl_epi64): Likewise.
569 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
570             Jinsong Ji <jji@us.ibm.com>
572         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
573         comparison operators with vec_cmp* for compatibility due to
574         unfortunate history; clean up formatting and use types more
575         appropriately.
576         (_mm_sll_epi32): Likewise.
577         (_mm_sll_epi64): Likewise.
578         (_mm_srl_epi16): Likewise.
579         (_mm_srl_epi32): Likewise.
580         (_mm_srl_epi64): Likewise.
582 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
583             Jinsong Ji <jji@us.ibm.com>
585         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
586         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
587         __vector __bool int.  Use vec_cmpgt in preference to deprecated
588         function vec_vcmpgtfp.
589         (_mm_max_ps): Likewise.
591 2018-10-25  Jeff Law  <law@redhat.com>
593         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
594         if returning false.
596 2018-10-25  Martin Sebor  <msebor@redhat.com>
598         * doc/extend.texi (aligned): Expand attribute description.
599         (Alignment): Rename section.  Discuss function arguments.
601 2018-10-25  Jan Hubicka  <jh@suse.cz>
603         * ipa-devirt.c (main_odr_variant): Remove.
604         (hash_odr_name, types_same_for_odr, types_odr_comparable,
605         odr_name_hasher::equal, odr_subtypes_equivalent_p):
606         Drop use of main_odr_variant.
607         (add_type_duplicate): Silence confused warnings on integer types.
608         (get_odr_type): Always look for main variant.
609         (register_odr_type): Simplify.
611 2018-10-25  Richard Biener  <rguenther@suse.de>
613         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
614         Initialize ng to silence error with release checking bootstrap.
616 2018-10-25  Richard Biener  <rguenther@suse.de>
618         * tree-if-conv.c: Include tree-ssa-sccvn.h.
619         (tree_if_conversion): Run CSE on the if-converted loop body.
621 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
623         * config/s390/constraints.md (ZL): New constraint.
624         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
625         operands.
626         * config/s390/s390.md (movdi_larl): Remove.
627         (movdi_64): Add the LARL alternative.
629 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
631         PR bootstrap/87747
632         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
633         (RTX_CODE_HWINT_P): New macro.
634         (rtx_code_size): Use RTX_CODE_HWINT_P ().
636 2018-10-25  Jan Hubicka  <jh@suse.cz>
638         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
639         is anonymous.
641 2018-10-25  Richard Biener  <rguenther@suse.de>
643         PR tree-optimization/87665
644         PR tree-optimization/87745
645         * tree-vectorizer.h (get_earlier_stmt): Remove.
646         (get_later_stmt): Pick up UID from the original non-pattern stmt.
648 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
650         * options.texi (Deprecated): Move list to Var section.
652 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
653             Jinsong Ji <jji@us.ibm.com>
655         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
656         __vector long to __vector long long.
657         (_mm_cvtpd_ps): Likewise.
658         (_mm_cvttpd_epi32): Likewise.
659         (_mm_cvtpi32_pd): Likewise.
660         (_mm_unpackhi_epi64): Likewise.
661         (_mm_unpacklo_epi64): Likewise.
663 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
665         PR rtl-optimization/87720
666         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
668 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
670         * gimple-ssa-isolate-paths.c
671         (find_implicit_erroneous_behavior): Do not change code if the
672         pass is running for warnings only.
673         (find_explicit_erroneous_behavior): Likewise.
675 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
677         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
678         Define as rs6000_mangle_decl_assembler_name.
679         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
680         long double to IEEE long double, switch the names of the long
681         double built-in functions to be <func>f128 instead of <func>l.
683 2018-10-24  Martin Sebor  <msebor@redhat.com>
685         * doc/extend.texi (nonnull): List no-argument form.  Reference
686         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
688 2018-10-24  Richard Biener  <rguenther@suse.de>
690         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
692 2018-10-24  Martin Liska  <mliska@suse.cz>
694         PR tree-optimization/84436
695         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
696         Remove.
697         (switch_conversion::contains_linear_function_p): New.
698         (switch_conversion::build_one_array): Support linear
699         transformation on input.
700         * tree-switch-conversion.h (struct switch_conversion): Add
701         contains_linear_function_p declaration.
703 2018-10-24  Richard Biener  <rguenther@suse.de>
705         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
706         if its argument is CONSTANT_CLASS_P.
708 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
710         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
711         it is wrong for forward declarations.
713 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
715         * config/s390/s390.c (s390_check_qrst_address): Add the missing
716         SYMBOL_REF_P () check.
718 2018-10-24  Richard Biener  <rguenther@suse.de>
720         PR tree-optimization/87105
721         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
722         dump classification.
723         (vect_analyze_data_ref_accesses): Handle duplicate loads and
724         stores by splitting the affected group after the fact.
725         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
726         fail the SLP build because of size constraints.
728 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
730         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
731         * configure: Regenerate.
732         * config.in: Regenerate.
733         * varasm.c (mergeable_string_section): Use readonly_data_section
734         if linker doesn't support SHF_MERGE with alignment > 8.
735         (mergeable_constant_section): Likewise.
737 2018-10-24  Richard Biener  <rguenther@suse.de>
739         PR tree-optimization/84013
740         * tree-ssa-structalias.c (struct msdi_data): New struct for
741         marshalling data to walk_stmt_load_store_ops.
742         (maybe_set_dependence_info): Refactor as callback for
743         walk_stmt_load_store_ops.
744         (compute_dependence_clique): Set restrict info on all stmt kinds.
746 2018-10-24  Martin Liska  <mliska@suse.cz>
748         * cgraph.c (cgraph_node::dump):
749         Remove reduntant dumps and make tp_first_run dump more compact.
751 2018-10-24  Richard Biener  <rguenther@suse.de>
753         PR tree-optimization/87665
754         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
755         to reflect reality.
757 2018-10-12  Jeff Law  <law@redhat.com>
759         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
760         for H8/S.
762 2018-10-23  Richard Biener  <rguenther@suse.de>
764         * tree-vrp.c (add_assert_info): Guard dump_printf with
765         dump_enabled_p.
766         * gimple-ssa-evrp-analyze.c
767         (evrp_range_analyzer::record_ranges_from_incoming_edge):
768         Use value_range::ignore_equivs_equal_p.
770 2018-10-23  Richard Biener  <rguenther@suse.de>
772         PR tree-optimization/87105
773         PR tree-optimization/87608
774         * passes.def (pass_all_early_optimizations): Add early phi-opt
775         after dce.
776         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
777         addition to debug stmts.
778         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
779         and abs replacement early.
780         * tree-cfg.c (gimple_empty_block_p): Likewise.
782 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
784         PR target/86383
785         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
786         specified to configure.
787         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
789 2018-10-23  Richard Biener  <rguenther@suse.de>
791         PR tree-optimization/87700
792         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
794 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
796         PR target/87674
797         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
798         second argument from __mmask16 to __mmask8.
799         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
800         _mm_mask_packs_epi32): Likewise.
801         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
802         Likewise.
803         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
805 2018-10-23  Richard Biener  <rguenther@suse.de>
807         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
809 2018-10-23  Richard Biener  <rguenther@suse.de>
811         PR tree-optimization/86144
812         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
813         over simd attribute.
815 2018-10-23  Richard Biener  <rguenther@suse.de>
817         PR tree-optimization/87693
818         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
819         the case we do not find the taken edge.
821 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
822             Jinsong Ji  <jji@us.ibm.com>
824         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
825         (_mm_store_pd): Use unaligned vector type for pointer cast.
826         (_mm_maskmoveu_si128): Likewise.
827         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
828         (_mm_store_ps): Use unaligned vector type for pointer cast.
830 2018-10-22  Paul Koning  <ni1d@arrl.net>
832         * symtab.c (symtab_node::increase_alignment): Correct max
833         alignment check.
835 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
837         PR tree-optimization/87633
838         * match.pd: Do not generate unordered integer comparisons.
840 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
842         PR rtl-optimization/87600
843         * combine.c: Add include of expr.h.
844         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
845         register to a pseudo.
846         (make_more_copies): New function, add a copy to a new pseudo after
847         the moves from hard registers into pseudos.
848         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
849         later.  Call make_more_copies.
851 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
853         * lra-constraints.c (process_alt_operands): New local array,
854         matching_early_clobber.  Check matching_early_clobber before
855         decrementing reject, and set matching_early_clobber after.
857 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
859         PR target/87598
860         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
861         call output_addr_const and hope for the best.
863 2018-10-22  Richard Biener  <rguenther@suse.de>
865         * gimple-ssa-evrp-analyze.c
866         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
867         smarter about what ranges to use.
868         * tree-vrp.c (add_assert_info): Dump here.
869         (register_edge_assert_for_2): Instead of here at multiple but
870         not all places.
872         * gcc.dg/tree-ssa/evrp12.c: New testcase.
873         * gcc.dg/predict-6.c: Adjust.
874         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
875         * gcc.dg/tree-ssa/vrp02.c: Likewise.
876         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
878 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
879         Richard Biener  <rguenther@suse.de>
881         * bitmap.h: Update data structure documentation, including a
882         description of bitmap views as either linked-lists or splay trees.
883         (struct bitmap_element_def): Update comments for splay tree bitmaps.
884         (struct bitmap_head_def): Likewise.
885         (bitmap_list_view, bitmap_tree_view): New prototypes.
886         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
887         tree_form fields.
888         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
889         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
890         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
891         released bitmap element here.
892         (bitmap_element_free): Remove.
893         (bitmap_elt_clear_from): Work on splay tree bitmaps.
894         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
895         this function similar ones such that linked-list bitmap implementation
896         functions are grouped.
897         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
898         and moved for grouping.
899         (bitmap_list_insert_element_after): Renamed from
900         bitmap_elt_insert_after, and moved for grouping.
901         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
902         (bitmap_tree_link_left, bitmap_tree_link_right,
903         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
904         bitmap_tree_link_element, bitmap_tree_unlink_element,
905         bitmap_tree_find_element): New functions for splay-tree bitmap
906         implementation.
907         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
908         Renamed and moved, see above entries.
909         (bitmap_tree_listify_from): New function to convert part of a splay
910         tree bitmap to a linked-list bitmap.
911         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
912         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
913         (bitmap_find_bit): Remove.
914         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
915         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
916         Handle splay tree bitmaps.
917         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
918         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
919         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
920         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
921         bitmap_intersect_compl_p, bitmap_ior_and_compl,
922         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
923         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
924         corresponding changes to use linked-list specific bitmap_element
925         manipulation functions as applicable for efficiency.
926         (bitmap_tree_to_vec): New function.
927         (debug_bitmap_elt_file): New function split out from ...
928         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
929         (bitmap_print): Likewise.
931         PR tree-optimization/63155
932         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
933         SSA edge worklists.
934         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
935         in tree-view.
937 2018-10-22  Martin Liska  <mliska@suse.cz>
939         PR tree-optimization/87686
940         Revert
941         2018-08-29  Martin Liska  <mliska@suse.cz>
943         * tree-switch-conversion.c (switch_conversion::expand):
944         Strenghten assumption about gswitch statements.
946 2018-10-22  Martin Liska  <mliska@suse.cz>
948         * ipa-icf.c (sem_item::compare_attributes): Remove.
949         (sem_item::compare_referenced_symbol_properties): Use
950         attribute_list_equal instead.
951         (sem_function::equals_wpa): Likewise.
952         * ipa-icf.h: Remove compare_attributes.
954 2018-10-22  Richard Biener  <rguenther@suse.de>
956         PR middle-end/87682
957         * mem-stats.h (mem_usage::operator==): Fix pasto.
959 2018-10-22  Richard Biener  <rguenther@suse.de>
961         PR tree-optimization/87640
962         * tree-vrp.c (set_value_range_with_overflow): Decompose
963         incomplete result.
964         (extract_range_from_binary_expr_1): Adjust.
966 2018-10-22  Martin Jambor  <mjambor@suse.cz>
968         * tree-eh.h (stmt_could_throw_p): Add function parameter.
969         (stmt_can_throw_external): Likewise.
970         (stmt_can_throw_internal): Likewise.
971         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
972         (lower_eh_constructs_2): Likewise.
973         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
974         (stmt_can_throw_external): Likewise.
975         (stmt_can_throw_internal): Likewise.
976         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
977         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
978         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
979         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
980         (pass_lower_eh_dispatch::execute): Pass cfun to
981         stmt_can_throw_external.
982         (cleanup_empty_eh): Likewise.
983         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
984         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
985         stmt_can_throw_external instead of pushing it to cfun.
986         (symbol_table::create_edge): Likewise.
987         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
988         stmt_can_throw_internal.
989         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
990         to stmt_could_throw_p.
991         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
992         stmt_can_throw_internal.
993         (pass_store_merging::execute): Likewise.
994         * gimple-ssa-strength-reduction.c
995         (find_candidates_dom_walker::before_dom_children): Pass cfun to
996         stmt_could_throw_p.
997         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
998         stmt_can_throw_internal.
999         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1000         to stmt_can_throw_external.
1001         (check_stmt): Pass cfun to stmt_could_throw_p.
1002         (check_stmt): Pass cfun to stmt_can_throw_external.
1003         (pass_nothrow::execute): Likewise.
1004         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1005         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1006         stmt_can_throw_internal.
1007         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1008         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1009         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1010         * tree-complex.c (expand_complex_libcall): Pass cfun to
1011         stmt_could_throw_p and to stmt_can_throw_internal.
1012         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1013         * tree-inline.c (copy_edges_for_bb): Likewise.
1014         (maybe_move_debug_stmts_to_successors): Likewise.
1015         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1016         stmt_could_throw_p.
1017         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1018         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1019         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1020         stmt_can_throw_internal.
1021         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1022         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1023         stmt_could_throw_p.
1024         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1025         stmt_can_throw_internal.
1026         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1027         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1028         stmt_could_throw_p.
1029         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1030         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1031         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1032         (convert_mult_to_fma_1): Likewise.
1033         (convert_to_divmod): Likewise.
1034         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1035         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1036         * tree-ssa-propagate.c
1037         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1038         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1039         (maybe_optimize_range_tests): Likewise.
1040         (linearize_expr_tree): Likewise.
1041         (reassociate_bb): Likewise.
1042         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1043         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1044         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1045         (handle_char_store): Likewise.
1046         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1047         stmt_can_throw_internal.
1048         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1049         stmt_could_throw_p.
1050         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1051         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1052         (vectorizable_simd_clone_call): Likewise.
1053         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1054         (gimple_stringop_fixed_value): Likewise.
1056 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1058         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1059         literal pool references.
1060         (s390_check_qrst_address): Adapt to the new behavior of
1061         s390_loadrelative_operand_p ().
1063 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1065         PR target/72782
1066         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1068 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1070         PR target/72782
1071         * config/i386/sse.md (*<code><mode>3_bcst): New.
1073 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1075         PR target/72782
1076         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1077         V4DI, V16SI and V8DI.
1078         (*sub<mode>3<mask_name>_bcst): New.
1079         (*add<mode>3<mask_name>_bcst): Likewise.
1081 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1082             Jinsong Ji  <jji@us.ibm.com>
1084         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1085         __m64 with __vector unsigned long long for compatibility.
1086         (_mm_movemask_epi8): Likewise.
1087         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1088         (_mm_cvttps_pi32): Likewise.
1089         (_mm_cvtpi32_ps): Likewise.
1090         (_mm_cvtps_pi16): Likewise.
1091         (_mm_loadh_pi): Likewise.
1092         (_mm_storeh_pi): Likewise.
1093         (_mm_movehl_ps): Likewise.
1094         (_mm_movelh_ps): Likewise.
1095         (_mm_loadl_pi): Likewise.
1096         (_mm_storel_pi): Likewise.
1097         (_mm_movemask_ps): Likewise.
1098         (_mm_shuffle_pi16): Likewise.
1100 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1102         PR target/72782
1103         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1104         __builtin_ia32_vfnmsubpd512_mask.
1105         (_mm512_mask_fnmsub_round_pd): Likewise.
1106         (_mm512_fnmsub_pd): Likewise.
1107         (_mm512_mask_fnmsub_pd): Likewise.
1108         (_mm512_maskz_fnmsub_round_pd): Use
1109         __builtin_ia32_vfnmsubpd512_maskz.
1110         (_mm512_maskz_fnmsub_pd): Likewise.
1111         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1112         (_mm512_mask_fnmsub_round_ps): Likewise.
1113         (_mm512_fnmsub_ps): Likewise.
1114         (_mm512_mask_fnmsub_ps): Likewise.
1115         (_mm512_maskz_fnmsub_round_ps): Use
1116         __builtin_ia32_vfnmsubps512_maskz.
1117         (_mm512_maskz_fnmsub_ps): Likewise.
1118         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1119         __builtin_ia32_vfnmsubpd256_mask.
1120         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1121         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1122         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1123         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1124         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1125         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1126         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1127         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1128         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1129         __builtin_ia32_vfnmsubpd.
1130         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1131         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1132         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1133         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1134         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1135         * config/i386/i386-builtin.def: Add
1136         __builtin_ia32_vfnmsubpd256_mask,
1137         __builtin_ia32_vfnmsubpd256_maskz,
1138         __builtin_ia32_vfnmsubpd128_mask,
1139         __builtin_ia32_vfnmsubpd128_maskz,
1140         __builtin_ia32_vfnmsubps256_mask,
1141         __builtin_ia32_vfnmsubps256_maskz,
1142         __builtin_ia32_vfnmsubps128_mask,
1143         __builtin_ia32_vfnmsubps128_maskz,
1144         __builtin_ia32_vfnmsubpd512_mask,
1145         __builtin_ia32_vfnmsubpd512_maskz,
1146         __builtin_ia32_vfnmsubps512_mask,
1147         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1148         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1149         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1150         __builtin_ia32_vfnmsubpd256.
1151         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1152         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1153         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1154         Likewise.
1155         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1156         Likewise.
1157         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1158         Likewise.
1159         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1161 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1163         PR target/72782
1164         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1165         __builtin_ia32_vfnmaddpd512_mask.
1166         (_mm512_mask_fnmadd_round_pd): Likewise.
1167         (_mm512_fnmadd_pd): Likewise.
1168         (_mm512_mask_fnmadd_pd): Likewise.
1169         (_mm512_maskz_fnmadd_round_pd): Use
1170         __builtin_ia32_vfnmaddpd512_maskz.
1171         (_mm512_maskz_fnmadd_pd): Likewise.
1172         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1173         (_mm512_mask_fnmadd_round_ps): Likewise.
1174         (_mm512_fnmadd_ps): Likewise.
1175         (_mm512_mask_fnmadd_ps): Likewise.
1176         (_mm512_maskz_fnmadd_round_ps): Use
1177         __builtin_ia32_vfnmaddps512_maskz.
1178         (_mm512_maskz_fnmadd_ps): Likewise.
1179         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1180         __builtin_ia32_vfnmaddpd256_mask.
1181         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1182         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1183         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1184         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1185         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1186         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1187         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1188         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1189         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1190         __builtin_ia32_vfnmaddpd.
1191         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1192         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1193         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1194         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1195         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1196         * config/i386/i386-builtin.def: Add
1197         __builtin_ia32_vfnmaddpd256_mask,
1198         __builtin_ia32_vfnmaddpd256_maskz,
1199         __builtin_ia32_vfnmaddpd128_mask,
1200         __builtin_ia32_vfnmaddpd128_maskz,
1201         __builtin_ia32_vfnmaddps256_mask,
1202         __builtin_ia32_vfnmaddps256_maskz,
1203         __builtin_ia32_vfnmaddps128_mask,
1204         __builtin_ia32_vfnmaddps128_maskz,
1205         __builtin_ia32_vfnmaddpd512_mask,
1206         __builtin_ia32_vfnmaddpd512_maskz,
1207         __builtin_ia32_vfnmaddps512_mask,
1208         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1209         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1210         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1211         __builtin_ia32_vfnmaddpd256.
1212         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1213         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1214         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1215         Likewise.
1216         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1217         Likewise.
1218         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1219         Likewise.
1220         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1222 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1224         PR target/72782
1225         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1226         __builtin_ia32_vfmsubpd512_mask.
1227         (_mm512_mask_fmsub_round_pd): Likewise.
1228         (_mm512_fmsub_pd): Likewise.
1229         (_mm512_mask_fmsub_pd): Likewise.
1230         (_mm512_maskz_fmsub_round_pd): Use
1231         __builtin_ia32_vfmsubpd512_maskz.
1232         (_mm512_maskz_fmsub_pd): Likewise.
1233         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1234         (_mm512_mask_fmsub_round_ps): Likewise.
1235         (_mm512_fmsub_ps): Likewise.
1236         (_mm512_mask_fmsub_ps): Likewise.
1237         (_mm512_maskz_fmsub_round_ps): Use
1238         __builtin_ia32_vfmsubps512_maskz.
1239         (_mm512_maskz_fmsub_ps): Likewise.
1240         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1241         __builtin_ia32_vfmsubpd256_mask.
1242         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1243         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1244         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1245         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1246         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1247         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1248         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1249         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1250         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1251         __builtin_ia32_vfmsubpd.
1252         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1253         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1254         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1255         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1256         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1257         * config/i386/i386-builtin.def: Add
1258         __builtin_ia32_vfmsubpd256_mask,
1259         __builtin_ia32_vfmsubpd256_maskz,
1260         __builtin_ia32_vfmsubpd128_mask,
1261         __builtin_ia32_vfmsubpd128_maskz,
1262         __builtin_ia32_vfmsubps256_mask,
1263         __builtin_ia32_vfmsubps256_maskz,
1264         __builtin_ia32_vfmsubps128_mask,
1265         __builtin_ia32_vfmsubps128_maskz,
1266         __builtin_ia32_vfmsubpd512_mask,
1267         __builtin_ia32_vfmsubpd512_maskz,
1268         __builtin_ia32_vfmsubps512_mask,
1269         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1270         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1271         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1272         __builtin_ia32_vfmsubpd256.
1273         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1274         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1275         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1276         Likewise.
1277         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1278         Likewise.
1279         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1280         Likewise.
1281         (fmai_vmfmsub_<mode><round_name>): Likewise.
1283 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1285         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1286         Remove plus.  Renamed to ...
1287         (*sub<mode>3<mask_name>_bcst): This.
1288         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1289         (*add<mode>3<mask_name>_bcst): This.
1291 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1293         PR target/72782
1294         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1296 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1298         PR target/87662
1299         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1300         (_mm_or_epi32): Likewise.
1301         (_mm256_xor_epi32): Likewise.
1302         (_mm_xor_epi32): Likewise.
1303         (_mm256_or_epi64): Likewise.
1304         (_mm_or_epi64): Likewise.
1305         (_mm256_xor_epi64): Likewise.
1306         (_mm_xor_epi64): Likewise.
1308 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1310         PR target/72782
1311         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1313 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1315         PR middle-end/87647
1316         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1318 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1320         * doc/ux.texi: Move @section directly after @node.
1322 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1324         PR middle-end/85488
1325         PR middle-end/87649
1326         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1327         depend closely nested inside of loop with ordered clause with
1328         a parameter.
1330 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1332         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1333         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1334         * doc/ux.texi: New file.
1336 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1338         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1339         be the first CR field allocated.
1341 2018-10-19  Richard Biener  <rguenther@suse.de>
1343         PR target/87657
1344         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1345         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1347 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1349         PR target/72782
1350         * config/i386/sse.md
1351         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1352         (*add<mode>3<mask_name>_bcst_2): Likewise.
1354 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1356         * config/i386/sse.md
1357         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1358         Replace nonimmediate_operand with register_operand.
1359         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1360         Likewise.
1361         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1362         Likewise.
1364 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1366         PR rtl-optimization/87596
1367         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1368         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1369         for instructions in FROM..TO range.
1371 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1373         * cfgexpand.c (expand_one_var): Use specific wording in error message
1374         for non-local frame variables.
1375         * stor-layout.c (layout_decl): Do not issue a warning for them.
1377 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1379         * haifa-sched.c (priority): Add force_recompute parameter.
1380         (apply_replacement): Call priority () with force_recompute = true.
1381         (restore_pattern): Likewise.
1383 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1385         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1386         HOST_BITS_PER_WIDE_INT.
1387         (test_vector_ops_duplicate): Likewise.
1389 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1391         PR target/72782
1392         * config/i386/sse.md (VF_AVX512): New.
1393         (avx512bcst): Likewise.
1394         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1395         Likewise.
1396         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1397         Likewise.
1398         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1399         Likewise.
1401 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
1403         * doc/invoke.texi (-dumpversion): Improve grammar.
1404         (-dumpfullversion): Make more consistent with -dumpversion.
1406 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
1408         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1409         Set JUMP_LABEL to the jump insn.
1410         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1411         Predict emitted jump and add label to jump insn.
1413 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
1415         PR tree-optimization/87562
1416         * input.c (get_substring_ranges_for_loc): Use
1417         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1418         getting the linemap for the endpoint.  Verify that it's either
1419         in the same linemap as the start point's spelling location, or
1420         at least in the same file.
1422 2018-10-18  Richard Biener  <rguenther@suse.de>
1424         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1425         feed width-specific load/store costs through ix86_vec_cost.
1426         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1427         (k8_cost): Likewise.
1428         (bdver_cost): Likewise.
1429         (znver1_cost): Likewise.
1430         (btver1_cost): Likewise.
1431         (btver2_cost): Likewise.
1433 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1435         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1436         to simplify subreg of vec_merge.
1438 2018-10-18  Richard Biener  <rguenther@suse.de>
1440         * config/i386/i386.c: Fix costing of vector FMA.
1442 2018-10-18  Richard Biener  <rguenther@suse.de>
1444         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1445         and argument.
1446         (ix86_builtin_vectorization_cost): For vec_construct properly
1447         cost insertion into SSE regs.
1448         (...): Adjust calls to ix86_vec_cost.
1450 2018-10-18  Richard Biener  <rguenther@suse.de>
1452         PR middle-end/87087
1453         Revert
1454         2018-02-07  Richard Biener  <rguenther@suse.de>
1456         PR tree-optimization/84204
1457         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1458         this place.
1460 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1462         PR target/87537
1463         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1464         of vec_duplicate.
1465         (test_vector_ops_duplicate): Add test for a scalar subreg of a
1466         VEC_MERGE of a VEC_DUPLICATE.
1468 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1470         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1471         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1472         * doc/standards.texi (C Language): Document C2X.
1473         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1474         (rl78_option_override): Handle "GNU C2X" language name.
1476 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1478         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1479         Document C17 as published in 2018.
1481 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
1483         PR middle-end/87623
1484         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1485         bail out if both sides do not have the same storage order.
1487 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1489         * bitmap.c (bitmap_head::dump): New.
1490         * bitmap.h (bitmap_head): Add dump().
1491         * gimple-ssa-evrp-analyze.c
1492         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1493         (evrp_range_analyzer::set_ssa_range_info): Same.
1494         (evrp_range_analyzer::record_ranges_from_phis): Same.
1495         (evrp_range_analyzer::record_ranges_from_stmt): Same.
1496         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1497         * gimple-ssa-sprintf.c (get_int_range): Same.
1498         (format_integer): Same.
1499         (sprintf_dom_walker::handle_gimple_call): Same.
1500         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1501         (ipcp_vr_lattice::top_p): Same.
1502         (ipcp_vr_lattice::bottom_p): Same.
1503         (ipcp_vr_lattice::set_to_bottom): Same.
1504         (ipa_vr_operation_and_type_effects): Same.
1505         (propagate_vr_across_jump_function): Same.
1506         (ipcp_store_vr_results): Same.
1507         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
1508         (ipa_print_node_jump_functions_for_edge): Same.
1509         (ipa_get_value_range): Same.
1510         (ipa_compute_jump_functions_for_edge): Same.
1511         (ipa_write_jump_function): Same.
1512         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
1513         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1514         Same.
1515         * vr-values.c (set_value_range_to_nonnegative): Same.
1516         (set_value_range_to_truthvalue): Same.
1517         (vr_values::get_value_range): Same.
1518         (vr_values::set_defs_to_varying): Same.
1519         (vr_values::update_value_range): Same.
1520         (symbolic_range_based_on_p): Same.
1521         (vr_values::op_with_boolean_value_range_p): Same.
1522         (vr_values::extract_range_for_var_from_comparison_expr): Same.
1523         (vr_values::extract_range_from_ssa_name): Same.
1524         (vr_values::extract_range_from_binary_expr): Same.
1525         (vr_values::extract_range_from_unary_expr): Same.
1526         (vr_values::extract_range_from_cond_expr): Same.
1527         (vr_values::extract_range_from_comparison): Same.
1528         (vr_values::check_for_binary_op_overflow): Same.
1529         (vr_values::extract_range_basic): Same.
1530         (vr_values::extract_range_from_assignment): Same.
1531         (compare_ranges): Same.
1532         (compare_range_with_value): Same.
1533         (vr_values::adjust_range_with_scev): Same.
1534         (vrp_valueize): Same.
1535         (vrp_valueize_1): Same.
1536         (vr_values::get_vr_for_comparison): Same.
1537         (vr_values::compare_name_with_value): Same.
1538         (vr_values::compare_names): Same.
1539         (vr_values::vrp_evaluate_conditional): Same.
1540         (find_case_label_ranges): Same.
1541         (vr_values::vrp_visit_switch_stmt): Same.
1542         (vr_values::extract_range_from_phi_node): Same.
1543         (vr_values::simplify_div_or_mod_using_ranges): Same.
1544         (vr_values::simplify_bit_ops_using_ranges): Same.
1545         (test_for_singularity): Same.
1546         (range_fits_type_p): Same.
1547         (vr_values::simplify_cond_using_ranges_1): Same.
1548         (vr_values::simplify_switch_using_ranges): Same.
1549         (vr_values::simplify_float_conversion_using_ranges): Same.
1550         (vr_values::two_valued_val_range_p): Same.
1551         (vr_values::add_equivalence): Move to value_range::equiv_add.
1552         * vr-values.h (vr_values::add_equivalence): Remove.
1553         (VR_INITIALIZER): Remove.
1554         * tree-vrp.c (value_range::set): New.
1555         (value_range::equiv_add): New.
1556         (value_range::value_range): New.
1557         (value_range::deep_copy): New.
1558         (value_range::check): New.
1559         (value_range::equal_p): New.
1560         (value_range::ignore_equivs_equal_p): New.
1561         (value_range::operator==): New.
1562         (value_range::operator!=): New.
1563         (value_range::symbolic_p): New.
1564         (value_range::numeric_p): New.
1565         (value_range::set_undefined): New.
1566         (value_range::set_varying): New.
1567         (value_range::may_contain_p): New.
1568         (value_range::equiv_clear): New.
1569         (value_range::singleton_p): New.
1570         (value_range::intersect): New.
1571         (value_range::dump): New.
1572         (value_range::set_and_canonicalize): New.
1573         (set_value_range): Adjust for value_range API.
1574         (set_value_range_to_undefined): Same.
1575         (set_value_range_to_varying): Same.
1576         (set_and_canonicalize_value_range): Same.
1577         (set_value_range_to_nonnull): Same.
1578         (set_value_range_to_null): Same.
1579         (range_is_null): Same.
1580         (range_is_nonnull): Same.
1581         (range_int_cst_p): Same.
1582         (range_int_cst_singleton_p): Same.
1583         (symbolic_range_p): Same.
1584         (range_includes_zero_p): Same.
1585         (value_range_constant_singleton): Same.
1586         (vrp_set_zero_nonzero_bits): Same.
1587         (ranges_from_anti_range): Same.
1588         (extract_range_into_wide_ints): Same.
1589         (extract_range_from_multiplicative_op): Same.
1590         (set_value_range_with_overflow): Same.
1591         (extract_range_from_binary_expr_1): Same.
1592         (extract_range_from_unary_expr): Same.
1593         (dump_value_range): Same.
1594         (debug_value_range): Same.
1595         (vrp_prop::check_array_ref): Same.
1596         (vrp_prop::check_mem_ref): Same.
1597         (vrp_prop::vrp_initialize): Same.
1598         (vrp_prop::visit_stmt): Same.
1599         (intersect_ranges): Same.
1600         (vrp_prop::visit_phi): Same.
1601         (vrp_prop::vrp_finalize): Same.
1602         (determine_value_range_1): Same.
1603         (determine_value_range): Same.
1604         (vrp_intersect_ranges_1): Rename to...
1605         (vrp_intersect_1): this.
1606         (vrp_intersect_ranges): Rename to...
1607         (value_range::intersect_helper): ...this.
1608         (vrp_meet_1): Rename to...
1609         (value_range::union_helper): ...this.
1610         (vrp_meet): Rename to...
1611         (value_range::union_): ...this.
1612         (copy_value_range): Remove.
1613         * tree-vrp.h (struct value_range): Rewrite into a proper class.
1614         (value_range::vrtype): New.
1615         (value_range::type): New.
1616         (value_range::equiv): New.
1617         (value_range::min): New.
1618         (value_range::max): New.
1619         (value_range::varying_p): New.
1620         (value_range::undefined_p): New.
1621         (value_range::null_p): New.
1622         (value_range::equiv_add): New.
1623         (copy_value_range): Remove.
1625 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
1627         * Makefile.in (SELFTEST_TARGETS): New.
1628         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
1629         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
1630         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
1631         c/Make-lang.in.
1632         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
1633         (selftest-c++-gdb, selftest-c++-valgrind): Move to
1634         cp/Make-lang.in.
1635         * configure: Regenerate.
1636         * configure.ac (selftest_languages): New.
1638 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1640         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
1641         overflow wraps argument.
1642         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
1643         wide_int_range_multiplicative_op.
1644         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
1645         overflow wraps argument.
1646         (wide_int_range_multiplicative_op): Same.
1647         (wide_int_range_lshift): Same.
1648         (wide_int_range_div): Same.
1649         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
1650         (wide_int_range_lshift): Same.
1651         (wide_int_range_div): Same.
1653 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1655         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
1656         use sign as argument.
1657         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
1658         wide_int_range_shift_undefined_p.
1660 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1662         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
1663         Rename to...
1664         (@despeculate_copy<ALLI_TI:mode>): ... This.
1665         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
1666         switch statement.
1668 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1670         * config.gcc: Obsolete *-*-solaris2.10*.
1671         * doc/install.texi (Specific, *-*-solaris2*): Document it.
1673 2018-10-12  Jeff Law  <law@redhat.com>
1675         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
1676         reg + sym +- const_int addressing modes.
1678 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1680         * common.opt (fdiagnostics-minimum-margin-width=): New option.
1681         * diagnostic-show-locus.c (layout::layout): Apply the minimum
1682         margin width.
1683         (layout::start_annotation_line): Only print up to 3 of the
1684         margin character, to avoid touching the left-hand side.
1685         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
1686         minimum margin width, as set by test_diagnostic_context's ctor.
1687         (selftest::test_fixit_insert_containing_newline): Likewise.
1688         (selftest::test_fixit_insert_containing_newline_2): Likewise.
1689         (selftest::test_line_numbers_multiline_range): Clear
1690         dc.min_margin_width.
1691         * diagnostic.c (diagnostic_initialize): Initialize
1692         min_margin_width.
1693         * diagnostic.h (struct diagnostic_context): Add field
1694         "min_margin_width".
1695         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
1696         * opts.c (common_handle_option): Handle
1697         OPT_fdiagnostics_minimum_margin_width_.
1698         * selftest-diagnostic.c
1699         (selftest::test_diagnostic_context::test_diagnostic_context):
1700         Initialize min_margin_width to 6.
1701         * toplev.c (general_init): Initialize global_dc->min_margin_width.
1703 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1705         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1706         Fix usage of "error_at_rich_loc" in the comment.
1708 2018-10-15  Renlin Li  <renlin.li@arm.com>
1710         PR target/87563
1711         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
1712         if-conversioned loop when it contains ifn with types not
1713         supported by backend.
1714         * internal-fn.c (expand_direct_optab_fn): Add an assert.
1715         (direct_internal_fn_supported_p): New helper function.
1716         * internal-fn.h (direct_internal_fn_supported_p): Declare.
1718 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
1720         PR target/87572
1721         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1722         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1723         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1725 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
1727         PR tree-optimization/87022
1728         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1729         bits in dist vector rather than the first one.
1731 2018-10-15  Richard Biener  <rguenther@suse.de>
1733         PR middle-end/87610
1734         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
1735         (visit_loadstore): When a used restrict tag escaped verify that
1736         the points-to solution of "other" pointers do not include
1737         escaped.
1738         (compute_dependence_clique): If a used restrict tag escaped
1739         communicated that down to visit_loadstore.
1741 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
1743         * config/s390/s390.c (s390_expand_vec_init): Force vector element
1744         into reg if it isn't a general operand.
1746 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1748         PR target/87599
1749         * config/i386/sse.md (*vec_dupv2di): Add register source to
1750         movddup.
1752 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1754         PR target/87572
1755         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1756         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
1757         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1758         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1760 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
1762         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
1763         (notice_args_size): Set it in the current trace if no insn that can
1764         throw internally has been seen yet.
1765         (connect_traces): When connecting args_size between traces, allow the
1766         incoming values not to match if there is an insn setting it before the
1767         first insn that can throw internally; in that case, force the creation
1768         of a CFI note on this latter insn.
1770 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
1772         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
1773         the base class, because using the injected-class-name was not clearly
1774         specified until DR 176.
1776 2018-10-12  Paul Koning  <ni1d@arrl.net>
1778         * config/pdp11/pdp11.md (doloop_end): New expander.
1779         (doloop_end_insn): renamed from "doloop_end".
1780         (addqi3): New pattern.
1781         (subqi3): New pattern.
1782         * config/pdp11/predicates.md (incdec_operand): New predicate.
1784 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
1786         PR middle-end/81376
1787         * real.c (format_helper::can_represent_integral_type_p): New function
1788         * real.h (format_helper::can_represent_integral_type_p): Ditto.
1789         * match.pd: New pattern.
1791 2018-10-12  Alexandre Oliva <oliva@adacore.com>
1793         * configure.ac: Introduce --enable-large-address-aware
1794         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1795         * doc/install.texi: Document it.
1796         * configure, config.in: Rebuilt.
1797         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1798         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1799         (LINK_SPEC): Insert it.
1800         * config/i386/mingw-264.h: Likewise.
1802         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1804 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
1806         PR rtl-optimization/87600
1807         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1809 2018-10-12  Paul Koning  <ni1d@arrl.net>
1811         * doc/md.texi (doloop_end): Document that the pattern code may
1812         need to check operand mode.
1814 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
1816         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1817         to zero-extend between int and floating-point registers.
1818         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1819         ldp into floating-point registers.  Add type and arch attributes.
1820         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1821         Use f_loads for type attribute.
1823 2018-10-11  Martin Sebor  <msebor@redhat.com>
1825         * doc/extend.texi (attribute packed): Correct typos.
1827 2018-10-11  Martin Sebor  <msebor@redhat.com>
1829         * doc/extend.texi (attribute flatten): Mention interaction with
1830         noinline.
1832 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1834         PR target/87156
1835         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1836         Set new_decl virtual flag to zero.
1838 2018-10-11  Martin Sebor  <msebor@redhat.com>
1840         PR middle-end/87593
1841         * doc/extend.texi (attribute format_arg): Discuss using multiple
1842         attributes on a single function.
1844 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1846         PR tree-optimization/86829
1847         * match.pd (sin (atan (x))): New simplification rules.
1848         (cos (atan (x))): Likewise.
1849         * real.c (build_sinatan_real): New function.
1850         * real.h (build_sinatan_real): Prototype.
1852 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
1854         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1855         function.
1856         (fold_mergeeo_helper): New helper function.
1857         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1858         intrinsics.  Correct some whitespace indentation issues.
1860 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
1862         PR target/87511
1863         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1864         Use HOST_WIDE_INT_1U for shift.
1866 2018-10-11  Doug Rupp  <rupp@adacore.com>
1867             Olivier Hainque  <hainque@adacore.com>
1869         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1870         Pass --relax to the linker for RTPs.
1871         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1873 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
1874             Jan Hubicka  <jh@suse.cz>
1875             Martin Jambor  <mjambor@suse.cz>
1877         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1878         the same elements are repeated rather than printing all of them.
1879         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1880         "repeated" elements.
1881         * read-rtl-function.c (test_loading_repeat): New function.
1882         (read_rtl_function_c_tests): Call test_loading_repeat.
1883         * rtl-tests.c (test_dumping_repeat): New function.
1884         (rtl_tests_c_tests): Call test_dumping_repeat.
1886 2018-10-11  Richard Biener  <rguenther@suse.de>
1888         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1889         bdver?_cost): Unify to ...
1890         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1891         * config/i386/i386.c (processor_cost_table): Adjust.
1893 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
1895         PR middle-end/87574
1896         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1897         the thunk when expanding to GIMPLE.
1899 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1901         * varasm.c (mergeable_string_section): Don't try to move zero-length
1902         strings to the merge section.
1904 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1906         PR target/87573
1907         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1909 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1911         PR target/87550
1912         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1913         to special_args set.
1915 2018-10-10  Richard Biener  <rguenther@suse.de>
1917         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1918         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1919         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1920         and recursing and pattern terminating the recursion on SSE
1921         vector width using ix86_expand_reduc.
1922         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1923         width and recursing and SSE2 vector variant doing the final
1924         reduction with ix86_expand_reduc.
1925         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1926         with terminating the recursion at AVX level, splitting that
1927         to SSE there.
1929 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1931         * genmatch.c (error_cb): Rename to...
1932         (diagnostic_cb): ...this, converting int params to enums.
1933         (fatal_at): Update for renaming.
1934         (warning_at): Likewise.
1935         (main): Likewise.
1936         * input.c (selftest::ebcdic_execution_charset::apply):
1937         Update for renaming of...
1938         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1939         to...
1940         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1941         converting level and reason to enums.
1942         (class selftest::lexer_error_sink): Rename to...
1943         (class selftest::lexer_test_options): ...this, renaming field
1944         "m_errors" to "m_diagnostics".
1945         (selftest::lexer_test_options::apply): Update for renaming of...
1946         (selftest::lexer_test_options::on_error): ...this, renaming to...
1947         (selftest::lexer_test_options::on_diagnostic): ...this
1948         converting level and reason to enums.
1949         (selftest::test_lexer_string_locations_raw_string_unterminated):
1950         Update for renamings.
1951         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1952         "reason".
1954 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
1956         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1957         * config/rs6000/pmmintrin.h: New file.
1959 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
1961         PR tree-optimization/86659
1962         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1964 2018-10-09  Richard Biener  <rguenther@suse.de>
1966         PR tree-optimization/63155
1967         * tree-ssa-structalias.c: Include tree-ssa.h.
1968         (get_constraint_for_ssa_var): For undefs return nothing_id.
1969         (find_func_aliases): Cleanup PHI handling.
1971 2018-10-09  Richard Biener  <rguenther@suse.de>
1973         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1974         replacements.
1976 2018-10-09  Martin Liska  <mliska@suse.cz>
1978         * asan.c (asan_emit_stack_protection): If a stack variable
1979         is located in a same file as current function, then emit
1980         line info into variable definition string.
1982 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1984         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1985         information.
1987 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1989         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1990         on the thunk.
1992 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1994         PR tree-optimization/86659
1995         * gimple-match.h (struct gimple_match_op): Add reverse field.
1996         (gimple_match_op::set_op): New overloaded method.
1997         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1998         the REF_REVERSE_STORAGE_ORDER flag on the value.
1999         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2000         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2001         
2002 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2004         PR middle-end/63155
2005         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2006         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2008 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2010         PR target/87517
2011         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2012         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2014 2018-10-08  Richard Biener  <rguenther@suse.de>
2016         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2017         cost the same as AVX128 ones.
2019 2018-10-08  Paul Koning  <ni1d@arrl.net>
2021         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2022         (expand_block_move): New function.
2023         * config/pdp11/pdp11.c (output_block_move): Remove.
2024         (expand_block_move): New function.
2025         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2026         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2027         (*movmemhi1): Remove.
2029 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2031         * config/s390/2827.md: Increase latencies for some FP instructions.
2033 2018-10-08  Richard Biener  <rguenther@suse.de>
2035         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2036         Open a dump scope.
2037         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2038         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2039         (add_stmt_cost): Dump return value of the hook.
2041 2018-10-08  Richard Biener  <rguenther@suse.de>
2043         PR tree-optimization/63155
2044         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2045         (ssa_propagation_engine::ssa_propagate): Remove redundant
2046         bitmap bit clearing.
2048 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2050         PR rtl-optimization/86939
2051         PR rtl-optimization/87479
2052         * ira.h (non_conflicting_reg_copy_p): New prototype.
2053         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2054         (make_hard_regno_dead): Don't add conflicts for register
2055         ignore_reg_for_conflicts.
2056         (make_object_dead): Likewise.
2057         (non_conflicting_reg_copy_p): New function.
2058         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2059         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2060         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2061         (make_hard_regno_dead): Don't add conflicts for register
2062         ignore_reg_for_conflicts.  Remove special conflict handling of
2063         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2064         check_pic_pseudo_p and update callers.
2065         (mark_pseudo_dead): Don't add conflicts for register
2066         ignore_reg_for_conflicts.
2067         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2069 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2070             Jim Wilson  <jimw@sifive.com>
2072         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2073         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2074         new pattern using HONOR_SNANS that emits one extra instruction.
2076 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2078         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2079         patterns): Merge SI and DI patterns to a GPR pattern.
2080         (unnamed define_insn and define_split for record form of that): Merge
2081         to a single define_insn_and_split pattern.
2083 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2085         PR c++/56856
2086         * input.c (expand_location_to_spelling_point): Add param "aspect"
2087         and use rather than hardcoding LOCATION_ASPECT_CARET.
2088         (get_substring_ranges_for_loc): Handle the case of a single token
2089         within a macro expansion.
2090         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2091         defaulting to LOCATION_ASPECT_CARET.
2093 2018-10-05  Paul Koning  <ni1d@arrl.net>
2095         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2096         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2097         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2098         (pdp11_guard_type): New function.
2100 2018-10-05  Paul Koning  <ni1d@arrl.net>
2102         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2103         * config/pdp11/pdp11.opt (mfloat32): Remove.
2104         (mfloat64): Remove.
2105         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2106         (pdp11 -mfloat64): Remove.
2108 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2110         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2111         (*cmp<mode>_cc_i387): Ditto.
2112         (*cmpu<mode>_cc_i387): Ditto.
2113         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2114         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2115         "scratch" argument.
2116         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2117         Emit x86_sahf_1 pattern.
2118         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2119         (ix86_expand_carry_flag_compare): Ditto.
2121 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2123         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2124         to reg_or_0_operand.  Add "C" constraint.
2125         (*cmpxf_cc_i387): Ditto.
2126         (*cmp<mode>_i387): Change operand 2 predicate
2127         to nonimm_or_0_operand.  Add "C" constraint.
2128         (*cmp<mode>_cc_i387): Ditto.
2129         (*cmp<mode>_0_i387): Remove insn pattern.
2130         (*cmp<mode>_0_cc_i387): Ditto.
2132 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2134         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2135         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2136         from vector_move_operand.  Update all uses.
2138 2018-10-05  Martin Sebor  <msebor@redhat.com>
2140         PR tree-optimization/87490
2141         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2142         consistently.
2144 2018-10-05  Richard Biener  <rguenther@suse.de>
2146         PR tree-optimization/63155
2147         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2148         vertical space in dumpfiles.
2149         * tree-ssa-propagate.h
2150         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2151         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2152         (ssa_edge_worklist_back): Likewise.
2153         (curr_order): Likewise.
2154         (cfg_blocks_get): Remove abstraction.
2155         (cfg_blocks_add): Likewise.
2156         (cfg_blocks_empty_p): Likewise.
2157         (add_ssa_edge): Add to current or next worklist based on
2158         RPO index.
2159         (add_control_edge): Likewise.
2160         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2161         into ...
2162         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2163         iteration from CFG and SSA edge worklist so we process
2164         everything in RPO order, prioritizing forward progress
2165         over iteration.
2166         (ssa_prop_init): Allocate new worklists, do not dump
2167         immediate uses.
2168         (ssa_prop_fini): Free new worklists.
2170 2018-10-05  Richard Biener  <rguenther@suse.de>
2172         * tree-core.h (tree_block::abstract_flag): Remove.
2173         (tree_block::block_num): Make full 32bits.
2174         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2175         * tree.h (BLOCK_ABSTRACT): Remove.
2176         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2177         resulting from BLOCK_ABSTRACT being always false.
2178         (gen_inlined_subroutine_die): Likewise.
2179         (gen_block_die): Likewise.
2180         * tree.c (block_ultimate_origin): Likewise.
2181         * tree-pretty-print.c (dump_block_node): Remove code dealing
2182         with BLOCK_ABSTRACT.
2183         * tree-ssa-live.c (dump_scope_block): Likewise.
2184         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2185         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2187 2018-10-05   Richard Biener  <rguenther@suse.de>
2189         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2190         is asked for initialize mode to the component mode of the
2191         vector type.
2193 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2195         PR target/87522
2196         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2197         assembler for -mavx.
2198         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2200 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2202         PR target/87509
2203         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2204         RS6000_BTM_DFP.
2205         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2206         to be DImode.  When using mffscrn, force the operand to a register.
2208 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2210         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2211         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2212         X87MODEF mode iterator.
2213         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2214         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2215         X87MODEF mode iterator.
2217 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2219         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2220         -Wno-prio-ctor-dtor.
2222 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2224         * Makefile.in (OBJS): Add opt-problem.o.
2225         * dump-context.h: Include "selftest.h.
2226         (selftest::temp_dump_context): New forward decl.
2227         (class dump_context): Make friend of class
2228         selftest::temp_dump_context.
2229         (dump_context::dump_loc_immediate): New decl.
2230         (class dump_pretty_printer): Move here from dumpfile.c.
2231         (class temp_dump_context): Move to namespace selftest.
2232         (temp_dump_context::temp_dump_context): Add param
2233         "forcibly_enable_dumping".
2234         (selftest::verify_dumped_text):
2235         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2236         (selftest::verify_item):
2237         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2238         (ASSERT_IS_TREE): Likewise.
2239         (ASSERT_IS_GIMPLE): Likewise.
2240         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2241         to...
2242         (dump_context::dump_loc_immediate): ...this new function.
2243         (class dump_pretty_printer): Move to dump-context.h.
2244         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2245         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2246         (temp_dump_context::temp_dump_context): Move to "selftest"
2247         namespace.  Add param "forcibly_enable_dumping", and use it to
2248         conditionalize the use of m_pp;
2249         (selftest::verify_dumped_text): Make non-static.
2250         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2251         (selftest::verify_item): Make non-static.
2252         (ASSERT_IS_TEXT): Move to dump-context.h.
2253         (ASSERT_IS_TREE): Likewise.
2254         (ASSERT_IS_GIMPLE): Likewise.
2255         (selftest::test_capture_of_dump_calls): Pass "true" for new
2256         param of temp_dump_context.
2257         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2258         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2259         TDF_COMPARE_DEBUG.
2260         * opt-problem.cc: New file.
2261         * opt-problem.h: New file.
2262         * optinfo-emit-json.cc
2263         (selftest::test_building_json_from_dump_calls): Pass "true" for
2264         new param of temp_dump_context.
2265         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2266         (optinfo::emit_for_opt_problem): New function.
2267         (optinfo::emit): Clarity which emit_item is used.
2268         * optinfo.h (optinfo::get_dump_location): New accessor.
2269         (optinfo::emit_for_opt_problem): New decl.
2270         (optinfo::emit): Make const.
2271         * selftest-run-tests.c (selftest::run_tests): Call
2272         selftest::opt_problem_cc_tests.
2273         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2274         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2275         bool to opt_result, converting fprintf messages to
2276         opt_result::failure_at calls.  Add "stmt" param for use by the
2277         failure_at calls.
2278         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2279         (runtime_alias_check_p): Convert return type from bool to
2280         opt_result, converting dump_printf calls to
2281         opt_result::failure_at, using the statement DDR_A for their
2282         location.
2283         (find_data_references_in_stmt): Convert return type from bool to
2284         opt_result, converting "return false" to opt_result::failure_at
2285         with a new message.
2286         * tree-data-ref.h: Include "opt-problem.h".
2287         (dr_analyze_innermost): Convert return type from bool to opt_result,
2288         and add a const gimple * param.
2289         (find_data_references_in_stmt): Convert return type from bool to
2290         opt_result.
2291         (runtime_alias_check_p): Likewise.
2292         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2293         dr_analyze_innermost.
2294         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2295         Convert return type from bool to opt_result, adding a message for
2296         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2297         (vect_analyze_data_ref_dependence): Convert return type from bool
2298         to opt_result.  Change sense of return type from "false"
2299         effectively meaning "no problems" to "false" meaning a problem,
2300         so that "return false" becomes "return opt_result::success".
2301         Convert "return true" calls to opt_result::failure_at, using
2302         the location of statement A rather than vect_location.
2303         (vect_analyze_data_ref_dependences): Convert return type from bool
2304         to opt_result.
2305         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2306         calls to opt_result::failure_at, using the stmt location rather
2307         than vect_location.
2308         (vect_verify_datarefs_alignment): Convert return type from bool
2309         to opt_result.
2310         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2311         into multiple more-tightly-scoped copies.
2312         (vect_analyze_data_refs_alignment): Convert return type from bool
2313         to opt_result.
2314         (vect_analyze_data_ref_accesses): Likewise, converting a
2315         "return false" to a "return opt_result::failure_at", adding a
2316         new message.
2317         (vect_prune_runtime_alias_test_list): Convert return type from
2318         bool to opt_result, converting dump_printf_loc to
2319         opt_result::failure_at.  Add a %G to show the pertinent statement,
2320         and use the stmt's location rather than vect_location.
2321         (vect_find_stmt_data_reference): Convert return type from
2322         bool to opt_result, converting dump_printf_loc to
2323         opt_result::failure_at, using stmt's location.
2324         (vect_analyze_data_refs):  Convert return type from bool to
2325         opt_result.  Convert "return false" to "return
2326         opt_result::failure_at", adding messages as needed.
2327         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2328         type from bool to opt_result.
2329         (vect_determine_vf_for_stmt): Likewise.
2330         (vect_determine_vectorization_factor): Likewise, converting
2331         dump_printf_loc to opt_result::failure_at, using location of phi
2332         rather than vect_location.
2333         (vect_analyze_loop_form_1): Convert return type from bool to
2334         opt_result, converting dump_printf_loc calls, retaining the use of
2335         vect_location.
2336         (vect_analyze_loop_form): Convert return type from loop_vec_info
2337         to opt_loop_vec_info.
2338         (vect_analyze_loop_operations): Convert return type from bool to
2339         opt_result, converting dump_printf_loc calls, using the location
2340         of phi/stmt rather than vect_location where available.  Convert
2341         various "return false" to "return opt_result::failure_at" with
2342         "unsupported phi" messages.
2343         (vect_get_datarefs_in_loop): Convert return type from bool to
2344         opt_result.  Add a message for the
2345         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2346         (vect_analyze_loop_2): Convert return type from bool to
2347         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2348         each "goto again;", adding new messages where needed.
2349         Add "unsupported grouped {store|load}" messages.
2350         (vect_analyze_loop): Convert return type from loop_vec_info to
2351         opt_loop_vec_info.
2352         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2353         bool to opt_result.
2354         * tree-vect-stmts.c (process_use): Likewise, converting
2355         dump_printf_loc call and using stmt location, rather than
2356         vect_location.
2357         (vect_mark_stmts_to_be_vectorized): Likeise.
2358         (vect_analyze_stmt): Likewise, adding a %G.
2359         (vect_get_vector_types_for_stmt): Convert return type from bool to
2360         opt_result, converting dump_printf_loc calls and using stmt
2361         location, rather than vect_location.
2362         (vect_get_mask_type_for_stmt): Convert return type from tree to
2363         opt_tree, converting dump_printf_loc calls and using stmt location.
2364         * tree-vectorizer.c: Include "opt-problem.h.
2365         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2366         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2367         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2368         enabled, use it to report at the top level "couldn't vectorize
2369         loop" followed by the problem.
2370         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2371         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2372         to opt_result.
2373         (vect_analyze_stmt): Likewise.
2374         (vect_get_vector_types_for_stmt): Likewise.
2375         (tree vect_get_mask_type_for_stmt): Likewise.
2376         (vect_analyze_data_ref_dependences): Likewise.
2377         (vect_enhance_data_refs_alignment): Likewise.
2378         (vect_analyze_data_refs_alignment): Likewise.
2379         (vect_verify_datarefs_alignment): Likewise.
2380         (vect_analyze_data_ref_accesses): Likewise.
2381         (vect_prune_runtime_alias_test_list): Likewise.
2382         (vect_find_stmt_data_reference): Likewise.
2383         (vect_analyze_data_refs): Likewise.
2384         (vect_analyze_loop): Convert return type from loop_vec_info to
2385         opt_loop_vec_info.
2386         (vect_analyze_loop_form): Likewise.
2387         (vect_analyze_slp): Convert return type from bool to opt_result.
2389 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2391         * doc/invoke.texi (-fopt-info): Document new "internals"
2392         sub-option.
2393         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2394         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2395         MSG_ALL_KINDS.
2396         (optinfo_verbosity_options): Add "internals".
2397         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2398         (dump_context::apply_dump_filter_p): New member function.
2399         (dump_context::dump_loc): Use apply_dump_filter_p rather than
2400         explicitly masking the dump_kind.
2401         (dump_context::begin_scope): Increment the scope depth first.  Use
2402         apply_dump_filter_p rather than explicitly masking the dump_kind.
2403         (dump_context::emit_item): Use apply_dump_filter_p rather than
2404         explicitly masking the dump_kind.
2405         (dump_dec): Likewise.
2406         (dump_hex): Likewise.
2407         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2408         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2409         (opt_info_switch_p): Update handling of default
2410         MSG_OPTIMIZED_LOCATIONS to cope with default of
2411         MSG_PRIORITY_USER_FACING.
2412         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2413         masking the dump_kind.
2414         (selftest::test_capture_of_dump_calls): Update test_dump_context
2415         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2416         than MSG_ALL.  Generalize scope test to be run at all four
2417         combinations of with/without MSG_PRIORITY_USER_FACING and
2418         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2419         for each of the two values.
2420         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2421         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
2422         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2423         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2424         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2425         with MSG_PRIORITY_*.
2426         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2427         dump messages as MSG_PRIORITY_USER_FACING.
2428         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2429         about the interaction with MSG_PRIORITY_*.
2431 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2433         * varasm.c (output_constant): Add new parameter merge_strings.
2434         Make strings properly zero terminated in merge string sections.
2435         (mergeable_string_section): Don't fail if the last char is non-zero.
2436         (assemble_variable_contents): Handle merge string sections.
2437         (assemble_variable): Likewise.
2438         (assemble_constant_contents): Likewise.
2439         (output_constant_def_contents): Likewise.
2440         (output_constructor_array_range,
2441         output_constructor_regular_field): Adjust call to output_constant.
2442         (output_object_block): Adjust call to assemble_constant_contents
2443         and assemble_variable_contents.
2445 2018-10-04  Martin Liska  <mliska@suse.cz>
2447         PR c/87483
2448         * cgraphunit.c (process_function_and_variable_attributes):
2449         Warn about a function with alias attribute and a body.
2451 2018-10-04  Martin Liska  <mliska@suse.cz>
2453         PR ipa/82625
2454         * multiple_target.c (redirect_to_specific_clone): New function.
2455         (ipa_target_clone): Use it.
2456         * tree-inline.c: Fix comment.
2458 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2460         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2461         fields.
2462         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2463         (gcc::dump_manager::register_pass): New member function, adapted
2464         from loop body in gcc::pass_manager::register_pass, adding a
2465         call to update_dfi_for_opt_info.
2466         (gcc::dump_manager::opt_info_enable_passes): Store the
2467         -fopt-info options into the new fields.  Move the loop
2468         bodies into...
2469         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2470         function.
2471         * dumpfile.h (struct opt_pass): New forward decl.
2472         (gcc::dump_manager::register_pass): New decl.
2473         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2474         (class gcc::dump_manager): Add fields "m_optgroup_flags",
2475         "m_optinfo_flags", and "m_optinfo_filename".
2476         * passes.c (gcc::pass_manager::register_pass): Move all of the
2477         dump-handling code to gcc::dump_manager::register_pass.
2479 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
2481         PR rtl-optimization/87466
2482         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2483         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2484         * doc/tm.texi: Regenerate.
2485         * ira-lives.c (process_bb_node_lives): Use the new target hook.
2486         * lra-lives.c (process_bb_lives): Likewise.
2487         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2488         Define.
2490 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
2492         * params.c (add_params): Fix initialization.
2494 2018-10-04  Martin Liska  <mliska@suse.cz>
2496         PR gcov-profile/84107
2497         * tree-profile.c (init_ic_make_global_vars):
2498         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2499         Come up with new ic_tuple* variables.  Emit
2500         __gcov_indirect_call{,_topn} variables.
2501         (gimple_gen_ic_profiler): Access the variable
2502         and emit gimple.
2503         (gimple_gen_ic_func_profiler): Access
2504         __gcov_indirect_call.callee field.
2505         (gimple_init_gcov_profiler): Use ptr_type_node.
2506         * value-prof.c (gimple_ic): Use ptr_type_node.
2508 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2510         PR tree-optimization/85787
2511         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
2512         into this function and add support for detecting multiple phis.
2513         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
2515 2018-10-04  Martin Liska  <mliska@suse.cz>
2517         PR ipa/87491
2518         * ipa-inline.c (inline_to_all_callers_1):
2519         Call ultimate_alias_target for node being inlined.
2521 2018-10-03  Jeff Law  <law@redhat.com>
2523         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
2524         target's wchar_t.
2525         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
2526         * tree.h (get_typenode_from_name): Prototype.
2528 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2530         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
2531         Change operand 2 predicate to nonimmediate_operand.
2532         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2534 2018-10-03  Martin Sebor  <msebor@redhat.com>
2535             Jeff Law  <law@redhat.com>
2537         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
2538         initialize it.
2539         (get_string_length): Detect unterminated arrays.
2540         (format_string): Same.
2541         (format_directive): Warn about unterminated arrays.
2542         (handle_gimple_call): Mark statements with no_warning as needed.
2544 2018-10-03  Jim Wilson  <jimw@sifive.com>
2546         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
2547         also define __riscv_abi_rve.  Delete trailing white space.
2549 2018-10-03  Paul Koning  <ni1d@arrl.net>
2551         Enable LRA register allocator for PDP11.
2552         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
2553         (R): Likewise.
2554         (D): Likewise.
2555         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
2556         * config/pdp11/pdp11.opt (-mlra): New option.
2557         * doc/invoke.texi (PDP-11 Options): Document -mlra.
2559 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2561         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
2562         (*<absneg:code>extend<mode>xf2): Ditto.
2564 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2566         PR tree-optimization/87415
2567         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
2568         precision fields.
2570 2018-10-02  Jeff Law  <law@redhat.com>
2572         * gimple-fold.c (get_range_strlen): Only set *nonstr when
2573         an unterminated string is discovered.  Bubble up range
2574         even for unterminated strings.
2575         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
2576         indicates the string was not terminated via NONSTR.
2578 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2580         * tree-vrp.c (extract_range_from_unary_expr): Special case all
2581         pointer conversions.
2582         Do not do anything special for anti-ranges.
2584 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
2586         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
2587         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
2589 2018-10-03  Martin Liska  <mliska@suse.cz>
2591         PR gcov-profile/86109
2592         * coverage.c (coverage_begin_function): Do not
2593         mark lambdas as artificial.
2594         * tree-core.h (struct GTY): Remove tm_clone_flag
2595         and introduce new lambda_function.
2596         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2598 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2600         PR target/87474
2601         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
2602         P8_VECTOR and VSX are enabled.
2604 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2606         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
2607         0x3907 as CPU model number.
2609 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2611         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
2612         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
2613         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
2614         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
2615         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
2616         * config/s390/s390.md: Likewise. Rename also the cpu attribute
2617         value from arch12 to z14.
2619 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2621         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
2622         (isinfxf2): Ditto.
2623         (isinf<mode>2): Ditto.
2625 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2627         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
2628         before emitting fxam.  Perform calculations in XFmode.
2630 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2632         * match.pd (((X /[ex] A) +- B) * A): New transformation.
2634 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2636         PR middle-end/87319
2637         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
2638         * tree.c (signed_or_unsigned_type_for): Handle complex.
2640 2018-10-02  Jeff Law  <law@redhat.com>
2642         * gimple-fold.c (get_range_strlen): Remove dead code.
2644 2018-10-02  Martin Sebor  <msebor@redhat.com>
2645             Jeff Law  <law@redhat.com>
2647         * builtins.c (unterminated_array): Add new arguments.
2648         If argument is not terminated, bubble up size and exact
2649         state to callers.
2650         (expand_builtin_strnlen): Detect, avoid expanding
2651         and diagnose unterminated arrays.
2652         (c_strlen): Fill in offset of start of unterminated strings.
2653         * builtins.h (unterminated_array): Update prototype.
2655 2018-10-02  Richard Biener  <rguenther@suse.de>
2657         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
2658         of haddv4df, first reduce to SSE width and exploit the fact
2659         that we only need element zero with the reduction result.
2660         (reduc_plus_scal_v2df): Likewise.
2662 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2664         * dojump.h (do_jump): Delete.
2665         (do_jump_1): Likewise.
2666         (split_comparison): Move around.
2667         * dojump.c (do_jump): Make static.
2668         (do_jump_1): Likewise.
2669         (jumpifnot): Move around.
2670         (jumpifnot_1): Likewise.
2671         (jumpif): Likewise.
2672         (jumpif_1): Likewise.
2673         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
2675 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2677         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
2678         insns in the delay slot and add_insn_after for the jump insn.
2680 2018-10-02  Richard Biener  <rguenther@suse.de>
2682         * tree-inline.c (expand_call_inline): Use the location of
2683         the callee declaration for the inline-entry marker.
2684         * final.c (notice_source_line): Remove special-casing of
2685         NOTE_INSN_INLINE_ENTRY.
2687 2018-10-01  Carl Love  <cel@us.ibm.com>
2689         PR 69431
2690         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
2691         (__builtin_mtfsb0): New.
2692         (__builtin_mtfsb1): New.
2693         ( __builtin_set_fpscr_rn): New.
2694         (__builtin_set_fpscr_drn): New.
2695         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
2696         (rs6000_expand_set_fpscr_rn_builtin): Add.
2697         (rs6000_expand_set_fpscr_drn_builtin): Add.
2698         (rs6000_expand_builtin): Add case statement entries for
2699         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
2700         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
2701         RS6000_BUILTIN_MFFSL.
2702         (rs6000_init_builtins): Add ftype initialization and def_builtin
2703         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
2704         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
2705         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
2706         rs6000_mffscdrn): Add define_insn.
2707         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
2708         * doc/extend.texi: Add documentation for the builtins.
2710 2018-10-01  Richard Biener  <rguenther@suse.de>
2712         PR tree-optimization/87465
2713         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2714         causing branch miscounts.
2716 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2718         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
2719         aarch64_option_default_param):  New.
2720         (params.h): Include.
2721         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
2722         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
2723         stack-clash protection validation code.
2725 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2727         * params.c (validate_param): New.
2728         (add_params): Use it.
2729         (set_param_value): Refactor param validation into validate_param.
2730         (diagnostic.h): Include.
2731         * diagnostic.h (diagnostic_ready_p): New.
2733 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2735         * params.c (set_param_value):
2736         Add index of parameter being validated.
2737         * common/common-target.def (option_validate_param): New.
2738         * common/common-targhooks.h (default_option_validate_param): New.
2739         * common/common-targhooks.c (default_option_validate_param): New.
2740         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
2741         * doc/tm.texi: Regenerate.
2743 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2745         PR target/86486
2746         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2747         Add validation for stack-clash parameters and set defaults.
2749 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2751         PR target/86486
2752         * configure.ac: Add stack-clash-protection-guard-size.
2753         * doc/install.texi: Document it.
2754         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
2755         * params.def: Update comment for guard-size.
2756         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
2757         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
2758         * configure: Regenerate.
2760 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2762         PR target/86486
2763         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
2764         STACK_DYNAMIC_OFFSET): New.
2765         * config/aarch64/aarch64.c (aarch64_layout_frame):
2766         Update outgoing args size.
2767         (aarch64_stack_clash_protection_alloca_probe_range,
2768         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2770 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2772         PR target/86486
2773         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
2774         probe ranges.
2775         * target.def (stack_clash_protection_alloca_probe_range): New.
2776         (stack_clash_protection_final_dynamic_probe): Remove.
2777         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
2778         (default_stack_clash_protection_final_dynamic_probe): Remove.
2779         * targhooks.c: Likewise.
2780         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2781         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
2782         * doc/tm.texi: Regenerate.
2784 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2786         PR target/86486
2787         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
2788         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2789         aarch64_clamp_to_uimm12_shift): New.
2790         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2791         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2793 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2795         PR target/86486
2796         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2798 2018-10-01  Jeff Law  <law@redhat.com>
2799             Richard Sandiford <richard.sandiford@linaro.org>
2800             Tamar Christina  <tamar.christina@arm.com>
2802         PR target/86486
2803         * config/aarch64/aarch64.md
2804         (probe_stack_range): Add k (SP) constraint.
2805         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2806         STACK_CLASH_MAX_UNROLL_PAGES): New.
2807         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2808         stack probes for stack clash.
2809         (aarch64_allocate_and_probe_stack_space): New.
2810         (aarch64_expand_prologue): Use it.
2811         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2812         (aarch64_sub_sp): Add emit_move_imm optional param.
2814 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
2816         PR tree-optimization/87261
2817         * match.pd: Remove trailing whitespace.
2818         Add (x & y) | ~(x | y) -> ~(x ^ y),
2819         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2821 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2823         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2824         constraints.
2825         (ashlsi3_insn): Update instruction constraints.
2826         (ashrsi3_insn): Likewise.
2827         (rotrsi3): Likewise.
2828         (add_shift): Likewise.
2829         * config/arc/constraints.md (Csz): New 32 bit constraint. It
2830         avoids placing in the limm field small constants which, otherwise,
2831         could end into a small instruction.
2833 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2835         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2836         destination register is not odd-even.
2837         (umaddsidi4_split): Likewise.
2839 2018-10-01  Richard Biener  <rguenther@suse.de>
2841         * tree-inline.c (expand_call_inline): Store origin of fn
2842         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2843         * tree.c (block_ultimate_origin): Simplify and do some
2844         checking.
2846 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
2848         * config/i386/mmx.md (EMMS): New int iterator.
2849         (emms): New int attribute.
2850         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2851         EMMS int iterator.  Explicitly declare clobbers.
2852         (mmx_emms): Remove expander.
2853         (mmx_femms): Ditto.
2854         * config/i386/predicates.md (emms_operation): Remove predicate.
2855         (vzeroall_pattern): New predicate.
2856         (vzeroupper_pattern): Rename from vzeroupper_operation.
2857         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2858         vzeroupper_pattern and vzeroall_pattern predicates.
2860 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
2862         PR rtl-optimization/86939
2863         * ira-lives.c (make_hard_regno_born): Rename from this...
2864         (make_hard_regno_live): ... to this.  Remove update to conflict
2865         information.  Update function comment.
2866         (make_hard_regno_dead): Add conflict information update.  Update
2867         function comment.
2868         (make_object_born): Rename from this...
2869         (make_object_live): ... to this.  Remove update to conflict information.
2870         Update function comment.
2871         (make_object_dead):  Add conflict information update.  Update function
2872         comment.
2873         (mark_pseudo_regno_live): Call make_object_live.
2874         (mark_pseudo_regno_subword_live): Likewise.
2875         (mark_hard_reg_dead): Update function comment.
2876         (mark_hard_reg_live): Call make_hard_regno_live.
2877         (process_bb_node_lives): Likewise.
2878         * lra-lives.c (make_hard_regno_born): Rename from this...
2879         (make_hard_regno_live): ... to this.  Remove update to conflict
2880         information.  Remove now uneeded check_pic_pseudo_p argument.
2881         Update function comment.
2882         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2883         to conflict information.  Update function comment.
2884         (mark_pseudo_live): Remove update to conflict information.  Update
2885         function comment.
2886         (mark_pseudo_dead): Add conflict information update.
2887         (mark_regno_live): Call make_hard_regno_live.
2888         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2889         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2891 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
2893         PR target/87370
2894         * config/i386/i386.c (construct_container): Use TImode for
2895         BLKmode values in 2 integer registers.
2897 2018-09-29  Jeff Law  <law@redhat.com>
2899         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2900         c_strlen rather than just a tree *.
2901         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2902         Update recursive calls appropriately.  If caller did not provide a
2903         suitable data pointer, create a local one.  When a non-terminated
2904         string is discovered, bubble up information about the string via the
2905         c_strlen_data object.
2906         * builtins.h (c_strlen): Update prototype.
2907         (c_strlen_data): New structure.
2908         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2909         For a type 2 call, if c_strlen indicates a non-terminated string
2910         use the length of the non-terminated string.
2911         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2913 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2915         PR target/87467
2916         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2917         __m512d type for __A argument rather than __m512.
2919 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2921         * match.pd (simple_comparison): Don't optimize if either operand is
2922         a function pointer when target needs function pointer canonicalization.
2924 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2926         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2927         power5 .. power9 to remove indirection.
2928         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2929         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2930         ASM_CPU_476_SPEC): Delete.
2931         (ASM_CPU_SPEC): Adjust.
2932         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2933         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2935 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2937         * config.in: Delete HAVE_AS_DCI.
2938         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2939         * config/rs6000/rs6000.h: Ditto.
2940         * configure.ac: Delete HAVE_AS_DCI.
2941         * configure: Regenerate.
2943 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2945         * config.in (HAVE_AS_LWSYNC): Delete.
2946         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2947         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2948         do it as a .long .
2949         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2950         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2951         as a .long .
2952         * configure.ac: Delete HAVE_AS_LWSYNC.
2953         * configure: Regenerate.
2955 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2956             Pierre-Marie de Rodat  <derodat@adacore.com>
2958         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2959         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2960         (cgraph_node::create_thunk): Add indirect_offset parameter.
2961         (thunk_adjust): Likewise.
2962         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2963         and initialize the corresponding field with it.
2964         (cgraph_node::dump): Dump indirect_offset field.
2965         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2966         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2967         (thunk_adjust): Add indirect_offset parameter and deal with it.
2968         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2969         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
2970         or if the thunk is external or local.  Fix formatting.  Do not chain
2971         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
2972         if any, in the GIMPLE representation.
2973         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2974         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2975         (input_node): Read indirect_offset field.
2976         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2977         call to thunk_adjust.
2978         * tree-nested.c (struct nesting_info): Add thunk_p field.
2979         (create_nesting_tree): Set it.
2980         (convert_all_function_calls): Copy static chain from targets to thunks.
2981         (finalize_nesting_tree_1): Return early for thunks.
2982         (unnest_nesting_tree_1): Do not finalize thunks.
2983         (gimplify_all_functions): Do not gimplify thunks.
2985 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2987         * opt-suggestions.c (option_proposer::build_option_suggestions):
2988         Release "option_values".
2990 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2992         * coverage.c (get_coverage_counts): Convert problem-reporting dump
2993         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2994         * dumpfile.c (kind_as_string): New function.
2995         (dump_loc): Rather than a hardcoded prefix of "note: ", use
2996         kind_as_string to vary the prefix based on dump_kind.
2997         (selftest::test_capture_of_dump_calls): Update for above.
2999 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3001         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3002         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3004 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3006         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3007         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3008         INVALID_REGNUM instead of FPSR_REG.
3009         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3010         * config/i386/i386.md: Update comment of FP compares.
3011         (fldenv): Do not clobber FPSR_REG.
3013 2018-09-28  Richard Biener  <rguenther@suse.de>
3015         * tree.h (BLOCK_ORIGIN): New.
3016         * omp-expand.c (grid_expand_target_grid_body): Assign
3017         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3018         * tree-inline.c (remap_block): Likewise.
3019         * auto-profile.c (get_function_decl_from_block): Simplify
3020         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3021         * langhooks.c (lhd_print_error_function): Likewise.
3022         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3023         Likewise.
3024         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3025         * tree.c (block_nonartificial_location): Likewise.
3026         (block_ultimate_origin): Likewise.
3027         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3028         no longer needed LTO case.
3030 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3031             Jan Hubicka  <jh@suse.cz>
3032             Martin Jambor  <mjambor@suse.cz>
3034         * simplify-rtx.c (simplify_merge_mask): New function.
3035         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3036         same masks are used in op1 or op2.
3037         (test_vec_merge): New function.
3038         (test_vector_ops): Call test_vec_merge.
3040 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3042         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3043         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3044         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3045         (cypress_costs): Set it.
3046         (supersparc_costs): Likewise.
3047         (hypersparc_costs): Likewise.
3048         (leon_cost): Likewise.
3049         (leon3_costs): Likewise.
3050         (sparclet_costs): Likewise.
3051         (ultrasparc_costs): Likewise.
3052         (ultrasparc_costs): Likewise.
3053         (niagara_costs): Likewise.
3054         (niagara2_costs): Likewise.
3055         (niagara3_costs): Likewise.
3056         (niagara4_costs): Likewise.
3057         (niagara7_costs): Likewise.
3058         (m8_costs): Likewise.
3059         (TARGET_CAN_FOLLOW_JUMP): Define.
3060         (pass_work_around_errata::gate): Minor tweak.
3061         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3062         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3063         Automaitcally clear MASK_FSMULD mask for V7 processors.
3064         (sparc_can_follow_jump): New static function.
3065         (output_ubranch): Deal with CROSSING_JUMP_P.
3066         (sparc_use_sched_lookahead): Rewrite using switch statement.
3067         (sparc_issue_rate): Reorder.
3068         (sparc_branch_cost): New function.
3070 2018-09-27  Martin Sebor  <msebor@redhat.com>
3072         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3073         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3074         (int_fits_type_p): Same.
3076 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3078         * config/i386/i386.md (FPCR_REG): Remove.
3079         (UNSPEC_FLDCW): Remove.
3080         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3081         (x86_fldcw_1): Remove insn pattern.
3082         (fnstenv): Do not clobber FPCR_REG.
3083         (fldenv): Ditto.
3084         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3085         (CALL_USED_REGISTERS): Ditto.
3086         (REG_ALLOC_ORDER): Ditto.
3087         (REG_CLASS_CONTENTS): Ditto.
3088         (HI_REGISTER_NAMES): Ditto.
3089         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3090         of numerical constants.
3091         * config/i386/i386.c (regclass_map): Remove fpsr register.
3092         (dbx_register_map): Ditto.
3093         (dbx64_register_map): Ditto.
3094         (svr4_dbx_register_map): Ditto.
3095         (print_reg): Do not handle FPCR_REG.
3097 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3099         PR target/87149
3100         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3101         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3102         Delete, always treat as true.
3103         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3104         Ditto.  Simplify remaining code.
3105         * config/powerpcspe/powerpcspe.h: Ditto.
3106         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3107         Simplify remaining code.
3108         (rs6000_expand_builtin): Ditto.
3109         * config/rs6000/rs6000.h: Ditto.
3110         * configure.ac: Ditto.
3111         * configure: Regenerate.
3113 2018-09-27  Martin Liska  <mliska@suse.cz>
3115         * coverage.c (get_coverage_counts): Revert the formatting
3116         of missing profile opt info.
3118 2018-09-27  Richard Biener  <rguenther@suse.de>
3120         PR debug/37801
3121         PR debug/87440
3122         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3123         block as we do not output that.
3124         (gen_inlined_subroutine_die): Elide the originally outermost
3125         block, matching what we do for concrete instances.
3126         (decls_for_scope): Add parameter specifying whether to recurse
3127         to subblocks.
3129 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3130             Tom de Vries  <tom@codesourcery.com>
3132         PR 82089
3134         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3135         STORE_FLAG_VALUE == 1.
3137 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3139         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3140         constant definitions.
3141         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3142         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3143         ("speculation_barrier"): New expander definition.
3145 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3147         PR gcov-profile/86957
3148         * common.opt: New warning option -Wmissing-profile.
3149         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3150         * doc/invoke.texi: Document -Wmissing-profile.
3152 2018-09-26  Jim Wilson  <jimw@sifive.com>
3154         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3155         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3157 2018-09-26  Martin Sebor  <msebor@redhat.com>
3159         * tree.c (zerop): Change return type to bool.
3160         (integer_zerop, integer_onep, integer_each_onep): Same.
3161         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3162         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3163         (real_onep, real_minus_onep, chain_index): Same.
3164         (print_type_hash_statistics, type_list_equal): Same.
3165         * tree.h (zerop): Same.
3166         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3167         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3168         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3169         (real_onep, real_minus_onep, chain_index): Same.
3170         (print_type_hash_statistics, type_list_equal): Same.
3172 2018-09-26  Jim Wilson  <jimw@sifive.com>
3174         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3176 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3178         PR target/87414
3179         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3180         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3181         call.
3183 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3185         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3187 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3189         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3190         and FP_SECOND_SSE_REGS.
3191         (REG_CLASS_NAMES): Ditto.
3192         (REG_CLASS_CONTENTS): Ditto.
3193         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3194         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3195         (ix86_preferred_output_reload_class): Ditto.
3196         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3197         clobber constraint to "=&f".
3198         (fix_truncdi_i387): Ditto.
3199         (lrintxfdi2): Ditto.
3200         (fistdi2_<rounding>): Ditto.
3201         (fpremxf4_i387): Change "=u" constraint to "=f".
3202         (fprem1xf4_i387): Ditto.
3203         (sincosxf3): Ditto.
3204         (fptanxf4_i387): Ditto.
3205         (fxtractxf3_i387): Ditto.
3206         (fscalexf4_i387): Ditto.
3207         (atan2xf3): Change "u" constraint to "f".
3208         (fyl2xxf3_i387): Ditto.
3209         (fyl2xp1xf3_i387): Ditto.
3211 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3213         PR target/87439
3214         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3215         for removed I387_MASK_PM entity.
3218 2018-09-26  Jeff Law  <law@redhat.com>
3219         Revert
3220         2018-09-26  Alexey Neyman  <stilor@att.net>
3222         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3223         headers are no longer pulled in by <isl/val.h>.
3225 2018-09-26  Richard Biener  <rguenther@suse.de>
3227         PR debug/87443
3228         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3229         or concrete instance DIE to the tree.  Create abstract origin
3230         attributes also for concrete instances.
3232 2018-09-26  Alexey Neyman  <stilor@att.net>
3234         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3235         headers are no longer pulled in by <isl/val.h>.
3237 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3239         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3240         Use new helper functions.
3241         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3242         Use new helper functions.
3243         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3244         aarch_mm_needs_release): New declarations.
3245         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3246         aarch_mm_needs_release): New.
3248 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3250         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3251         (arm32_output_mi_thunk): Deal with long calls.
3253 2018-09-26  Richard Biener  <rguenther@suse.de>
3255         PR debug/87428
3256         PR debug/87362
3257         * tree-inline.c (expand_call_inline): When the location
3258         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3259         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3260         the inserted BLOCK to make inlined_function_outer_scope_p
3261         recognize it.
3262         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3263         coords for reserved locations.
3265 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3267         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3268         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3269         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3270         (*sibcall_nonlocal_sysv<mode>): Ditto.
3271         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3272         (<bd>_<mode>): Ditto.
3273         (<bd>tf_<mode>): Ditto.
3275 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3277         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3278         control string as a list of templates instead of as C code.
3279         (*altivec_movti): Ditto.
3280         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3282 2018-09-25  Jim Wilson  <jimw@sifive.com>
3284         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3285         when target symbol is weak.
3287 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3289         PR c/87387
3290         * builtins.c (unterminated_array): Simplify.
3291         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3292         where pointer arithmetic is safe.
3294 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3296         PR target/86987
3297         * config/rs6000/altivec.md (altivec_vspltb): Use
3298         const_0_to_15_operand instead of u5bit_cint_operand.
3299         (*altivec_vspltb_internal): Ditto.
3300         (altivec_vspltb_direct): Ditto.
3301         (altivec_vsplth): Use const_0_to_7_operand instead of
3302         u5bit_cint_operand.
3303         (*altivec_vsplth_internal): Ditto.
3304         (altivec_vsplth_direct): Ditto.
3305         (altivec_vspltw): Use const_0_to_3_operand instead of
3306         u5bit_cint_operand.
3307         (*altivec_vspltw_internal): Ditto.
3308         (altivec_vspltw_direct): Ditto.
3309         (altivec_vspltsf): Ditto.
3310         (*altivec_vspltsf_internal): Ditto.
3311         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3312         various splats with the proper size immediate.  Reorder the various
3313         cases by ascending size of immediate, and put all such together.
3315 2018-09-25  Richard Biener  <rguenther@suse.de>
3317         PR debug/83941
3318         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3319         GC-ification.
3320         (maybe_create_die_with_external_ref): Do not create
3321         DW_TAG_imported_unit here.
3322         (add_abstract_origin_attribute): Handle external BLOCK refs.
3323         (dwarf2out_abstract_function): Simplify LTO case.
3324         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3325         rather than using maybe_create_die_with_external_ref.
3327 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3329         PR target/71278
3330         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3331         (frndintxf2_mask_pm_i387): Ditto.
3332         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3333         Enable for !flag_trapping_math.
3334         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3335         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3336         Change operand 1 predicate to nonimmediate_operand.
3337         (attr "i387_cw"): Remove mask_pm.
3338         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3339         (enum ix86_entity): Remove I387_MASK_PM.
3340         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3341         handle I387_MASK_PM.
3342         (ix86_mode_needed): Ditto.
3343         (ix86_mode_after): Ditto.
3344         (ix86_mode_entry): Ditto.
3345         (ix86_mode_exit): Ditto.
3346         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3348 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3350         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3351         to_update_switch_stmts to vNULL instead of calling create on them
3352         immediately.
3354 2018-09-25  Richard Biener  <rguenther@suse.de>
3356         PR tree-optimization/87402
3357         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3358         (visit_phi): Re-instantiate handling of supposed to be VARYING
3359         but non-VARYING backedge value.
3361 2018-09-25  Richard Biener  <rguenther@suse.de>
3363         PR debug/83941
3364         * dwarf2out.c (struct sym_off_pair): New.
3365         (external_die_map): New global.
3366         (lookup_decl_die): When in LTO create DIEs lazily from the
3367         external_die_map.
3368         (lookup_block_die): New function, create DIEs lazily in LTO.
3369         (equate_block_to_die): New function.
3370         (dwarf2out_die_ref_for_decl): During WPA get the association
3371         from the external DIE map.
3372         (dwarf2out_register_external_die): Record mapping into the
3373         external DIE map.
3374         (maybe_create_die_with_external_ref): New function split out from
3375         DIE generation part of old dwarf2out_register_external_die.
3376         (add_abstract_origin_attribute): Do not return the DIE.  When
3377         in LTO reference externals directly.
3378         (dwarf2out_abstract_function): When in LTO ignore calls for
3379         decls with external DIEs (already present abstract instances).
3380         (gen_call_site_die): Adjust.
3381         (add_high_low_attributes): Likewise.
3382         (gen_lexical_block_die): Likewise.
3383         (gen_inlined_subroutine_die): Likewie.
3384         (gen_block_die): Likewise.
3385         (dwarf2out_inline_entry): Likewise.
3386         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3387         DIEs.
3389 2018-09-25  Martin Liska  <mliska@suse.cz>
3391         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3392         integers and not by a float value.
3394 2018-09-25  Martin Liska  <mliska@suse.cz>
3396         PR fortran/87394
3397         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3398         instead of NULL.
3399         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3400         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3401         * godump.c (go_format_type): Remove extra parenthesis.
3403 2018-09-25  Martin Liska  <mliska@suse.cz>
3405         * alias.c (set_dest_equal_p): Remove unused function.
3406         * config/i386/i386.c (def_builtin_pure2): Likewise.
3407         * diagnostic-show-locus.c (class layout): Remove
3408         unused field.
3409         (layout::layout): Likewise here.
3410         * dump-context.h (class temp_dump_context): Likewise.
3411         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3412         (add_AT_loclistsptr): Likewise.
3413         (add_AT_offset): Likewise.
3414         (get_AT_hi_pc): Likewise.
3415         (is_comdat_die): Likewise.
3416         (type_is_enum): Likewise.
3417         (ceiling): Likewise.
3418         (add_AT_vms_delta): Likewise.
3419         (is_class_die): Likewise.
3420         * edit-context.c (class line_event): Remove unused field.
3421         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3422         unused function.
3423         * ipa-cp.c (ipa_get_vr_lat): Likewise.
3424         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3425         (ok_for_base_p_nonstrict): Likewise.
3426         * tree-chrec.c (is_not_constant_evolution): Likewise.
3427         (chrec_fold_poly_cst): Likewise.
3428         * tree-if-conv.c (has_pred_critical_p): Likewise.
3429         * tree-ssa-coalesce.c (print_exprs): Likewise.
3430         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3431         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3432         * tree-vrp.c (value_ranges_intersect_p): Likewise.
3433         (value_range_nonnegative_p): Likewise.
3435 2018-09-25  Martin Liska  <mliska@suse.cz>
3437         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3438         Do not handle "GNU Pascal".
3439         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3440         Likewise.
3441         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3442         from documentation. Likewise.
3443         * dbxout.c (dbxout_range_type): Likewise.
3444         * doc/cpp.texi: Likewise.
3445         * doc/extend.texi: Likewise.
3446         * doc/frontends.texi: Likewise.
3447         * doc/invoke.texi: Remove Pascal entry.
3448         * tree.def (CLEANUP_POINT_EXPR): Likewise.
3449         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3451 2018-09-25  Martin Liska  <mliska@suse.cz>
3453         PR middle-end/86078
3454         * doc/invoke.texi: Document all parameters and remove default
3455         of the parameters.
3457 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3459         PR bootstrap/87417
3460         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3461         contains HOST_WIDE_INTs when computing its size.
3463 2018-09-24  Jim Wilson  <jimw@sifive.com>
3465         PR target/87391
3466         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3467         not TARGET_RVE.
3468         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3470 2018-09-24  Andrew Pinski  <apinski@marvell.com>
3472         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3473         access prev before checking it for NULLness in the
3474         AARCH64_FUSE_CMP_BRANCH case.
3476 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
3478         PR target/82699
3479         * config/i386/i386.c (rest_of_insert_endbranch): Set
3480         endbr_queued_at_entrance to true and don't insert ENDBR if
3481         x86_function_profiler will be called.
3482         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3483         is true.
3484         * config/i386/i386.h (machine_function): Add
3485         endbr_queued_at_entrance.
3487 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3489         * genattrtab.c (mk_attr_alt): Use alternative_mask.
3490         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3491         types.
3492         (check_attr_test): Use alternative_mask.
3493         (get_attr_value): Likewise.
3494         (compute_alternative_mask): Use alternative_mask and XWINT.
3495         (make_alternative_compare): Use alternative_mask.
3496         (attr_alt_subset_p): Use XWINT.
3497         (attr_alt_subset_of_compl_p): Likewise.
3498         (attr_alt_intersection): Use alternative_mask and XWINT.
3499         (attr_alt_union): Likewise.
3500         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3501         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3502         (simplify_test_exp): Use alternative_mask and XWINT.
3503         (write_test_expr): Use alternative_mask and XWINT, adjust bit
3504         number calculation to support 64 bits.  Generate code that
3505         checks 64-bit masks.
3506         (main): Use alternative_mask.
3507         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
3509 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3511         PR target/80080
3512         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
3513         RETURN+USE when returning via %r14.
3515 2018-09-24  Martin Liska  <mliska@suse.cz>
3517         * gcov.c (output_lines): Print colorization legend
3518         for both flag_use_colors and flag_use_hotness_colors.
3519         Reword the help.
3521 2018-09-24  Martin Liska  <mliska@suse.cz>
3523         * coverage.c (get_coverage_counts): Use warning_at
3524         with current_function_decl location. Use %qD in warning
3525         message.
3527 2018-09-24  Martin Liska  <mliska@suse.cz>
3529         * memory-block.h (memory_block_pool::release): Annotate with
3530         valgrind that the memory is not accessible.
3532 2018-09-24  Martin Liska  <mliska@suse.cz>
3534         PR sanitizer/85774
3535         * asan.c: Make asan_handled_variables extern.
3536         * asan.h: Likewise.
3537         * cfgexpand.c (expand_stack_vars): Make sure
3538         a representative is unpoison if another
3539         variable in the partition is handled by
3540         use-after-scope sanitization.
3542 2018-09-24  Richard Biener  <rguenther@suse.de>
3544         PR tree-optimization/63155
3545         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
3546         the worklist when the edge of the respective argument isn't
3547         executable.
3549 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3551         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
3552         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
3553         (MASK_CLASS_P): Update for rename.
3554         (MAYBE_MASK_CLASS_P): Ditto.
3555         (REG_CLASS_NAMES): Update.
3556         (REG_CLASS_CONTENT): Update.
3557         * config/i386/i386.c (regclass_map): Update for MASK_REG
3558         and ALL_MASK_REGS rename.
3559         * config/i386/constraints.md (Yk): Update for rename.
3560         (k): Ditto.
3562 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3564         * config/i386/i386.h (enum reg_class): Remove
3565         EVEX_SSE_REGS and MOD4_SSE_REGS.
3566         (REG_CLASS_NAMES): Update.
3567         (REG_CLASS_CONTENT): Update.
3568         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
3569         registers as ALL_SSE_REGS.
3570         (ix86_additional_allocno_class_p): Remove.
3571         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
3572         (ix86_register_priority): Lower priority of EVEX SSE registers.
3573         Use IN_RANGE macro where appropriate.
3574         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
3575         AVX-5124VNNIW checks.
3576         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
3577         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
3578         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
3579         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
3580         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
3581         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
3582         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
3583         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
3584         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
3585         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
3586         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
3587         * config/i386/constraints.md (Yh): Remove.
3589 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3591         * config/i386/i386.c (regclass_map): Declare integer REX registers
3592         as GENERAL_REGS.
3594 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
3596         * doc/service.texi (Service): Switch the fsf.org link to https.
3598 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3600         PR target/86798
3601         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3602         Define to speculation_safe_value_not_needed.
3604 2018-09-21  Florian Weimer  <fweimer@redhat.com>
3606         PR middle-end/81035
3607         * doc/extend.texi (Common Function Attributes): Mention that
3608         noreturn suppresses tail call optimization.
3610 2018-09-21  Jeff Law  <law@redhat.com>
3612         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
3613         vr_values::cleanup_edges_and_switches.
3614         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
3615         vr_values class.
3616         (identify_jump_threads): Remove EDGE_IGNORE handling.
3617         (execute_vrp): Move handling of to_remove_edges and
3618         to_update_switch_stmts into vr_values class member functions.
3619         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
3620         (to_update_switch_stmts): Likewise.
3621         * vr-values.c: Include cfghooks.h.
3622         (vr_values::vr_values): Initialize to_remove_edges and
3623         to_update_switch_stmts.
3624         (vr_values::~vr_values): Verify to_remove_edges and
3625         to_update_switch_stmts are empty.
3626         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
3627         (vr_values::cleanup_edges_and_switches): New member function.
3628         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
3629         function.  Add new data members.
3631 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
3633         PR tree-optimization/87309
3634         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
3635         calls with pflags and alt_flags.
3636         (selftest::test_capture_of_dump_calls): Add test of interaction of
3637         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
3639 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3641         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
3643 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3645         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
3647 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3649         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
3650         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
3651         (VXWORKS_LIBS_RTP): Minor reordering.
3653 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3655         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
3656         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
3658 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3660         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
3661         (PTRDIFF_TYPE): Likewise.
3663 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3665         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
3666         triplet, similar to support for VxWorks7.
3667         * config/vxworks-dummy.h: Provide a default definition
3668         of TARGET_VXWORKS64 to 0.
3670 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3672         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
3673         * config/vxworks-dummy.h: here.
3675 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3677         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
3679 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
3680             Bo Zhou  <zbo.zhou@hisilicon.com>
3682         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
3683         * config/aarch64/aarch64-tune.md: Regenerated.
3684         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
3685         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
3686         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
3688 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
3689             Julian Brown  <julian@codesourcery.com>
3691         * builtins.c (get_builtin_sync_mem): Handle address spaces.
3693 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3695         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
3696         if the call takes a static chain.
3698 2018-09-21  Martin Liska  <mliska@suse.cz>
3700         * auto-profile.c (autofdo_source_profile::read): Do not
3701         set sum_all.
3702         (read_profile): Do not add working sets.
3703         (read_autofdo_file): Remove sum_all.
3704         (afdo_callsite_hot_enough_for_early_inline): Remove const
3705         qualifier.
3706         * coverage.c (struct counts_entry): Remove gcov_summary.
3707         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
3708         do not support GCOV_TAG_PROGRAM_SUMMARY.
3709         (get_coverage_counts): Remove summary and expected
3710         arguments.
3711         * coverage.h (get_coverage_counts): Likewise.
3712         * doc/gcov-dump.texi: Remove -w option.
3713         * gcov-dump.c (dump_working_sets): Remove.
3714         (main): Do not support '-w' option.
3715         (print_usage): Likewise.
3716         (tag_summary): Likewise.
3717         * gcov-io.c (gcov_write_summary): Do not dump
3718         histogram.
3719         (gcov_read_summary): Likewise.
3720         (gcov_histo_index): Remove.
3721         (gcov_histogram_merge): Likewise.
3722         (compute_working_sets): Likewise.
3723         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
3724         it not obsolete.
3725         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
3726         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
3727         (GCOV_HISTOGRAM_SIZE): Remove.
3728         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
3729         (struct gcov_summary): Simplify rapidly just
3730         to runs and sum_max fields.
3731         (gcov_histo_index): Remove.
3732         (NUM_GCOV_WORKING_SETS): Likewise.
3733         (compute_working_sets): Likewise.
3734         * gcov-tool.c (print_overlap_usage_message): Remove
3735         trailing empty line.
3736         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
3737         (output_lines): Remove program related line.
3738         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
3739         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
3740         histogram.
3741         (input_profile_summary): Do not read it.
3742         (merge_profile_summaries): And do not merge it.
3743         (input_symtab): Do not call removed function.
3744         * modulo-sched.c (sms_schedule): Do not print sum_max.
3745         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
3746         removed when histogram method was invented.
3747         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
3748         mode.
3749         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
3750         GCOV coding style.
3751         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
3752         and dump selected value.
3753         * profile.c (add_working_set): Remove.
3754         (get_working_sets): Likewise.
3755         (find_working_set): Likewise.
3756         (get_exec_counts): Do not work with working sets.
3757         (read_profile_edge_counts): Do not inform as sum_max is removed.
3758         (compute_branch_probabilities): Likewise.
3759         (compute_value_histograms): Remove argument for call of
3760         get_coverage_counts.
3761         * profile.h: Do not make gcov_summary const.
3763 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
3765         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
3767 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3769         PR tree-optimization/86990
3770         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
3771         Check that the entire merged store group is made of constants only for
3772         overlapping stores.
3774 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3776         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
3777         (VTABLE_VERIFICATION_SPEC): Likewise.
3778         (SANITIZER_EARLY_SPEC): Likewise.
3779         (SANITIZER_SPEC): Likewise.
3780         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
3781         * doc/invoke.texi (Link Options): Document -r.
3783 2018-09-20  Richard Biener <rguenther@suse.de>
3785         PR middle-end/87054
3786         * gimplify.c (gimplify_expr): Retain alignment of
3787         addressable lvalue in dereference.
3789 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
3791         PR bootstrap/87013
3792         * configure.ac: Check for .loc is_stmt support.
3793         * configure, config.in: Rebuilt.
3794         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3795         if not supported.
3797 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3799         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3800         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3801         -misel=no.
3803 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3805         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3806         VECTOR_OTHER.
3807         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3808         case VECTOR_OTHER.
3810 2018-09-20  Marek Polacek  <polacek@redhat.com>
3812         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3814 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3816         PR tree-optimization/87288
3817         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3818         into account when determining PEELING_FOR_NITERS.
3820 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3822         PR tree-optimization/86877
3823         * tree-vect-loop.c (vect_analyze_loop_2): Call
3824         vect_verify_datarefs_alignment.
3826 2018-09-19  Marek Polacek  <polacek@redhat.com>
3828         * doc/invoke.texi: Document -Wclass-conversion.
3830 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
3832         * config/pa/pa.c (pa_adjust_priority): Delete.
3833         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3835         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3836         (atomic_storehi): Likewise.
3837         (atomic_storesi): Likewise.
3838         (atomic_loaddi): Restore compare and swap exchange loop code.
3840 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
3842         PR rtl-optimization/86902
3843         * combine.c (try_combine): When changing the CC mode used, don't change
3844         an unrelated mode in other_insn to that new CC mode.
3846 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
3848         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3849         with %T in place of calls to dump_generic_expr.
3850         (prune_runtime_alias_test_list): Likewise.
3851         (create_runtime_alias_checks): Likewise.
3852         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3853         (vect_analyze_data_ref_dependence): Likewise.
3854         (vect_slp_analyze_data_ref_dependence): Likewise.
3855         (vect_record_base_alignment): Likewise.  Use %G in place of call
3856         to dump_gimple_stmt.
3857         (vect_compute_data_ref_alignment): Likewise.
3858         (verify_data_ref_alignment): Likewise.
3859         (vect_find_same_alignment_drs): Likewise.
3860         (vect_analyze_group_access_1): Likewise.
3861         (vect_analyze_data_ref_accesses): Likewise.
3862         (dependence_distance_ge_vf): Likewise.
3863         (dump_lower_bound): Likewise.
3864         (vect_prune_runtime_alias_test_list): Likewise.
3865         (vect_find_stmt_data_reference): Likewise.
3866         (vect_analyze_data_refs): Likewise.
3867         (vect_create_addr_base_for_vector_ref): Likewise.
3868         (vect_create_data_ref_ptr): Likewise.
3869         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3870         (vect_can_advance_ivs_p): Likewise.
3871         (vect_update_ivs_after_vectorizer): Likewise.
3872         (vect_gen_prolog_loop_niters): Likewise.
3873         (vect_prepare_for_masked_peels): Likewise.
3874         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3875         (vect_determine_vectorization_factor): Likewise.
3876         (vect_is_simple_iv_evolution): Likewise.
3877         (vect_analyze_scalar_cycles_1): Likewise.
3878         (vect_analyze_loop_operations): Likewise.
3879         (report_vect_op): Likewise.
3880         (vect_is_slp_reduction): Likewise.
3881         (check_reduction_path): Likewise.
3882         (vect_is_simple_reduction): Likewise.
3883         (vect_create_epilog_for_reduction): Likewise.
3884         (vect_finalize_reduction:): Likewise.
3885         (vectorizable_induction): Likewise.
3886         (vect_transform_loop_stmt): Likewise.
3887         (vect_transform_loop): Likewise.
3888         (optimize_mask_stores): Likewise.
3889         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3890         (vect_split_statement): Likewise.
3891         (vect_recog_over_widening_pattern): Likewise.
3892         (vect_recog_average_pattern): Likewise.
3893         (vect_determine_min_output_precision_1): Likewise.
3894         (vect_determine_precisions_from_range): Likewise.
3895         (vect_determine_precisions_from_users): Likewise.
3896         (vect_mark_pattern_stmts): Likewise.
3897         (vect_pattern_recog_1): Likewise.
3898         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3899         (vect_record_max_nunits): Likewise.
3900         (vect_build_slp_tree_1): Likewise.
3901         (vect_build_slp_tree_2): Likewise.
3902         (vect_print_slp_tree): Likewise.
3903         (vect_analyze_slp_instance): Likewise.
3904         (vect_detect_hybrid_slp_stmts): Likewise.
3905         (vect_detect_hybrid_slp_1): Likewise.
3906         (vect_slp_analyze_operations): Likewise.
3907         (vect_slp_analyze_bb_1): Likewise.
3908         (vect_transform_slp_perm_load): Likewise.
3909         (vect_schedule_slp_instance): Likewise.
3910         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3911         (vect_mark_stmts_to_be_vectorized): Likewise.
3912         (vect_init_vector_1): Likewise.
3913         (vect_get_vec_def_for_operand): Likewise.
3914         (vect_finish_stmt_generation_1): Likewise.
3915         (vect_check_load_store_mask): Likewise.
3916         (vectorizable_call): Likewise.
3917         (vectorizable_conversion): Likewise.
3918         (vectorizable_operation): Likewise.
3919         (vectorizable_load): Likewise.
3920         (vect_analyze_stmt): Likewise.
3921         (vect_is_simple_use): Likewise.
3922         (vect_get_vector_types_for_stmt): Likewise.
3923         (vect_get_mask_type_for_stmt): Likewise.
3924         * tree-vectorizer.c (increase_alignment): Likewise.
3926 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3928         * doc/rtl.texi: Adjust vec_select description.
3929         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3930         non-constant selectors.
3932 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3934         * config/aarch64/aarch64-protos.h
3935         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3936         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3937         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3938         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3939         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3940         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3941         (AARCH64_ISA_RCPC8_4): New flag.
3942         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3943         to aarch64_offset_9bit_signed_unscaled_p.
3944         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3945         and use stlur.
3946         * config/aarch64/constraints.md (Ust): New constraint.
3947         * config/aarch64/predicates.md.
3948         (aarch64_9bit_offset_memory_operand): New predicate.
3949         (aarch64_rcpc_memory_operand): New predicate.
3951 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
3953         PR rtl-optimization/87361
3954         * rtlanal.c (nonzero_bits1): Revert accidental change.
3956 2018-09-19  Richard Biener  <rguenther@suse.de>
3958         PR tree-optimization/87349
3959         PR tree-optimization/87342
3960         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3962 2018-09-18  Marek Polacek  <polacek@redhat.com>
3964         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3965         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3967 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3969         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3971 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3973         PR rtl-optimization/86882
3974         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3976 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
3978         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3979         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3981 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
3983         PR other/87353
3984         * doc/invoke.texi (Link Options): Fix formatting and grammar.
3986 2018-09-18  Richard Biener  <rguenther@suse.de>
3988         PR middle-end/63155
3989         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3990         (compute_samebase_partition_bases): Likewise.
3991         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3992         (gimple_can_coalesce_p): Simplify.
3994 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
3996         Handle a library implementation of ffs calling __builtin_ffs.
3997         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3998         (mmix_init_libfuncs): New function: make __builtin_ffs expand
3999         to __ffsdi2.
4001 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4003         * diagnostic-show-locus.c (class layout_range): Add field
4004         "m_original_idx".
4005         (layout_range::layout_range): Add "original_idx" param and use it
4006         to initialize new field.
4007         (make_range): Use 0 for original_idx.
4008         (layout::layout): Pass in index to calls to
4009         maybe_add_location_range.
4010         (layout::maybe_add_location_range): Add param "original_idx" and
4011         pass it on to layout_range.
4012         (layout::print_any_labels): Pass on range->m_original_idx to
4013         get_text call.
4014         (gcc_rich_location::add_location_if_nearby): Use 0 for
4015         original_idx.
4016         * gcc-rich-location.h (text_range_label::get_text): Update for new
4017         param.
4018         (range_label_for_type_mismatch::get_text): Likewise.
4020 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4022         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4024 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4026         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4027         format_string_diagnostic_t.
4028         (fmtwarn_n): Likewise.
4029         * substring-locations.c
4030         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4031         (format_warning_n_va): Convert to...
4032         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4033         (format_warning_va): Convert to...
4034         (format_string_diagnostic_t::emit_warning_va): ...this.
4035         (format_warning_at_substring): Convert to...
4036         (format_string_diagnostic_t::emit_warning): ...this.
4037         (format_warning_at_substring_n): Convert to...
4038         (format_string_diagnostic_t::emit_warning_n): ...this.
4039         * substring-locations.h (class format_string_diagnostic_t): New
4040         class.
4041         (format_warning_va): Convert to
4042         format_string_diagnostic_t::emit_warning_va.
4043         (format_warning_n_va): Convert to
4044         format_string_diagnostic_t::emit_warning_n_va.
4045         (format_warning_at_substring): Convert to
4046         format_string_diagnostic_t::emit_warning.
4047         (format_warning_at_substring_n): Convert to
4048         format_string_diagnostic_t::emit_warning_n.
4050 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4051             Bernd Schmidt <bernds_cb1@t-online.de>
4053         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4054         SImode args.
4056 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4058         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4059         operand 0 predicate to nonimmediate operand.
4060         (rint<mode>2_frndint): Remove insn pattern.
4061         (rint<mode>2): Change operand 1 predicate to general_operand.
4062         Extend operand 1 to XFmode and generate rintxf2 insn.
4063         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4064         Do not use X87MODEF mode macro.
4065         (frndintxf2_<rounding>_i387): Rename from
4066         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4067         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4068         to XFmode and generate significandxf3 insn.
4070 2018-09-17  Richard Biener  <rguenther@suse.de>
4072         PR tree-optimization/87328
4073         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4074         visiting unexecutable backedges when not iterating.
4075         (do_rpo_vn): Mark all edges not executable even when not
4076         iterating.
4078 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4080         PR c/63886
4081         * doc/invoke.texi (Warning Options): Likewise.
4083 2018-09-17  Richard Biener  <rguenther@suse.de>
4085         PR tree-optimization/87301
4086         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4087         clean EH info from leftover copy assignments.
4089 2018-09-17  Martin Liska  <mliska@suse.cz>
4091         PR gcov-profile/85871
4092         * gcov.c (output_intermediate_file): Fix out of bounds
4093         access.
4095 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4097         * config/arc/arc.c: Object attributes for core4 not reflected
4098         correctly.
4099         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4100         core3).
4102 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4104         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4106 2018-09-17  Martin Liska  <mliska@suse.cz>
4108         * doc/gcov.texi: Document new option --use-hotness-colors.
4109         * gcov.c (struct source_info): Declare new field.
4110         (source_info::source_info): Set default for maximum_count.
4111         (print_usage): Add new -q option.
4112         (process_args): Process it.
4113         (accumulate_line_info): Save src->maximum_count.
4114         (output_line_beginning): Make color line number if
4115         flag_use_hotness_colors is set.
4116         (output_line_details): Pass default argument value.
4117         (output_lines): Pass src->maximum_count.
4119 2018-09-17  Martin Liska  <mliska@suse.cz>
4121         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4122         Use processor_names table.
4123         * config/i386/i386.c (ix86_default_align): Use
4124         processor_cost_table for alignment values.
4125         (ix86_option_override_internal): Use processor_names.
4126         (ix86_function_specific_print): Likewise.
4127         * config/i386/i386.h (struct processor_costs):
4128         Add alignment values.
4129         (struct ptt): Remove and replace with const char *.
4130         * config/i386/x86-tune-costs.h (struct processor_costs):
4131         Declare default alignments for all costs.
4133 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4135         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4136         symbolics or VR_VARYING ranges for ABS_EXPR.
4137         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4138         when range will wrap.
4140 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4142         PR middle-end/86864
4143         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4144         before and after a JUMP_TABLE_DATA.
4146 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4148         PR middle-end/87188
4149         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4150         when one operand is a function pointer.  Use POINTER_TYPE_P and
4151         FUNC_OR_METHOD_TYPE_P.
4152         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4153         * fold-const.c (build_range_check): Likewise.
4154         * match.pd (simple_comparison): Likewise.
4156 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4158         PR c/82967
4159         * spellcheck.c (get_edit_distance_cutoff): New function.
4160         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4161         (selftest::test_get_edit_distance_one_way): ...this.
4162         (selftest::test_get_edit_distance_unit): Rename to...
4163         (selftest::test_get_edit_distance_both_ways): ...this.
4164         (selftest::test_edit_distances): Move tests to this new function,
4165         and test some more pairs of strings.  Update for above renaming.
4166         (selftest::get_old_cutoff): New function.
4167         (selftest::test_get_edit_distance_cutoff): New function.
4168         (selftest::assert_suggested_for): New function.
4169         (ASSERT_SUGGESTED_FOR): New macro.
4170         (selftest::assert_not_suggested_for): New function.
4171         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4172         (selftest::test_suggestions): New function.
4173         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4174         tests to selftest::test_edit_distances and call it.  Add calls to
4175         selftest::test_get_edit_distance_cutoff and
4176         selftest::test_suggestions.
4177         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4178         (best_match::consider): Replace hard-coded cutoff calculation with
4179         a call to...
4180         (best_match::get_cutoff): New declaration.
4181         (best_match::get_best_meaningful_candidate): Likewise.
4183 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4185         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4187 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4189         revert:
4190         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4192         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4193         terminated string literal.
4195 2018-09-14  Martin Sebor  <msebor@redhat.com>
4197         * builtins.c (unterminated_array): Handle ARRAY_REF.
4198         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4199         * builtins.h (unterminated_array): Declare extern.
4200         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4201         arrays.
4202         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4203         calls.
4205 2018-09-14  Martin Sebor  <msebor@redhat.com>
4206             Jeff Law  <law@redhat.com>
4208         * builtins.c (unterminated_array): New.
4209         (expand_builtin_strcpy): Adjust.
4210         (expand_builtin_strcpy_args): Detect unterminated arrays.
4211         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4212         unterminated arrays.
4213         * gimple-fold.h (get_maxval_strlen): Add argument.
4214         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4216         * gimple-fold.c (get_range_strlen): Add argument.
4217         (get_maxval_strlen): Adjust.
4218         * gimple-fold.h (get_range_strlen): Add argument.
4220 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4222         * config/i386/movdirintrin.h: Fix copyright year.
4224 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4226         * reg-stack.c: Include regs.h.
4227         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4228         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4229         FIRST_STACK_REG, not DFmode.
4230         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4231         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4232         (change stack): Default register mode to the reg_raw_mode of
4233         FIRST_STACK_REG, not DFmode.
4234         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4235         (*swapxf): Rename from swapxf.
4237 2018-09-14  Carl Love  <cel@us.ibm.com>
4239         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4240         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4242 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4244         PR target/87224
4245         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4246         alternatives.
4248 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4250         PR target/85628
4251         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4253 2018-09-14  Jason Merrill  <jason@redhat.com>
4255         Fix --enable-gather-detailed-mem-stats.
4256         * hash-table.c (hash_table_usage): Change from variable to function.
4257         * hash-table.h: Adjust.
4258         * Makefile.in: Add missing dependencies on hash-table.h.
4260 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4262         PR tree-optimization/87259
4263         PR lto/87283
4264         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4265         execute_cse_reciprocals_1 has tried transforming.
4267 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4269         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4270         VR_VARYING for PLUS/MINUS_EXPR.
4272 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4274         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4275         formatting.
4277 2018-09-14  Richard Biener  <rguenther@suse.de>
4279         PR middle-end/63155
4280         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4281         bits for the merged partition.
4283 2018-09-13  Martin Sebor  <msebor@redhat.com>
4284             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4286         * builtins.h (c_srlen): Add argument.
4287         * builtins.c (warn_string_no_nul): New function.
4288         (c_strlen): Add argument and use it.  Update recursive calls.
4289         Pass DECL argument to string_constant to get info on non
4290         terminated strings.  Update *NONSTR as needed.
4291         (fold_builtin_strlen): Add argument to calls to c_strlen.
4292         Warn for unterminated arrays.
4293         (warn_string_no_null): Add prototype.
4294         * expr.c (string_constant): Update arguments.  Update recursive
4295         calls appropriately.  Detect missing NUL terminator and outermost
4296         declaration its missing in.
4297         Improve checks for arrays with nonzero lower bound or elements
4298         that are not a single byte.  Simplify offset computation.
4299         Simplify checks for non-NUL terminated strings.
4300         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4301         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4303 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4305         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4306         correctly.
4307         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4308         argument.  Fix range checks.
4309         * fold-const.h (c_getstr): Adjust protoype.
4310         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4311         string is constant but contains no NUL byte.
4313         * expr.c (string_constant): Adjust function comment.
4314         Remove bogus check for zero termination.
4316         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4318         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4319         (get_constant_size): Don't make STRING_CSTs larger than they are.
4320         (check_string_literal): New check function for STRING_CSTs.
4321         (output_constant): Use it.
4323 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4325         PR target/86812
4326         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4328 2018-09-13  Richard Biener  <rguenther@suse.de>
4330         PR tree-optimization/87263
4331         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4332         (struct unwind_state): Add max_rpo field.
4333         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4334         Compute max_rpo, the max RPO number a block can be backwards reached
4335         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4336         separating it from the iterating mode.
4338 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4340         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4341         (rfs_decision): New scheduling decision.
4343 2018-09-13  Richard Biener  <rguenther@suse.de>
4345         PR bootstrap/87134
4346         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4347         (vn_nary_op_insert_pieces_predicated): Do not write useless
4348         valid_dominated_by_p entry outside of the allocated storage.
4350 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4351             Tom de Vries  <tdevries@suse.de>
4353         PR debug/86985
4354         * dwarf2out.c (is_c): New function.
4355         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4357 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4359         PR target/85628
4360         * config/aarch64/aarch64.md (*aarch64_bfxil):
4361         Define.
4362         * config/aarch64/constraints.md (Ulc): Define.
4363         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4364         Define.
4365         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4366         New function.
4368 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4370         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4371         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4372         aarch64_layout_frame call.
4373         (aarch64_expand_epilogue): Likewise.
4374         (aarch64_initial_elimination_offset): Likewise.
4375         (aarch64_get_separate_components): Likewise.
4376         (aarch64_use_return_insn_p): Likewise.
4377         (aarch64_layout_frame): Remove unneeded check.
4379 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4381         * configure.ac: Only append
4382         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4383         gcc_config_arguments if it was never reconfigured or last reconfigure
4384         was with different arguments.
4385         * configure: Regenerated.
4387 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4388             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4390         PR middle-end/87290
4391         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4392         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4394 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4396         PR tree-optimization/87287
4397         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4398         X % C == 0 to X % (unsigned) C == 0 optimization to ...
4399         * match.pd (X % C == 0): ... here.  New optimization.
4401 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4403         PR middle-end/82853
4404         * expr.h (maybe_optimize_mod_cmp): Declare.
4405         * expr.c (mod_inv): New function.
4406         (maybe_optimize_mod_cmp): New function.
4407         (do_store_flag): Use it.
4408         * cfgexpand.c (expand_gimple_cond): Likewise.
4410 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
4411             Julian Brown  <julian@codesourcery.com>
4413         PR middle-end/86336
4414         * gimplify.c (gimplify_scan_omp_clauses): Set
4415         target_firstprivatize_array_bases in OpenACC parallel and kernels
4416         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4417         OpenACC data regions.
4419 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
4421         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4422         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4424 2018-09-12  Richard Biener  <rguenther@suse.de>
4426         PR tree-optimization/87280
4427         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4428         edge but unreachable target.
4429         (do_rpo_vn): For conservatively handling a PHI only mark
4430         the backedge executable but not the block reachable.
4432 2018-09-12  Richard Biener  <rguenther@suse.de>
4434         PR tree-optimization/87266
4435         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4436         visited blocks.
4438 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
4440         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4441         constants.
4442         ("trunc<BFP:mode><DFP_ALL:mode>2")
4443         ("trunc<DFP_ALL:mode><BFP:mode>2")
4444         ("extend<BFP:mode><DFP_ALL:mode>2")
4445         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4446         according to the target operand type.
4448 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4449             Andreas Krebbel  <krebbel@linux.ibm.com>
4451         PR tree-optimization/86844
4452         * gimple-ssa-store-merging.c
4453         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4454         there are any overlapping stores in between them, make sure they are
4455         also coalesced or we give up completely.
4457 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4459         PR middle-end/87248
4460         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4461         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
4463 2018-09-12  Tom de Vries  <tdevries@suse.de>
4465         * common.opt (gdescribe-dies): Add option.
4466         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4467         attribute for artifical and nameless decls.
4468         (dwarf2out_register_external_die): Add description attribute to
4469         external reference die.
4470         (add_desc_attribute): New functions.
4471         (gen_subprogram_die): Add description attribute to
4472         DW_TAG_call_site_parameter.
4473         * tree-pretty-print.c (print_generic_expr_to_str): New function.
4474         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4475         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4476         -gno-describe-dies.
4477         (@item -gdescribe-dies): Add.
4479 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4481         * tree-vrp.c (vrp_shift_undefined_p): Remove.
4482         (extract_range_from_binary_expr_1: Call
4483         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4484         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4485         depend on sign.
4487 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4489         * gimple-ssa-warn-alloca.c
4490         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4491         field for ALLOCA_BOUND_*_LARGE.
4493 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4495         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4497 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4499         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4500         for clobbers.  Remove obsolete comment.
4502 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4504         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4505         mpxchk, mpxld and mpxst types.
4506         (define_attr length_immediate): Remove all processing of mpx types.
4507         (define_attr prefix_0f): Ditto.
4508         (define_attr memory): Ditto.
4510 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4512         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
4513         (log<mode>2): Change operand 1 predicate to general_operand.
4514         Extend operand 1 to XFmode and generate logxf3 insn.
4515         (log10<mode>2): Change operand 1 predicate to general_operand.
4516         Extend operand 1 to XFmode and generate log10xf3 insn.
4517         (log2<mode>2): Change operand 1 predicate to general_operand.
4518         Extend operand 1 to XFmode and generate log2xf3 insn.
4519         (fyl2xp1_extend<mode>xf3_i387): Remove.
4520         (log1p<mode>2): Change operand 1 predicate to general_operand.
4521         Extend operand 1 to XFmode and generate log1pxf3 insn.
4522         (fxtract_extend<mode>xf3_i387): Remove.
4523         (logb<mode>2): Change operand 1 predicate to general_operand.
4524         Extend operand 1 to XFmode and generate logbxf3 insn.
4525         (ilogb<mode>2): Change operand 1 predicate to general_operand.
4526         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
4527         (significand<mode>2): Change operand 1 predicate to general_operand.
4528         Extend operand 1 to XFmode and generate significandxf3 insn.
4530 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4532         * gcc.c (perror_with_name, pfatal_with_name): Delete.
4533         (load_specs): Use fatal_error.
4534         (DELETE_IF_ORDINARY, process_command): Use error.
4535         (execute, run_attempt): Use fatal_error.
4537 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
4539         * diagnostic-core.h (sorry_at): New prototype.
4540         * diagnostic.c (sorry_at): New function.
4542 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
4544         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
4545         by zero as VR_UNDEFINED.
4547 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4549         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
4550         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
4551         (<sincos>mode2): New expander.
4552         (sincos_extend<mode>xf3_i387): Remove insn pattern.
4553         (sincos -> sin, cos splitters): Remove splitter patterns.
4554         (sincos<mode>3): Change operand 2 predicate to general_operand.
4555         Extend operand 2 to XFmode and generate sincosxf3 insn.
4556         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
4557         Change operand 3 predicate to const1_operand.
4558         (fptan_extend<mode>xf4_i387): Remove insn pattern.
4559         (tanxf2): Update operands in the call to fptanxf4_i387.
4560         (tan<mode>2): Change operand 1 predicate to general_operand.
4561         Extend operand 1 to XFmode and generate tanxf3 insn.
4562         (atan2xf3): Rename from *fpatanxf3_i387.
4563         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
4564         (atan2xf3): Remove expander.
4565         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
4566         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
4567         (atan<mode>2): Change operand 1 predicate to general_operand.
4568         Extend operand 1 to XFmode and generate atanxf3 insn.
4570 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4572         * config/i386/i386.md (x87/SSE constant load splitter): Use
4573         memory_operand instead of nonimmediate_operand for input operand
4574         predicate.
4576 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
4578         * config/i386/i386.md (float partial SSE register stall splitter): Move
4579         splitter near its instruction pattern.
4580         (float_extend partial SSE register stall splitter): Ditto.
4581         (float_truncate partial SSE register stall splitter): Ditto.
4583 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
4585         PR target/86794
4586         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4587         to speculation_safe_value_not_needed.
4589         PR target/85666
4590         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
4591         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
4592         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
4593         leaf_function_p, instead use has_hard_reg_initial_val.
4595 2018-09-09  Nathan Sidwell  <nathan@acm.org>
4597         * gcc.h (pfatal_with_name): Don't declare here.
4598         * gcc.c (pfatal_with_name): Make static.
4600 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4602         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
4603         earlyclobber.
4605 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
4607         PR rtl-optimization/85458
4608         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
4609         priority hook to reduce the priority of EXPR.
4611 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
4613         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
4614         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
4615         with X87_ENABLE_FLOAT.  Remove preparation code.
4616         (*float<SWI48:mode><MODEF:mode>2): Rename from
4617         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
4618         math using "enabled" attribute.
4619         (*floatdi<MODEF:mode>2_i387): Rename from
4620         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
4621         enable for 32bit targets only.
4622         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
4623         splitter.
4624         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
4625         as operand 1 predicate.  Rewrite as define_insn_and_split.
4626         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
4628 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4630         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
4631         to fallthru to FLOAT case.
4633 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4635         PR target/86731
4636         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
4637         around folding of vec_sl to handle out of range shift values.
4639 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4641         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
4642         Update callers to gen_fix_trunc<mode>_i387_fisttp
4643         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
4644         nonimmediate_operand.
4645         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
4646         and corresponding splitters.
4647         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
4648         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
4649         (fix_truncdi_i387_with_temp): Remove insn pattern
4650         and corresponding splitters.
4651         (fix_trunc<mode>_i387): Change operand 0 predicate to
4652         nonimmediate_operand.
4653         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
4654         and corresponding splitters.
4655         (*fistdi2_1): Remove.
4656         (fistdi2): Ditto.
4657         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
4658         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
4659         (*fist<mode>2_1): Remove.
4660         (fist<mode>2): Ditto.
4661         (fist<mode>2_with_temp): Remove insn pattern and corresponding
4662         splitters.
4663         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
4664         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
4665         (fistdi2_<rounding>): Change operand 0 predicate to
4666         nonimmediate_operand.
4667         (fistdi2_<rounding>_with_temp): Remove insn pattern
4668         and corresponding splitters.
4669         (fist<mode>2_<rounding>): Change operand 0 predicate to
4670         nonimmediate_operand.
4671         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
4672         and corresponding splitters.
4674         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
4676 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4678         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
4679         the init value.
4681 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4683         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4684         early gimple folding of vec_splat().
4685         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
4686         * gimple-fold.h: Add an extern define for tree_vec_extract().
4688 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4690         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
4691         wrappers around TREE_TYPE comparisons.
4693 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4695         PR target/80080
4696         * config/s390/predicates.md: Add nonsym_memory_operand.
4697         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
4698         contains a SYMBOL_REF, load it into an intermediate pseudo.
4699         (s390_emit_compare_and_swap): Legitimize operand.
4700         * config/s390/s390.md: Use the new nonsym_memory_operand
4701         with UNSPECV_CAS patterns.
4703 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4705         PR target/80080
4706         * config/s390/s390-passes.def: New file.
4707         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
4708         declaration.
4709         (make_pass_s390_early_mach): Add declaration.
4710         * config/s390/s390.c (make_pass_s390_early_mach):
4711         (s390_option_override): Remove dynamic registration.
4712         * config/s390/t-s390: Add s390-passes.def.
4714 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4716         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4717         Remove UNSPEC_LTREL_BASE check.
4718         (annotate_constant_pool_refs): Likewise.
4719         (find_constant_pool_ref): Likewise.
4720         (find_ltrel_base): Removed.
4721         (replace_ltrel_base): Removed.
4722         (s390_mainpool_finish): Remove replace_ltrel_base call.
4723         (s390_chunkify_start): Remove pending LTREL_BASE logic.
4724         (s390_chunkify_finish): Remove replace_ltrel_base call.
4725         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
4727 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
4729         PR target/86779
4730         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4731         to speculation_safe_value_not_needed.
4733 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
4734             Bernd Schmidt  <bernds_cb1@t-online.de>
4736         * config/nvptx/nvptx-opts.h: New file.
4737         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
4738         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
4739         (ASM_SPEC): Define.
4740         (TARGET_SM35): New macro.
4741         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
4742         correct predicate.
4743         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
4744         values.
4745         (misa=): New option.
4746         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
4748 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
4750         * config/i386/i386.md (truncdfsf2): Remove expander.
4751         (truncdfsf2_with_temp): Ditto.
4752         (truncxf<mode>2): Ditto.
4753         (*truncdfsf_fast_mixed): Remove insn pattern.
4754         (*truncdfsf_fast_i387): Ditto.
4755         (*truncdfsf_mixed): Ditto.
4756         (*truncdfsf_i387): Ditto.
4757         (*truncdfsf2_i387_1): Ditto.
4758         (*truncxfsf2_mixed): Ditto.
4759         (*truncxfdf2_mixed): Ditto.
4760         (*truncxf<mode>2_i387_noop): Ditto. Update callers
4761         to call gen_truncxf<mode>2 instead.
4762         (*truncxf<mode>2_i387): Remove.
4763         (reg->reg splitters): Remove splitter pattern.
4764         (reg->mem splitters): Ditto.
4766         (truncdfsf2): New insn pattern.
4767         (truncxf<mode>2): Ditto.
4769 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4771         * tree-ssa-math-opts.c (is_mult_by): New function.
4772         (is_square_of): Use the above.
4773         (optimize_recip_sqrt): New function.
4774         (pass_cse_reciprocals::execute): Use the above.
4776 2018-09-05  Richard Biener  <rguenther@suse.de>
4778         PR bootstrap/87134
4779         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
4780         to zero-init the emplaced vec.
4782 2018-09-05  Martin Liska  <mliska@suse.cz>
4784         PR tree-optimization/87205
4785         * tree-switch-conversion.c (pass_lower_switch::execute):
4786         Group cases for switch statements.
4788 2018-09-05  Richard Biener  <rguenther@suse.de>
4790         PR tree-optimization/87217
4791         * tree-ssa-sccvn.c (vuse_valueize): New.
4792         (vn_reference_lookup_pieces): Use it.
4793         (vn_reference_lookup): Likewise.
4795 2018-09-05  Nathan Sidwell  <nathan@acm.org>
4797         PR c++/87137
4798         * stor-layout.c (place_field): Scan forwards to check last
4799         bitfield when ms_bitfield_placement is in effect.
4801 2018-09-05  Richard Biener  <rguenther@suse.de>
4803         PR bootstrap/87225
4804         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4805         return.
4807 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4808             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4810         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4811         * config.gcc (extra_objs): Build it.
4812         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4813         Likewise.
4814         * config/aarch64/aarch64-passes.def
4815         (pass_tag_collision_avoidance): New pass.
4816         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4817         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4818         (aarch64_classify_address): Remove static qualifier.
4819         (aarch64_address_info, aarch64_address_type): Move to...
4820         * config/aarch64/aarch64-protos.h: ... here.
4821         (make_pass_tag_collision_avoidance): New function.
4822         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4823         New tuning flag.
4825 2018-09-05  Martin Liska  <mliska@suse.cz>
4827         * doc/gcov.texi: Update documentation of humar
4828         readable mode.
4829         * gcov.c (format_count): Print one decimal place, it provides
4830         more fine number of situations like '1G' vs. '1.4G'.
4832 2018-09-05  Martin Liska  <mliska@suse.cz>
4834         PR target/87164
4835         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4836         * optc-gen.awk: Allow 'Var' for Deprecated options in order
4837         to generate a MASK value.
4839 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
4841         PR debug/86593
4842         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4843         if frame pointer isn't used.
4844         (compute_frame_pointer_to_fb_displacement): Likewise.
4846 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
4848         PR target/87198
4849         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4850         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4851         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4852         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4853         and OPTION_MASK_ISA_XSAVEC_UNSET.
4855 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
4857         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4858         XOR operations in NAND case.
4860 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4862         * wide-int-range.cc (wide_int_range_convert): New.
4863         * wide-int-range.h (wide_int_range_convert): New.
4864         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4865         code into wide_int_range_convert.
4866         (extract_range_into_wide_ints): Do not munge anti range constants
4867         into the entire domain.  Just return the range back.
4869 2018-09-04  Martin Liska  <mliska@suse.cz>
4871         * genmatch.c (output_line_directive): Add new argument
4872         fnargs.
4873         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4875 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4877         * doc/invoke.texi (Option Summary): Add whitespace.
4879         * doc/invoke.texi (Option Summary): Add -Waligned-new.
4881 2018-09-04  Richard Biener  <rguenther@suse.de>
4883         PR tree-optimization/87211
4884         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4885         backedge value we're supposed to treat as VARYING also number
4886         the PHI to VARYING in case it got a different value-number already.
4888 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4890         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4891         (extract_range_from_binary_expr_1): Do not call
4892         vrp_can_optimize_bit_op.
4893         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4894         static.
4895         (wide_int_range_get_mask_and_bounds): New.
4896         (wide_int_range_optimize_bit_op): New.
4897         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4898         (wide_int_range_bit_and): Same.
4899         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4900         (wide_int_range_optimize_bit_op): New.
4901         (wide_int_range_get_mask_and_bounds): New.
4903 2018-09-04  Richard Biener  <rguenther@suse.de>
4905         PR tree-optimization/87176
4906         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4907         variable.  When value-numbering a virtual PHI node make sure
4908         to not value-number to the backedge value.
4910 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4912         * doc/extend.texi (Long Long, Hex Floats): Document support for
4913         long long and hex floats in more recent versions of ISO C++.
4915 2018-09-03  Richard Biener  <rguenther@suse.de>
4917         PR tree-optimization/87177
4918         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4919         cleanup.
4921 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4923         * bb-reorder.c (edge_order): Convert to C-qsort-style
4924         tri-state comparator.
4925         (reorder_basic_blocks_simple): Change std::stable_sort to
4926         gcc_stablesort.
4928 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4930         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4931         tri-state comparator.
4932         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4934 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4936         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4937         (mergesort): ... here as maximum count for using netsort.
4938         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4939         (gcc_stablesort): New.
4940         * system.h (gcc_stablesort): Declare.
4942 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4944         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4945         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
4946         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
4948 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
4950         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4951         lxsdx and stxsdx alternatives.
4952         (*mov<mode>_hardfloat64): Ditto.
4953         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4955 2018-09-03  Richard Biener  <rguenther@suse.de>
4957         PR tree-optimization/87200
4958         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4959         simplify result.
4961 2018-09-03  Martin Liska  <mliska@suse.cz>
4963         PR tree-optimization/87201
4964         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4965         Fix parenthesis in an expression.
4967 2018-09-03  Richard Biener  <rguenther@suse.de>
4969         PR tree-optimization/87197
4970         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4971         visited.  CSE the VN_INFO hashtable lookup.
4973         PR tree-optimization/87169
4974         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4975         iterating make sure there's no extra backedges from irreducible
4976         regions feeding the header.  Mark the destination block
4977         executable.
4979 2018-09-03  Martin Liska  <mliska@suse.cz>
4981         PR driver/83193
4982         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4983         * common/common-targhooks.c (default_get_valid_option_values):
4984         New function.
4985         * common/common-targhooks.h (default_get_valid_option_values):
4986         Likewise.
4987         * common/config/i386/i386-common.c: Move processor_target_table
4988         from i386.c.
4989         (ix86_get_valid_option_values): New function.
4990         (TARGET_GET_VALID_OPTION_VALUES): New macro.
4991         * config/i386/i386.c (struct ptt): Move to i386-common.c.
4992         (PTA_*): Move all defined masks into i386-common.c.
4993         (ix86_function_specific_restore): Use new processor_cost_table.
4994         * config/i386/i386.h (struct ptt): Moved from i386.c.
4995         (struct pta): Likewise.
4996         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4997         * doc/tm.texi.in: Likewise.
4998         * opt-suggestions.c (option_proposer::suggest_option):
4999         Pass prefix to build_option_suggestions.
5000         (option_proposer::get_completions): Likewise.
5001         (option_proposer::build_option_suggestions): Use the new target
5002         hook.
5003         * opts.c (struct option_help_tuple): New struct.
5004         (print_filtered_help): Use the new target hook.
5006 2018-09-03  Martin Liska  <mliska@suse.cz>
5008         PR middle-end/59521
5009         * predict.c (set_even_probabilities): Add likely_edges
5010         argument and handle cases where we have precisely one
5011         likely edge.
5012         (combine_predictions_for_bb): Catch also likely_edges.
5013         (tree_predict_by_opcode): Handle gswitch statements.
5014         * tree-cfg.h (find_case_label_for_value): New declaration.
5015         (find_taken_edge_switch_expr): Likewise.
5016         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5017         Find pivot in decision tree based on probabily, not by number of
5018         nodes.
5020 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5022         * doc/standards.texi (Standards): Update Objective-C reference.
5024 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5026         * doc/install.texi (Prerequisites): Update link for MPC.
5028 2018-09-01  Michael Matz  <matz@suse.de>
5030         PR tree-optimization/87074
5031         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5032         PHIs for outer-loop uses.
5034 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5036         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5037         * doc/invoke.texi (C Dialect Options): Ditto.
5039 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5041         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5043 2018-08-31  Richard Biener  <rguenther@suse.de>
5045         PR tree-optimization/87168
5046         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5047         (rpo_elim::eliminate_avail): When OP was not visited it must
5048         be available.
5050 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5052         * tree-vrp.c (copy_value_range): Convert param "from" from
5053         "value_range *" to "const value_range *".
5054         (range_is_null): Likewise for param "vr".
5055         (range_int_cst_p): Likewise.
5056         (range_int_cst_singleton_p): Likewise.
5057         (symbolic_range_p): Likewise.
5058         (value_ranges_intersect_p): Likewise for both params.
5059         (value_range_nonnegative_p): Likewise for param "vr".
5060         (value_range_constant_singleton): Likewise.
5061         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5062         (extract_range_into_wide_ints): Likewise for param "vr".
5063         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5064         and "vr1".
5065         (vrp_can_optimize_bit_op): Likewise.
5066         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5067         "vr1_".
5068         (extract_range_from_unary_expr): Likewise.
5069         (debug_value_range): Likewise for param "vr".
5070         (value_range::dump): Add "const" qualifier.
5071         (vrp_prop::check_array_ref): Convert local "vr" from
5072         "value_range *" to "const value_range *".
5073         (vrp_prop::check_mem_ref): Likewise.
5074         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5075         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5076         (vrp_intersect_ranges): Likewise.
5077         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5078         (vrp_prop::vrp_finalize): Likewise.
5079         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5080         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5081         (extract_range_from_unary_expr): Likewise.
5082         (value_range_constant_singleton): Likewise.
5083         (symbolic_range_p): Likewise.
5084         (copy_value_range): Likewise.
5085         (extract_range_from_binary_expr_1): Likewise.
5086         (range_int_cst_p): Likewise.
5087         (vrp_set_zero_nonzero_bits): Likewise.
5088         (range_int_cst_singleton_p): Likewise.
5090 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5092         * config/aarch64/arm_neon.h (vabsd_s64): New.
5093         (vnegd_s64): Likewise.
5095 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5097         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5099 2018-08-31  Martin Liska  <mliska@suse.cz>
5101         * ipa-icf.c (sem_item::add_type): Use
5102         sem_item::m_type_hash_cache.
5103         * ipa-icf.h: Move the cache from sem_item_optimizer
5104         to sem_item.
5106 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5108         * doc/extend.texi (Backwards Compatibility): Remove implicit
5109         extern C leeway of () being (...).
5111 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5113         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5115 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5117         PR target/86684
5118         PR target/87149
5119         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5121 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5123         PR middle-end/87138
5124         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5125         gen_int_mode.  Formatting fixes.
5127 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5129         * target.def (custom_function_descriptors): Improve documentation.
5130         * doc/tm.texi.in (Trampolines): Expand discussion of function
5131         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5132         beginning of the section.
5133         * doc/tm.texi: Regenerated.
5135 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5137         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5138         base class in the initializer list.  This is a workaround for
5139         building with older GCC.
5140         (class auto_bb_flag): Likewise.
5142 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5144         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5145         (altivec_vcmpequ<VI_char>_p): Remove star.
5146         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5147         vector load modes.
5148         (expand_strncmp_vec_sequence): New function.
5149         (emit_final_str_compare_vec): New function.
5150         (expand_strn_compare): Add support for vector strncmp.
5151         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5152         length specification to bytes.
5153         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5154         (vcmpnezb_p): New pattern.
5155         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5156         for option -mstring-compare-inline-limit.
5158 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5160         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5161         (PTA_SKYLAKE): Add PTA_AES.
5162         (PTA_GOLDMONT): Likewise.
5164 2018-08-29  Jan Hubicka  <jh@suse.cz>
5166         PR lto/86517
5167         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5168         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5170 2018-08-29  Jan Hubicka  <jh@suse.cz>
5172         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5173         TYPE_STUB_DECL.
5174         (hash_tree): Do not visit TYPE_STUB_DECL.
5175         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5176         stream TYPE_STUB_DECL.
5177         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5178         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5179         after free_lang_data.
5180         (type_in_anonymous_namespace_p): Likewise.
5182 2018-08-29  Jan Hubicka  <jh@suse.cz>
5184         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5185         comment that it has to be even number.
5186         (class sreal): Change m_sig type to int32_t.
5187         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5188         int64_t for temporary calculations.
5189         (sreal_verify_basics): Drop one bit from minimum and maximum.
5191 2018-08-30  Richard Biener  <rguenther@suse.de>
5193         PR tree-optimization/87147
5194         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5195         (visit_phi): When the degenerate result is from the backedge and
5196         we didn't visit its definition yet drop to VARYING.
5197         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5199 2018-08-29  Jan Hubicka  <jh@suse.cz>
5201         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5202         DECL_VINDEX.
5203         (hash_tree): Likewise.
5205 2018-08-29  Jan Hubicka  <jh@suse.cz>
5207         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5208         and TYPE_NEXT_REF_TO.
5210 2018-08-29  Jan Hubicka  <jh@suse.cz>
5212         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5213         comment that it has to be even number.
5214         (class sreal): Change m_sig type to int32_t.
5215         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5216         int64_t for temporary calculations.
5217         (sreal_verify_basics): Drop one bit from minimum and maximum.
5219 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5221         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5223 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5225         PR middle-end/86995
5226         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5227         if to_add is negative.
5229 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5231         PR middle-end/87053
5232         * builtins.c (c_strlen): Improve range checks.
5234 2018-08-29  Martin Sebor  <msebor@redhat.com>
5235             Jeff Law  <law@redhat.com>
5237         PR tree-optimization/86714
5238         PR tree-optimization/86711
5239         * builtins.c (c_strlen): Add arguments to call to string_constant.
5240         * expr.c (string_constant): Add argument.  Detect missing nul
5241         terminator and outermost declaration it's missing in.
5242         * expr.h (string_constant): Add argument.
5243         * fold-const.c (read_from_constant_string): Add arguments to call to
5244         string_constant.
5245         (c_getstr): Likewise.
5246         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5247         to string_constant.
5248         * tree-ssa-strlen.c (get_stridx): Likewise.
5250 2018-08-29  Jan Hubicka  <jh@suse.cz>
5252         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5253         Do not stream DECL_VINDEX.
5254         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5255         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5256         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5258 2018-08-29  Richard Biener  <rguenther@suse.de>
5260         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5261         virtual operands that are not default defs to honor region
5262         boundaries.
5263         (rpo_vn_valueize): Remove ineffective code here.
5265 2018-08-29  Richard Biener  <rguenther@suse.de>
5267         PR tree-optimization/87132
5268         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5269         when skipping defs reachable over backedges.
5271 2018-08-29  Richard Biener  <rguenther@suse.de>
5273         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5274         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5275         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5276         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5277         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5278         (vn_reference_lookup_3): Remove use of const_parms.
5279         (free_rpo_vn): Do not free const_parms.
5280         (do_rpo_vn): Do not call init_const_parms.
5281         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5282         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5283         (call_may_clobber_ref_p_1): Likewise.
5285 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5287         PR other/86726
5288         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5289         (-O): Ditto.
5290         (-ftree-scev-cprop): Document.
5292 2018-08-29  Jan Hubicka  <jh@suse.cz>
5294         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5295         parameters.
5296         (sreal constructor): Update.
5297         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5298         sreal:operator/): Update.
5300 2018-08-29  Martin Liska  <mliska@suse.cz>
5302         * tree-switch-conversion.c (switch_conversion::expand):
5303         Strenghten assumption about gswitch statements.
5305 2018-08-29  Richard Biener  <rguenther@suse.de>
5307         PR tree-optimization/87117
5308         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5309         re-value-number released SSA VDEFs.
5311 2018-08-29  Richard Biener  <rguenther@suse.de>
5313         PR tree-optimization/87126
5314         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5316 2018-08-28  Jim Wilson  <jimw@sifive.com>
5318         * config/riscv/pic.md: Rewrite.
5319         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5320         invalid address.
5321         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5322         (SOFTF, default_load, softload, softstore): New.
5324 2018-08-28  Jeff Law  <law@redhat.com>
5326         * fold-const.c (fold_binary_loc): Remove recently added assert.
5328 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5330         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5331         to OP parmeter of generated function.
5333 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5335         PR tree-optimization/87009
5336         * match.pd: Add boolean optimizations.
5338 2018-08-28  Martin Sebor  <msebor@redhat.com>
5340         PR middle-end/86631
5341         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5342         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5343         (pass_walloca::gate): Use it.
5344         (alloca_call_type): Same.
5345         (pass_walloca::execute): Same.
5346         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5348 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5350         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5351         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5352         >= 3005 to >= 9000.
5354 2018-08-28  Richard Biener  <rguenther@suse.de>
5356         PR tree-optimization/87124
5357         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5358         constants before looking up avail.
5360 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5362         PR middle-end/87099
5363         * calls.c (maybe_warn_nonstring_arg): Punt early if
5364         warn_stringop_overflow is zero.  Don't call get_range_strlen
5365         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5366         Swap comparison operands to have constants on rhs.  Only use
5367         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5368         increment lenrng[0].
5370 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5372         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5373         use of tree_to_shwi.  Remove duplicated test for the size being
5374         a whole number of bytes.
5376 2018-08-28  Richard Biener  <rguenther@suse.de>
5378         PR tree-optimization/87117
5379         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5380         Handle removed stmt without LHS (GIMPLE_NOP).
5382 2018-08-28  Richard Biener  <rguenther@suse.de>
5384         PR tree-optimization/87117
5385         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5386         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5388 2018-08-28  Richard Biener  <rguenther@suse.de>
5390         PR tree-optimization/87117
5391         * tree-ssa-pre.c (compute_avail): Do not make expressions
5392         with predicated values available.
5393         (get_expr_value_id): Assert we do not run into predicated value
5394         expressions.
5396 2018-08-28  Richard Biener  <rguenther@suse.de>
5398         PR tree-optimization/87117
5399         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5400         get virtual operands.
5401         (get_expr_operands): Handle STRING_CST like other decls.
5403 2018-08-28  Martin Liska  <mliska@suse.cz>
5405         * tree.h: Update documentation of fndecl_built_in_p
5406         functions.
5409 2018-08-27  Jeff Law  <law@redhat.com>
5410         PR tree-optimization/87110
5411         * tree-ssa-dse.c (compute_trims): Handle non-constant
5412         TYPE_SIZE_UNIT.
5414 2018-08-27  Martin Sebor  <msebor@redhat.com>
5416         PR tree-optimization/86914
5417         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5419 2018-08-27  Martin Sebor  <msebor@redhat.com>
5421         PR tree-optimization/87112
5422         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5423         the type of the bound argument.
5425 2018-08-27  Jeff Law  <law@redhat.com>
5427         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5428         type does not have a TYPE_SIZE_UNIT.
5430 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
5432         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5433         with include of backend.h.
5435 2018-08-27  Richard Biener  <rguenther@suse.de>
5437         PR tree-optimization/86927
5438         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5439         use const cond reduction code.
5441 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
5443         PR tree-optimization/85758
5444         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5446 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5448         PR c++/87091
5449         * diagnostic-show-locus.c (class layout_range): Update for
5450         conversion of show_caret_p to a tri-state.
5451         (layout_range::layout_range): Likewise.
5452         (make_range): Likewise.
5453         (layout::maybe_add_location_range): Likewise.
5454         (layout::should_print_annotation_line_p): Don't show annotation
5455         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5456         (layout::get_state_at_point): Update for conversion of
5457         show_caret_p to a tri-state.  Bail out early for
5458         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5459         underlining or source colorization.
5460         (gcc_rich_location::add_location_if_nearby): Update for conversion
5461         of show_caret_p to a tri-state.
5462         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5463         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5464         Likewise.
5465         (selftest::test_one_liner_labels): Likewise.
5466         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5467         conversion of show_caret_p to a tri-state.
5468         * pretty-print.c (text_info::set_location): Likewise.
5469         * pretty-print.h (text_info::set_location): Likewise.
5470         * substring-locations.c (format_warning_n_va): Likewise.
5471         * tree-diagnostic.c (default_tree_printer): Likewise.
5472         * tree-pretty-print.c (newline_and_indent): Likewise.
5474 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5476         PR c++/87091
5477         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5478         line above for line-insertion fix-it hints.
5479         (selftest::test_fixit_insert_containing_newline): Update the
5480         expected results, and add a test with line-numbering enabled.
5482 2018-08-27  Martin Liska  <mliska@suse.cz>
5484         PR sanitizer/86962
5485         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5486         params with DECL_HAS_VALUE_EXPR_P.
5488 2018-08-27  Martin Liska  <mliska@suse.cz>
5490         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5491         selected expansion strategy.
5493 2018-08-27  Martin Liska  <mliska@suse.cz>
5495         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5496         * builtins.c (is_builtin_fn): Likewise.
5497         * attribs.c (diag_attr_exclusions): Use new function
5498         fndecl_built_in_p and remove check for FUNCTION_DECL if
5499         possible.
5500         (builtin_mathfn_code): Likewise.
5501         (fold_builtin_expect): Likewise.
5502         (fold_call_expr): Likewise.
5503         (fold_builtin_call_array): Likewise.
5504         (fold_call_stmt): Likewise.
5505         (set_builtin_user_assembler_name): Likewise.
5506         (is_simple_builtin): Likewise.
5507         * calls.c (gimple_alloca_call_p): Likewise.
5508         (maybe_warn_nonstring_arg): Likewise.
5509         * cfgexpand.c (expand_call_stmt): Likewise.
5510         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
5511         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
5512         (cgraph_node::verify_node): Likewise.
5513         * cgraphclones.c (build_function_decl_skip_args): Likewise.
5514         (cgraph_node::create_clone): Likewise.
5515         * config/arm/arm.c (arm_insert_attributes): Likewise.
5516         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
5517         * dse.c (scan_insn): Likewise.
5518         * expr.c (expand_expr_real_1): Likewise.
5519         * fold-const.c (operand_equal_p): Likewise.
5520         (fold_binary_loc): Likewise.
5521         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
5522         * gimple-low.c (lower_stmt): Likewise.
5523         * gimple-pretty-print.c (dump_gimple_call): Likewise.
5524         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5525         Likewise.
5526         * gimple.c (gimple_build_call_from_tree): Likewise.
5527         (gimple_call_builtin_p): Likewise.
5528         (gimple_call_combined_fn): Likewise.
5529         * gimplify.c (gimplify_call_expr): Likewise.
5530         (gimple_boolify): Likewise.
5531         (gimplify_modify_expr): Likewise.
5532         (gimplify_addr_expr): Likewise.
5533         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
5534         * ipa-cp.c (determine_versionability): Likewise.
5535         * ipa-fnsummary.c (compute_fn_summary): Likewise.
5536         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
5537         * ipa-split.c (visit_bb): Likewise.
5538         (split_function): Likewise.
5539         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5540         * lto-cgraph.c (input_node): Likewise.
5541         * lto-streamer-out.c (write_symbol): Likewise.
5542         * omp-low.c (setjmp_or_longjmp_p): Likewise.
5543         (lower_omp_1): Likewise.
5544         * predict.c (strip_predict_hints): Likewise.
5545         * print-tree.c (print_node): Likewise.
5546         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
5547         * trans-mem.c (is_tm_irrevocable): Likewise.
5548         (is_tm_load): Likewise.
5549         (is_tm_simple_load): Likewise.
5550         (is_tm_store): Likewise.
5551         (is_tm_simple_store): Likewise.
5552         (is_tm_abort): Likewise.
5553         (tm_region_init_1): Likewise.
5554         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
5555         * tree-cfg.c (verify_gimple_call): Likewise.
5556         (move_stmt_r): Likewise.
5557         (stmt_can_terminate_bb_p): Likewise.
5558         * tree-eh.c (lower_eh_constructs_2): Likewise.
5559         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
5560         * tree-inline.c (remap_gimple_stmt): Likewise.
5561         (copy_bb): Likewise.
5562         (estimate_num_insns): Likewise.
5563         (fold_marked_statements): Likewise.
5564         * tree-sra.c (scan_function): Likewise.
5565         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
5566         (optimize_stack_restore): Likewise.
5567         (pass_fold_builtins::execute): Likewise.
5568         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
5569         (mark_all_reaching_defs_necessary_1): Likewise.
5570         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
5571         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5572         (pass_forwprop::execute): Likewise.
5573         * tree-ssa-loop-im.c (stmt_cost): Likewise.
5574         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5575         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
5576         * tree-ssa-strlen.c (get_string_length): Likewise.
5577         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
5578         (find_func_aliases_for_call): Likewise.
5579         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
5580         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
5581         * tree-tailcall.c (find_tail_calls): Likewise.
5582         * tree.c (need_assembler_name_p): Likewise.
5583         (free_lang_data_in_decl): Likewise.
5584         (get_call_combined_fn): Likewise.
5585         * ubsan.c (is_ubsan_builtin_p): Likewise.
5586         * varasm.c (incorporeal_function_p): Likewise.
5587         * tree.h (DECL_BUILT_IN): Remove and replace with
5588         fndecl_built_in_p.
5589         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
5590         (fndecl_built_in_p): New.
5592 2018-08-27  Martin Liska  <mliska@suse.cz>
5594         PR tree-optimization/86847
5595         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
5596         Dump also subtree probability.
5597         (switch_decision_tree::do_jump_if_equal): New function.
5598         (switch_decision_tree::emit_case_nodes): Handle special
5599         situations in balanced tree that can be emitted much simpler.
5600         Fix calculation of probabilities that happen in tree expansion.
5601         * tree-switch-conversion.h (struct cluster): Add
5602         is_single_value_p.
5603         (struct simple_cluster): Likewise.
5604         (struct case_tree_node): Add new function has_child.
5605         (do_jump_if_equal): New.
5607 2018-08-27  Martin Liska  <mliska@suse.cz>
5609         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
5610         Add new argument to bit_test_cluster constructor.
5611         (bit_test_cluster::emit): Set bits really number of values
5612         handlel by a test.
5613         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
5614         probability argument.
5615         * tree-switch-conversion.h (struct bit_test_cluster):
5616         Add m_handles_entire_switch.
5618 2018-08-27  Martin Liska  <mliska@suse.cz>
5620         PR tree-optimization/86702
5621         * tree-switch-conversion.c (jump_table_cluster::emit):
5622         Make probabilities even for values in jump table
5623         according to number of cases handled.
5624         (switch_decision_tree::compute_cases_per_edge): Pass
5625         argument to reset_out_edges_aux function.
5626         (switch_decision_tree::analyze_switch_statement): Likewise.
5627         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
5628         Make it static.
5630 2018-08-27  Martin Liska  <mliska@suse.cz>
5632         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
5633         cfun argument explicitly.
5634         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
5635         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
5636         function gimple_switch_default_bb.
5637         (convert_switch_statements):
5638         (expand_builtins):
5639         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
5640         * stmt.c (label_to_block_fn): Use label_to_block and pass
5641         cfun argument explicitly and use gimple_switch_label_bb.
5642         (expand_case): Likewise.
5643         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
5644         cfun argument explicitly. Likewise.
5645         (make_edges_bb): Likewise.
5646         (make_cond_expr_edges): Likewise.
5647         (get_cases_for_edge): Likewise.
5648         (make_gimple_switch_edges): Likewise.
5649         (label_to_block_fn): Likewise.
5650         (label_to_block): Likewise.
5651         (make_goto_expr_edges): Likewise.
5652         (make_gimple_asm_edges): Likewise.
5653         (main_block_label): Likewise.
5654         (group_case_labels_stmt): Likewise.
5655         (find_taken_edge_computed_goto): Likewise.
5656         (find_taken_edge_switch_expr): Likewise.
5657         (gimple_verify_flow_info): Likewise.
5658         (gimple_redirect_edge_and_branch): Likewise.
5659         (gimple_switch_label_bb): New function.
5660         (gimple_switch_default_bb): Likewise.
5661         (gimple_switch_edge): Likewise.
5662         (gimple_switch_default_edge): Likewise.
5663         * tree-cfg.h (label_to_block_fn): Remove and replace ...
5664         (label_to_block): ... with this.
5665         (gimple_switch_label_bb): New.
5666         (gimple_switch_default_bb): Likewise.
5667         (gimple_switch_edge): Likewise.
5668         (gimple_switch_default_edge): Likewise.
5669         * tree-cfgcleanup.c (convert_single_case_switch): Use
5670         new gimple functions and pass new argument to label_to_block.
5671         (cleanup_control_flow_bb):
5672         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
5673         cfun argument explicitly.
5674         (make_eh_edges): Likewise.
5675         (redirect_eh_dispatch_edge): Likewise.
5676         (lower_resx): Likewise.
5677         (lower_eh_dispatch): Likewise.
5678         (maybe_remove_unreachable_handlers): Likewise.
5679         (unsplit_eh): Likewise.
5680         (cleanup_empty_eh): Likewise.
5681         (verify_eh_edges): Likewise.
5682         (verify_eh_dispatch_edge): Likewise.
5683         * tree-ssa-dom.c (record_edge_info): Likewise.
5684         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
5685         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
5686         (thread_through_normal_block): Likewise.
5687         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
5688         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
5689         * tree-switch-conversion.c (switch_conversion::collect): Use new
5690         gimple functions.
5691         (switch_conversion::check_final_bb): Likewise.
5692         (switch_conversion::gather_default_values): Pass new argument
5693         to label_to_block.
5694         (switch_conversion::build_constructors): Likewise.
5695         (switch_decision_tree::compute_cases_per_edge): Use new
5696         gimple_switch_edge function.
5697         (switch_decision_tree::analyze_switch_statement): Pass new argument
5698         to label_to_block.
5699         (switch_decision_tree::try_switch_expansion): Use
5700         gimple_switch_default_edge.
5701         * tree-vrp.c (find_switch_asserts): Pass new argument
5702         to label_to_block.
5703         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
5704         (vr_values::simplify_switch_using_ranges): Likewise.
5706 2018-08-27  Richard Biener  <rguenther@suse.de>
5708         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
5709         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
5711         * tree-ssa-sccvn.h (struct vn_pval): New structure.
5712         (struct vn_nary_op_s): Add unwind_to member.  Add
5713         predicated_values flag and put result into a union together
5714         with a linked list of vn_pval.
5715         (struct vn_ssa_aux): Add name member to make maintaining
5716         a map of SSA name to vn_ssa_aux possible.  Remove no longer
5717         needed info, dfsnum, low, visited, on_sccstack, use_processed
5718         and range_info_anti_range_p members.
5719         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
5720         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
5721         New functions.
5722         (vn_valueize): New global.
5723         (vn_context_bb): Likewise.
5724         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
5725         VN_INFO_PTR_INFO): Remove.
5726         * tree-ssa-sccvn.c: ... (rewrite)
5727         (pass_fre::execute): For -O2+ initialize loops and run
5728         RPO VN in optimistic mode (iterating).  For -O1 and -Og
5729         run RPO VN in non-optimistic mode.
5730         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
5731         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
5732         * doc/invoke.texi (sccvn-max-scc-size): Remove.
5733         (rpo-vn-max-loop-depth): Document.
5734         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
5735         when valuezing the VUSE signals we walked out of the region.
5736         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
5737         (phi_translate): Set VN context block to use for availability
5738         lookup.
5739         (compute_avail): Likewise.
5740         (pre_valueize): New function.
5741         (pass_pre::execute): Adjust to the RPO VN API.
5743         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
5744         (propagate_constants_for_unrolling): Remove.
5745         (tree_unroll_loops_completely): Perform value-numbering
5746         on the unrolled bodies loop parent.
5748 2018-08-27  Richard Biener  <rguenther@suse.de>
5750         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
5751         for partial antic compute.
5753 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
5755         PR rtl-optimization/87065
5756         * combine.c (simplify_if_then_else): Formatting fix.
5757         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
5758         check.
5759         (known_cond): Don't return const_true_rtx for vector modes.  Use
5760         CONST0_RTX instead of const0_rtx.  Formatting fixes.
5762 2018-08-27  Martin Liska  <mliska@suse.cz>
5764         PR gcov-profile/87069
5765         * gcov.c (process_file): Record files already processed
5766         and warn about a file being processed multiple times.
5768 2018-08-27  Martin Liska  <mliska@suse.cz>
5770         PR driver/83193
5771         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5772         Set default values for x_aarch64_*_string strings.
5773         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
5774         prefix.  For -mabi do not print '=ABI' in help and use
5775         <option_value> format for -msve-vector-bits and -moverride
5776         options.
5778 2018-08-26  Jeff Law  <law@redhat.com>
5780         * config/mips/frame-header-opt.c: Include "backend.h" rather than
5781         "cfg.h"
5783 2018-08-26  Marek Polacek  <polacek@redhat.com>
5785         PR c++/87029, Implement -Wredundant-move.
5786         * doc/invoke.texi: Document -Wredundant-move.
5788 2018-08-25  Martin Sebor  <msebor@redhat.com>
5790         PR tree-optimization/87059
5791         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5792         to the same type as the other.
5793         * fold-const.c (fold_binary_loc): Assert expectation.
5795 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
5797         * config/darwin.c (machopic_legitimize_pic_address): Clean up
5798         extraneous parentheses, dead code section and formatting.
5800 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
5802         PR c++/87091
5803         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5804         wide enough for jumps in the line-numbering to be visible.
5805         (layout::print_gap_in_line_numbering): New member function.
5806         (layout::calculate_line_spans): When using line numbering, merge
5807         line spans that are only 1 line apart.
5808         (diagnostic_show_locus): When printing line numbers, show gaps in
5809         line numbering directly, rather than printing headers.
5810         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5811         line-numbering with multiple line spans.
5812         (selftest::test_fixit_insert_containing_newline_2): Add test of
5813         line-numbering, in which the spans are close enough to be merged.
5815 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5817         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5818         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
5819         * tree-vrp.c (range_is_nonnull): Remove.
5820         (range_includes_zero_p): Accept value_range instead of min/max.
5821         (extract_range_from_binary_expr_1): Do not early bail on
5822         POINTER_PLUS_EXPR.
5823         Use range_includes_zero_p instead of range_is_nonnull.
5824         (extract_range_from_unary_expr): Use range_includes_zero_p instead
5825         of range_is_nonnull.
5826         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
5827         special case VR_ANTI_RANGE.
5828         (vrp_finalize): Same.
5829         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5830         instead of min/max.
5831         (range_is_nonnull): Remove.
5832         * vr-values.c (vrp_stmt_computes_nonzero): Use
5833         range_includes_zero_p instead of range_is_nonnull.
5834         (extract_range_basic): Pass value_range to range_includes_zero_p
5835         instead of range_is_nonnull.
5837 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
5839         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5840         * emit-rtl.h (rtl_data): Remove return_bnd.
5841         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5842         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5843         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5844         (POINTER_BOUNDS_MODE): Remove definition.
5845         (make_pointer_bounds_mode): Remove.
5846         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5847         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5848         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5849         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5850         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5851         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5852         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5854         * config/i386/i386-modes.def (BND32, BND64): Remove.
5855         * config/i386/i386.c (dbx_register_map): Remove bound registers.
5856         (dbx64_register_map): Ditto.
5857         (svr4_dbx_register_map): Ditto.
5858         (indirect_thunk_bnd_needed): Remove.
5859         (indirect_thunks_bnd_used): Ditto.
5860         (indirect_return_bnd_needed): Ditto.
5861         (indirect_return_via_cx_bnd): Ditto.
5862         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5863         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5864         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
5865         (output_indirect_thunk_function): Remove handling of
5866         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5867         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5868         (ix86_save_reg): Remove handling of crtl->return_bnd.
5869         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5870         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5871         and UNSPEC_BNDLX_ADDR.
5872         (ix86_output_indirect_branch_via_reg): Remove handling of
5873         indirect_thunk_prefix_bnd.
5874         (ix86_output_indirect_branch_via_push): Ditto.
5875         (ix86_output_function_return): Ditto.
5876         (ix86_output_indirect_function_return): Ditto.
5877         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5878         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5879         (CALL_USED_REGISTERS): Ditto.
5880         (REG_ALLOC_ORDER): Update for removal of bound registers.
5881         (HI_REGISTER_NAMES): Ditto.
5882         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5883         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5884         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5885         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5886         (FIRST_PSEUDO_REG): Update.
5887         (BND): Remove mode iterator.
5888         * config/i386/predicates.md (bnd_mem_operator): Remove.
5890 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5892         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5893         vectors.
5895 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5897         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5898         the case in which the permute needs only a single element and
5899         repeats for every vector of the result.  Extend that case to
5900         handle variable-length vectors.
5901         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5903 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5905         PR debug/79342
5906         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5907         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5909 2018-08-24  Richard Biener  <rguenther@suse.de>
5911         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5912         bb_flags_allocated members.
5913         (auto_flag): New RAII class for allocating flags.
5914         (auto_edge_flag): New RAII class for allocating edge flags.
5915         (auto_bb_flag): New RAII class for allocating bb flags.
5916         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5917         flag dynamically.
5918         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5919         in favor of temporarily allocated BB flag.
5920         * hsa-brig.c: Re-order includes.
5921         * hsa-dump.c: Likewise.
5922         * hsa-regalloc.c: Likewise.
5923         * print-rtl.c: Likewise.
5924         * profile-count.c: Likewise.
5926 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5928         PR target/86989
5929         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5930         the TOC register.
5932 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5934         PR 87073/bootstrap
5935         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5936         from wide_int_range_multiplicative_op.
5938 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5940         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5941         "permutaion".
5943 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5945         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5946         to 'expanded'.
5948 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
5950         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5951         full GENERIC expression used for replacement.
5953 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
5955         * tree-vrp.c (abs_extent_range): Remove.
5956         (extract_range_into_wide_ints): Pass wide ints by reference.
5957         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5958         Pass wide ints by reference in all calls to
5959         extract_range_into_wide_ints.
5960         * wide-int-range.cc (wide_int_range_div): New.
5961         * wide-int-range.h (wide_int_range_div): New.
5962         (wide_int_range_includes_zero_p): New.
5963         (wide_int_range_zero_p): New.
5965 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
5967         * config/aarch64/aarch64.md (arches): New enum.
5968         (arch): New enum attr.
5969         (arch_enabled): New attr.
5970         (enabled): Now uses arch_enabled only.
5971         (simd, sve, fp16): Removed attribute.
5972         (fp): Attr now defined in terms of 'arch'.
5973         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5974         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5975         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5976         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5977         attributes into 'arch'.
5978         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5979         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5980         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5981         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5982         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5983         'simd' attribute into 'arch'.
5984         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5985         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5986         Convert use of 'fp' attribute to 'arch'.
5987         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5988         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5989         into 'arch'.
5990         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5991         (different modes) Merge 'fp' and 'simd' into 'arch'.
5992         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5993         'simd' into 'arch'.
5995 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
5997         PR rtl-optimization/87026
5998         * expmed.c (canonicalize_comparison): If we can no longer create
5999         pseudoregisters, don't.
6001 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6003         PR target/86951
6004         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6005         prototype.
6006         * config/arm/arm.c (speculation_barrier_libfunc): New static
6007         variable.
6008         (arm_init_libfuncs): Initialize it.
6009         (arm_emit_speculation_barrier): New function.
6010         * config/arm/arm.md (speculation_barrier): Call
6011         arm_emit_speculation_barrier for architectures that do not have
6012         DSB or ISB.
6013         (speculation_barrier_insn): Only match on Armv7 or later.
6015 2018-08-23  Richard Biener  <rguenther@suse.de>
6017         PR middle-end/87024
6018         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6019         calls.
6021 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6023         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6024         of single-vector TBLs.
6025         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6026         one input is given.
6028 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6030         PR target/85910
6031         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6032         aarch64_evpc_tbl guard.
6034 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6036         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6037         behaviour.
6039 2018-08-22  Martin Sebor  <msebor@redhat.com>
6041         PR middle-end/87052
6042         * tree-pretty-print.c (pretty_print_string): Add argument.
6043         (dump_generic_node): Call to pretty_print_string with string size.
6045 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6047         PR rtl-optimization/86771
6048         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6049         of two SETs into those two SETs, one to be placed at i2, if that SETs
6050         destination is modified between i2 and i3.
6052 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6054         PR tree-optimization/86725
6055         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6056         function.
6057         (vect_analyze_scalar_cycles_1): Check it.
6059 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6061         PR tree-optimization/86725
6062         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6063         an outer loop phi as a double reduction, make sure that the
6064         single user of the phi result is an inner loop phi.
6066 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6068         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6069         grouped stores with gaps to a strided group.
6071 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6073         * tree-vect-stmts.c (get_group_load_store_type)
6074         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6075         first statement in a group.
6077 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6079         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6080         the sequence used in gcc/gcc.c.
6082 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6084         PR other/704
6085         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6086         building it.
6088 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6090         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6091         Darwin10-specific unwinder-shim.
6092         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6093         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6094         New to cater for Darwin10 Rosetta.
6096 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6098         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6099         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6101 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6103         PR bootstrap/81033
6104         PR target/81733
6105         PR target/52795
6106         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6107         (dwarf2out_switch_text_section): Generate a local label for the
6108         second function sub-section and apply it as the second FDE start
6109         label.
6110         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6111         second sub-section start.
6113 2018-08-22  Richard Biener  <rguenther@suse.de>
6115         PR tree-optimization/86988
6116         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6118 2018-08-22  Richard Biener  <rguenther@suse.de>
6120         PR tree-optimization/86945
6121         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6123 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6125         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6126         a comment about how uses of r2 for .sdata2 come about.
6128 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6130         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6132 2018-08-21  Marek Polacek  <polacek@redhat.com>
6134         PR c++/86981, Implement -Wpessimizing-move.
6135         * doc/invoke.texi: Document -Wpessimizing-move.
6137 2018-08-21  Jan Hubicka  <jh@suse.cz>
6139         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6140         * tree.h (is_redundant_typedef): Remove.
6141         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6143 2018-08-21  Jan Hubicka  <jh@suse.cz>
6145         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6146         when possible.
6148 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6150         * expmed.c (extract_low_bits): Reject invalid subregs early.
6152 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6154         PR middle-end/86121
6155         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6156         behaviour.
6158 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6160         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6161         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6162         * config/vxworks.c: Likewise.
6164 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6166         * config/vxworks.c: Set targetm.have_ctors_dtors
6167         if HAVE_INITFINI_ARRAY_SUPPORT.
6168         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6169         if HAVE_INITFINI_ARRAY_SUPPORT.
6171 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6173         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6174         default search path for VxWorks < 7.
6176 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6178         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6179         (get_format_string): Refer to c_getstr.
6181 2018-08-21  Tom de Vries  <tdevries@suse.de>
6183         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6184         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6185         (debuginfo_early_stop): Declare.
6186         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6187         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6188         (debuginfo_early_stop): New function.
6189         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6190         and debuginfo_early_stop.
6191         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6192         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6193         (general_init): Call debuginfo_early_init.
6194         (finalize): Call debuginfo_fini.
6195         (do_compile): Call debuginfo_init.
6196         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6197         -fdump-early-debug.
6198         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6200 2018-08-21  Tom de Vries  <tdevries@suse.de>
6202         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6203         flag_dump_noaddr and flag_dump_unnumbered.
6205 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6207         * wide-int-range.cc (wide_int_range_abs): New.
6208         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6209         * wide-int-range.h (wide_int_range_abs): New.
6210         (wide_int_range_min_max): New.
6211         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6212         case to call wide_int_range_abs.
6213         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6214         (extract_range_from_abs_expr): Delete.
6216 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6218         PR target/87033
6219         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6220         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6221         for indexed loads.
6223 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6224             Jeff Law <law@redhat.com>
6226         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6227         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6229 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6231         PR other/84889
6232         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6233         (decl_attributes): Likewise.
6234         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6235         instance.
6236         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6237         * diagnostic-core.h (class auto_diagnostic_group): New class.
6238         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6239         (diagnostic_report_diagnostic): Handle the first diagnostics within
6240         a group.
6241         (emit_diagnostic): Add auto_diagnostic_group instance.
6242         (inform): Likewise.
6243         (inform_n): Likewise.
6244         (warning): Likewise.
6245         (warning_at): Likewise.
6246         (warning_n): Likewise.
6247         (pedwarn): Likewise.
6248         (permerror): Likewise.
6249         (error): Likewise.
6250         (error_n): Likewise.
6251         (error_at): Likewise.
6252         (sorry): Likewise.
6253         (fatal_error): Likewise.
6254         (internal_error): Likewise.
6255         (internal_error_no_backtrace): Likewise.
6256         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6257         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6258         * diagnostic.h (struct diagnostic_context): Add fields
6259         "diagnostic_group_nesting_depth",
6260         "diagnostic_group_emission_count", "begin_group_cb",
6261         "end_group_cb".
6262         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6263         Add auto_diagnostic_group instance(s).
6264         (find_explicit_erroneous_behavior): Likewise.
6265         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6266         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6267         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6268         (gimplify_va_arg_expr): Likewise.
6269         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6270         (HSA_SORRY_AT): Likewise.
6271         * ipa-devirt.c (compare_virtual_tables): Likewise.
6272         (warn_odr): Likewise.
6273         * multiple_target.c (expand_target_clones): Likewise.
6274         * opts-common.c (cmdline_handle_error): Likewise.
6275         * reginfo.c (globalize_reg): Likewise.
6276         * substring-locations.c (format_warning_n_va): Likewise.
6277         * tree-inline.c (expand_call_inline): Likewise.
6278         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6279         * tree-ssa-loop-niter.c
6280         (do_warn_aggressive_loop_optimizations): Likewise.
6281         * tree-ssa-uninit.c (warn_uninit): Likewise.
6282         * tree.c (warn_deprecated_use): Likewise.
6284 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6286         PR target/87014
6287         * config/i386/i386.md (eh_return): Always update EH return
6288         address in word_mode.
6290 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6292         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6293         TARGET_SPLIT_COMPLEX_ARG is defined.
6295 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6297         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6299 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6301         PR target/86984
6302         * expr.c (expand_assignment): Assert that bitpos is positive.
6303         (store_field): Likewise
6304         (expand_expr_real_1): Make sure that bitpos is positive.
6305         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6306         integer overflow.
6308 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6310         * Makefile.in (CPP_ID_DATA_H): Delete.
6311         (CPP_INTERNAL_H): Don't add it.
6312         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6313         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6315 2018-08-20  Richard Biener  <rguenther@suse.de>
6317         PR tree-optimization/78655
6318         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6319         pointer + offset nonnull if either operand is nonnull work.
6321 2018-08-20  Tom de Vries  <tdevries@suse.de>
6323         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6324         unless the referenced die describes the added property using
6325         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6326         Otherwise, add a DW_AT_location to the referenced die.
6328 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6330         PR target/86994
6331         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6332         register_operand when calling ix86_set_reg_reg_cost.
6333         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6334         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6335         predicate and to 1 otherwise.
6337 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6339         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6340         emit a diagnostic that it is not supported and reset the option.
6341         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6342         supported and consume the option.  (ASM_FINAL_SPEC): New.
6344 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6346         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6347         a sentence.
6349 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6351         C-SKY port: Documentation
6353         * doc/extend.texi (C-SKY Function Attributes): New section.
6354         * doc/invoke.texi (Option Summary): Add C-SKY options.
6355         (C-SKY Options): New section.
6356         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6358 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6359             Huibin Wang  <huibin_wang@c-sky.com>
6360             Sandra Loosemore  <sandra@codesourcery.com>
6361             Chung-Lin Tang  <cltang@codesourcery.com>
6363         C-SKY port: Backend implementation
6365         * config/csky/*: New.
6366         * common/config/csky/*: New.
6368 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6369             Huibin Wang  <huibin_wang@c-sky.com>
6370             Sandra Loosemore  <sandra@codesourcery.com>
6371             Chung-Lin Tang  <cltang@codesourcery.com>
6372             Andrew Jenner  <andrew@codesourcery.com>
6374         C-SKY port: Configury
6376         * config.gcc (csky-*-*): New.
6377         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6378         * configure: Regenerated.
6380 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6382         * dump-context.h: Include "dumpfile.h".
6383         (dump_context::dump_printf_va): Convert final param from va_list
6384         to va_list *.  Convert from ATTRIBUTE_PRINTF to
6385         ATTRIBUTE_GCC_DUMP_PRINTF.
6386         (dump_context::dump_printf_loc_va): Likewise.
6387         * dumpfile.c: Include "stringpool.h".
6388         (make_item_for_dump_printf_va): Delete.
6389         (make_item_for_dump_printf): Delete.
6390         (class dump_pretty_printer): New class.
6391         (dump_pretty_printer::dump_pretty_printer): New ctor.
6392         (dump_pretty_printer::emit_items): New member function.
6393         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6394         function.
6395         (dump_pretty_printer::emit_item): New member function.
6396         (dump_pretty_printer::stash_item): New member function.
6397         (dump_pretty_printer::format_decoder_cb): New member function.
6398         (dump_pretty_printer::decode_format): New member function.
6399         (dump_context::dump_printf_va): Reimplement in terms of
6400         dump_pretty_printer.
6401         (dump_context::dump_printf_loc_va): Convert final param from va_list
6402         to va_list *.
6403         (dump_context::begin_scope): Reimplement call to
6404         make_item_for_dump_printf.
6405         (dump_printf): Update for change to dump_printf_va.
6406         (dump_printf_loc): Likewise.
6407         (selftest::test_capture_of_dump_calls): Convert "stmt" from
6408         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
6409         with %T, %E, and %G.
6410         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6411         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6412         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6413         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6414         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6415         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6416         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6417         within a dump_printf_loc call to "%wu".
6418         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6419         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
6420         missing space after "=".
6421         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6422         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6423         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6424         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6425         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
6426         duplicate "vectorized" from message.
6428 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
6430         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6431         polyNxK_t element's TYPE_STRING_FLAG.
6433 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6435         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6436         (they were unnamed before).  Fix comments.
6438 2018-08-17  Nathan Sidwell  <nathan@acm.org>
6440         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6442 2018-08-17  Richard Biener  <rguenther@suse.de>
6444         PR tree-optimization/86841
6445         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6447 2018-08-17  Martin Liska  <mliska@suse.cz>
6449         * common.opt: Remove Warn, Init and Report for options with
6450         Ignore/Deprecated flag. Warning is done automatically for
6451         Deprecated flags.
6452         * config/i386/i386.opt: Likewise.
6453         * config/ia64/ia64.opt: Likewise.
6454         * config/rs6000/rs6000.opt: Likewise.
6455         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6456         Remove usage of flag_check_pointer_bounds.
6457         * lto-wrapper.c (merge_and_complain): Do not handle
6458         OPT_fcheck_pointer_bounds.
6459         (append_compiler_options): Likewise.
6460         * opt-functions.awk: Do not handle Deprecated.
6461         * optc-gen.awk: Check that Var, Report and Init are not
6462         used for an option with Ignore/Deprecated flag.
6463         * opts-common.c (decode_cmdline_option): Do not report
6464         CL_ERR_DEPRECATED.
6465         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6466         options.
6467         * opts.h (struct cl_option): Remove cl_deprecated flag.
6468         (CL_ERR_DEPRECATED): Remove error enum value.
6470 2018-08-17  Richard Biener  <rguenther@suse.de>
6472         PR middle-end/86505
6473         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6474         across a va-arg-pack using call adjust its return value accordingly.
6476 2018-08-16  Martin Sebor  <msebor@redhat.com>
6478         PR tree-optimization/86853
6479         * gimple-ssa-sprintf.c (struct format_result): Rename member.
6480         (struct fmtresult): Add member and initialize it in ctors.
6481         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
6482         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
6483         (format_directive): Set POSUNDER4K when MAYFAIL is set.
6484         (parse_directive): Handle %C same as %c.
6485         (sprintf_dom_walker::compute_format_length): Adjust.
6486         (is_call_safe): Adjust.
6488 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6490         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
6491         for determining how to count the elements.
6492         * builtins.h (c_strlen): Adjust prototype.
6493         * expr.c (string_constant): Add new parameter mem_size.
6494         Set *mem_size appropriately.
6495         * expr.h (string_constant): Adjust protoype.
6496         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6497         * gimple-fold.h (get_range_strlen): Adjust prototype.
6498         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6499         (format_string): Call get_string_length with eltsize.
6501 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6503         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6504         to emit the span, rather than setting it as the prefix.
6506 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6508         * diagnostic-show-locus.c (layout::start_annotation_line): Add
6509         "margin_char" parameter, defaulting to space.  Use it in place
6510         of pp_space for the initial part of the margin.
6511         (layout::print_leading_fixits): Use '+' when filling the margin
6512         of line-insertion fix-it hints.
6514 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6516         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
6517         Delete.
6519 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6521         * config/rs6000/altivec.md: Don't set length attribute to the default
6522         value.
6523         * config/rs6000/darwin.md: Ditto.
6524         * config/rs6000/dfp.md: Ditto.
6525         * config/rs6000/htm.md: Ditto.
6526         * config/rs6000/rs6000.md: Ditto.
6527         * config/rs6000/sync.md: Ditto.
6528         * config/rs6000/vsx.md: Ditto.
6530 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6532         * config/rs6000/altivec.md: Don't set length attribute to the default
6533         value, for branch instructions.
6534         * config/rs6000/darwin.md: Ditto.
6535         * config/rs6000/rs6000.md: Ditto.
6537 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6539         * config/rs6000/rs6000.md (length): Always define as const_int 4.
6540         (unnamed conditional branch define_insn): Set length to 4 or 8
6541         depending on offset.
6542         (<bd>_<mode>): Similar, for alternative 0.
6543         (<bd>tf_<mode>): Ditto.
6545 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6547         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
6549 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
6551         * doc/rtl.texi: Replace old RTX class names with new names.
6554 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
6556         * expmed.h (canonicalize_comparison): New declaration.
6557         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
6558         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6559         * optabs.c (prepare_cmp_insn): Likewise.
6560         * rtl.h (unsigned_condition_p): New function which checks if a
6561         comparison operator is unsigned.
6563 2018-08-16  Nathan Sidwell  <nathan@acm.org>
6565         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
6566         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
6568 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6570         PR target/84711
6571         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
6572         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
6573         (mov<mov>): ..this and enable unconditionally.
6575 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6577         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
6579 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
6581         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
6582         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
6583         "Common" with "Target".
6585 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6587         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
6588         * doc/invoke.texi (mmitigate-rop): Remove.
6589         * config/i386/i386.c: Do not include "regrename.h".
6590         (ix86_rop_should_change_byte_p, reg_encoded_number)
6591         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
6592         Remove.
6593         (ix86_reorg): Remove call to ix86_mitigate_rop.
6594         * config/i386/i386.md (attr "modrm_class"): Remove.
6595         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
6596         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
6597         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
6599 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
6601         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
6602         allow folding of mergeh() and mergel() for the float and double types.
6603         (fold_mergehl_helper): Rework to handle building a permute tree
6604         for float vectors.
6606 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6608         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
6609         for TARGET_SSE.
6611 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
6613         * common.opt (fdiagnostics-show-labels): New option.
6614         * diagnostic-show-locus.c (class layout_range): Add field
6615         "m_label".
6616         (class layout): Add field "m_show_labels_p".
6617         (layout_range::layout_range): Add param "label" and use it to
6618         initialize m_label.
6619         (make_range): Pass in NULL for new "label" param of layout_range's
6620         ctor.
6621         (layout::layout): Initialize m_show_labels_p.
6622         (layout::maybe_add_location_range): Pass in loc_range->m_label
6623         when constructing layout_range instances.
6624         (struct line_label): New struct.
6625         (layout::print_any_labels): New member function.
6626         (layout::print_line): Call it if label-printing is enabled.
6627         (selftest::test_one_liner_labels): New test.
6628         (selftest::test_diagnostic_show_locus_one_liner): Call it.
6629         * diagnostic.c (diagnostic_initialize): Initialize
6630         context->show_labels_p.
6631         * diagnostic.h (struct diagnostic_context): Add field
6632         "show_labels_p".
6633         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6634         -fno-diagnostics-show-labels.
6635         * dwarf2out.c (gen_producer_string): Add
6636         OPT_fdiagnostics_show_labels to the ignored options.
6637         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
6638         param.
6639         (gcc_rich_location::maybe_add_expr): Likewise.
6640         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
6641         label" param, defaulting to NULL.
6642         (gcc_rich_location::add_expr): Add "label" param.
6643         (gcc_rich_location::maybe_add_expr): Likewise.
6644         (class text_range_label): New class.
6645         (class range_label_for_type_mismatch): New class.
6646         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
6647         of format_warning_va.
6648         (fmtwarn_n): Likewise for new params of format_warning_n_va.
6649         * lto-wrapper.c (merge_and_complain): Add
6650         OPT_fdiagnostics_show_labels to the "pick one setting" options.
6651         (append_compiler_options): Likewise to the dropped options.
6652         (append_diag_options): Likewise to the passed-on options.
6653         * opts.c (common_handle_option): Handle the new option.
6654         * selftest-diagnostic.c
6655         (test_diagnostic_context::test_diagnostic_context): Enable
6656         show_labels_p.
6657         * substring-locations.c: Include "gcc-rich-location.h".
6658         (format_warning_n_va): Add "fmt_label" and "param_label" params
6659         and use them as appropriate.
6660         (format_warning_va): Add "fmt_label" and "param_label" params,
6661         passing them on to format_warning_n_va.
6662         (format_warning_at_substring): Likewise.
6663         (format_warning_at_substring_n): Likewise.
6664         * substring-locations.h (format_warning_va): Add "fmt_label" and
6665         "param_label" params.
6666         (format_warning_n_va): Likewise.
6667         (format_warning_at_substring): Likewise.
6668         (format_warning_at_substring_n): Likewise.
6669         * toplev.c (general_init): Initialize global_dc->show_labels_p.
6671 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
6673         PR testsuite/86519
6674         * builtins.c (expand_builtin_memcmp): Do not expand the call
6675         when overflow is detected.
6677 2018-08-15  Martin Sebor  <msebor@redhat.com>
6679         PR tree-optimization/71625
6680         * config/aarch64/aarch64-builtins.c
6681         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
6683 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
6685         * config/s390/s390.c (s390_reorg): Remove loop.
6687 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6689         * config/darwin.c
6690          (darwin_function_switched_text_sections): Delete.
6691         * gcc/config/darwin.h
6692          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
6694 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6696         PR target/81685
6697         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
6698         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
6699         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
6701 2018-08-15  Martin Liska  <mliska@suse.cz>
6703         PR tree-optimization/86925
6704         * predict.c (expr_expected_value_1): When taking
6705         later predictor, assign also probability.
6706         Use fold_build2_initializer_loc in order to fold
6707         the expression in -frounding-math.
6709 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
6711         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
6712         patterns.
6713         (expand_vec_perm_1): Try the new method.
6715 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6717         PR target/86547
6718         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
6719         Check whether lra_live_max_point is 0 before dividing.
6721 2018-08-14  Martin Sebor  <msebor@redhat.com>
6723         PR tree-optimization/86650
6724         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
6725         (vrp_prop::check_mem_ref): Same.
6727 2018-08-13  Liu Hao <lh_mouse@126.com>
6729         * pretty-print.c (eat_esc_sequence): Swap the foreground and
6730         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
6731         and clear it thereafter, as it only works for DBCS.
6733 2018-08-13  Liu Hao <lh_mouse@126.com>
6735         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
6736         handle returned by _get_osf_handle().
6738 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6740         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
6741         for folding vec_perm.
6743 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6745         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
6746         Add support for gimple-folding of vec_pack() and vec_unpack()
6747         intrinsics.
6749 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
6751         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
6752         vec_xst variants to the list.
6753         (rs6000_gimple_fold_builtin): Add support for folding unaligned
6754         vector loads and stores.
6756 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
6758         * config.gcc (rs6000-ibm-aix4.x): Delete.
6759         (rs6000-ibm-aix5.1): Delete.
6760         (rs6000-ibm-aix5.2): Delete.
6761         (rs6000-ibm-aix5.3): Delete.
6762         * config/rs6000/aix43.h: Delete.
6763         * config/rs6000/aix51.h: Delete.
6764         * config/rs6000/aix52.h: Delete.
6765         * config/rs6000/t-aix43: Delete.
6767 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
6769         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6770         New function.
6771         (s390_decompose_address): Factor out constant pool ref
6772         decomposition.
6774 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6776         * config/nds32/nds32-predicates.c
6777         (nds32_can_use_bclr_p): Change return type as bool.
6778         (nds32_can_use_bset_p): Ditto.
6779         (nds32_can_use_btgl_p): Ditto.
6780         (nds32_can_use_bitci_p): Ditto.
6781         * config/nds32/nds32-protos.h
6782         (nds32_can_use_bclr_p): Change declaration.
6783         (nds32_can_use_bset_p): Ditto.
6784         (nds32_can_use_btgl_p): Ditto.
6785         (nds32_can_use_bitci_p): Ditto.
6787 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6789         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6790         Support -msched-prolog-epilog option.
6791         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6793 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6795         * common/config/nds32/nds32-common.c
6796         (nds32_option_optimization_table): Enalbe -malways-align.
6798 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6800         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6801         extra_headers.
6802         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6803         OPT_misr_secure_ case.
6804         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6805         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6806         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6807         secure attribute.
6808         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6809         (nds32_isr_info): New field security_level.
6810         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6811         * config/nds32/nds32.md (return_internal): Consider critical attribute.
6812         * config/nds32/nds32.opt (misr-secure): New option.
6813         * config/nds32/nds32_init.inc: New file.
6814         * config/nds32/nds32_isr.h: New file.
6816 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
6818         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6819         Update comment for atomic instructions.
6820         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6821         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6822         Remove.
6823         (atomic_loaddi): Revise fence expansion to only emit fence prior to
6824         load for __ATOMIC_SEQ_CST model.
6825         (atomic_loaddi_1): Remove float register target.
6826         (atomic_storedi): Handle CONST_INT values.
6827         (atomic_storedi_1): Remove float register source.  Add special case
6828         for zero value.
6829         (memory_barrier): New expander and insn.
6831 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
6833         PR tree-optimization/86835
6834         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6835         new_stmt after def_gsi, make sure to insert new_square_stmt after
6836         that stmt, not 2 stmts before it.
6838 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
6840         PR target/82418
6841         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6842         instead of SWI48.
6844 2018-08-10  Martin Liska  <mliska@suse.cz>
6846         PR target/83610
6847         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6848         function type.
6849         * builtins.c (expand_builtin_expect_with_probability):
6850         New function.
6851         (expand_builtin_expect_with_probability): New function.
6852         (build_builtin_expect_predicate): Add new argumnet probability
6853         for BUILT_IN_EXPECT_WITH_PROBABILITY.
6854         (fold_builtin_expect):
6855         (fold_builtin_2):
6856         (fold_builtin_3):
6857         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6858         * builtins.h (fold_builtin_expect): Set new argument.
6859         * doc/extend.texi: Document __builtin_expect_with_probability.
6860         * doc/invoke.texi: Likewise.
6861         * gimple-fold.c (gimple_fold_call): Pass new argument.
6862         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6863         also BUILT_IN_EXPECT_WITH_PROBABILITY.
6864         * predict.c (get_predictor_value): New function.
6865         (expr_expected_value): Add new argument probability. Assume
6866         that predictor and probability are always non-null.
6867         (expr_expected_value_1): Likewise.  For __builtin_expect and
6868         __builtin_expect_with_probability set probability.  Handle
6869         combination in binary expressions.
6870         (tree_predict_by_opcode): Simplify code by simply calling
6871         get_predictor_value.
6872         (pass_strip_predict_hints::execute): Add handling of
6873         BUILT_IN_EXPECT_WITH_PROBABILITY.
6874         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6875         new predictor.
6876         * tree.h (DECL_BUILT_IN_P): New function.
6878 2018-08-10  Martin Liska  <mliska@suse.cz>
6880         PR tree-optimization/85799
6881         * passes.def: Add argument for pass_strip_predict_hints.
6882         * predict.c (class pass_strip_predict_hints): Add new argument
6883         early_p.
6884         (strip_predictor_early): New function.
6885         (pass_strip_predict_hints::execute): Call the function to
6886         strip predictors.
6887         (strip_predict_hints): New function.
6888         * predict.def: Fix comment.
6890 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6892         * Makefile.in: Clarify which tm.texi to copy over to assert the
6893         right to grant a GFDL license for all.
6895 2018-08-09  Jeff Law  <law@redhat.com>
6897         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6898         unused variable.
6900 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6902         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6903         prototype.
6905 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6907         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6908         reductions for variable-length vectors.
6910 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6912         PR other/84889
6913         * common.opt (fdiagnostics-show-line-numbers): New option.
6914         * diagnostic-show-locus.c (class layout): Add fields
6915         "m_show_line_numbers_p" and "m_linenum_width";
6916         (num_digits): New function.
6917         (test_num_digits): New function.
6918         (layout::layout): Initialize new fields.  Update m_x_offset
6919         logic to handle any left margin.
6920         (layout::print_source_line): Print line number when requested.
6921         (layout::start_annotation_line): New member function.
6922         (layout::print_annotation_line): Call it.
6923         (layout::print_leading_fixits): Likewise.
6924         (layout::print_trailing_fixits): Likewise.  Update calls to
6925         move_to_column for new parameter.
6926         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6927         it to potentially call start_annotation_line.
6928         (layout::show_ruler): Call start_annotation_line.
6929         (selftest::test_line_numbers_multiline_range): New selftest.
6930         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6931         and selftest::test_line_numbers_multiline_range.
6932         * diagnostic.c (diagnostic_initialize): Initialize
6933         show_line_numbers_p.
6934         * diagnostic.h (struct diagnostic_context): Add field
6935         "show_line_numbers_p".
6936         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6937         -fno-diagnostics-show-line-numbers.
6938         * dwarf2out.c (gen_producer_string): Add
6939         OPT_fdiagnostics_show_line_numbers to the ignored options.
6940         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6941         one setting" options.
6942         (append_compiler_options): Likewise to the dropped options.
6943         (append_diag_options): Likewise to the passed-on options.
6944         * opts.c (common_handle_option): Handle the new option.
6945         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6947 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6949         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6950         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
6951         third argument of type "const signed char" to descriptions of
6952         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6953         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6954         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6955         __builtin_bcdsub_ov functions.
6957 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6959         PR tree-optimization/86858
6960         * tree-vect-loop.c (vect_is_simple_reduction): Restore
6961         flow_bb_inside_loop_p calls.
6963 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6965         PR tree-optimization/86871
6966         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6967         instead of gimple_assign_lhs.
6969 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
6971         PR target/86887
6972         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6973         register constraint to operand 0.
6974         (add<mode>3_carryinC): Likewise.
6975         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6977 2018-08-09  Martin Liska  <mliska@suse.cz>
6979         PR c/86895
6980         * common.opt: Remove extra line.
6982 2018-08-09  Martin Liska  <mliska@suse.cz>
6984         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6985         at the end of a line, make first letter capital and end up
6986         a sentence with a dot.
6987         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6988         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6989         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6990         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6991         (PARAM_MAX_ISL_OPERATIONS): Likewise.
6992         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6993         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6994         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6995         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6996         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6997         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6998         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6999         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7000         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7001         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7002         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7003         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7004         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7006 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7008         PR target/84332
7009         * config/s390/s390.c (s390_option_override_internal): Reduce the
7010         stack-clash-protection-probe-interval param if it would be too big
7011         for z900.
7013 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7015         PR target/46179
7016         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7017         * config/m68k/m68k.c (handle_move_double): Don't call
7018         m68k_final_prescan_insn.
7019         (m68k_adjust_decorated_operand): Renamed from
7020         m68k_final_prescan_insn, remove first and third operand and
7021         simplify.
7022         (print_operand): Call it.
7023         (print_operand_address): Call it.
7025 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7027         * diagnostic.c (diagnostic_report_current_module): Use
7028         linemap_included_from & linemap_included_from_linemap.
7030 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7032         * config/aarch64/aarch64-cores.def: Add phecda core.
7033         * config/aarch64/aarch64-tune.md: Regenerate.
7034         * doc/invoke.texi: Add phecda core.
7036 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7038         PR target/85295
7039         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7040         definitions.
7041         * config/s390/s390.md ("movti"): Add more alternatives for
7042         constant to GPR copies.
7044 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7046         * config/s390/s390.c: Fix whitespace damage throughout the file.
7047         * config/s390/s390.h: Likewise.
7048         * config/s390/tpf.h: Likewise.
7050 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7052         * config/s390/s390.c (s390_loadrelative_operand_p):
7053         Remove TARGET_CPU_ZARCH usages.
7054         (s390_rtx_costs): Likewise.
7055         (s390_legitimate_constant_p): Likewise.
7056         (s390_cannot_force_const_mem): Likewise.
7057         (legitimate_reload_constant_p): Likewise.
7058         (s390_preferred_reload_class): Likewise.
7059         (legitimize_pic_address): Likewise.
7060         (legitimize_tls_address): Likewise.
7061         (s390_split_branches): Removed.
7062         (s390_add_execute): Removed.
7063         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7064         (s390_mainpool_start): Likewise.
7065         (s390_mainpool_finish): Likewise.
7066         (s390_mainpool_cancel): Removed.
7067         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7068         (s390_chunkify_cancel): Likewise.
7069         (s390_return_addr_rtx): Likewise.
7070         (s390_register_info): Remove split_branches_pending_p uages.
7071         (s390_optimize_register_info): Likewise.
7072         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7073         split_branches_pending_p usages.
7074         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7075         (s390_load_got): Likewise.
7076         (s390_expand_split_stack_prologue): Likewise.
7077         (output_asm_nops): Likewise.
7078         (s390_function_profiler): Likewise.
7079         (s390_emit_call): Likewise.
7080         (s390_conditional_register_usage): Likewise.
7081         (s390_optimize_prologue): Likewise.
7082         (s390_reorg): Remove TARGET_CPU_ZARCH and
7083         split_branches_pending_p usages.
7084         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7085         usages.
7086         (s390_output_indirect_thunk_function): Likewise.
7087         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7088         (TARGET_CPU_ZARCH_P): Removed.
7089         (struct machine_function): Remove split_branches_pending_p.
7090         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7092 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7094         * common/config/s390/s390-common.c (processor_flags_table):
7095         Remove flags.
7096         * config.gcc: Remove with_arch/with_tune support.
7097         * config/s390/2064.md: Remove cpu attribute comparisons.
7098         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7099         Remove MTN.
7100         * config/s390/linux.h (ASM_SPEC):
7101         Remove -march support.
7102         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7103         Use a table to get an arch level.
7104         * config/s390/s390-opts.h (enum processor_type):
7105         Remove enum values.
7106         * config/s390/s390.c
7107         (processor_table): Remove entries, add arch_level values.
7108         (s390_issue_rate): Remove cases.
7109         (s390_option_override): Adjust
7110         s390_option_override_internal() call.
7111         (s390_option_override_internal): Remove deprecation warning.
7112         (s390_valid_target_attribute_tree): Adjust
7113         s390_option_override_internal() call.
7114         * config/s390/s390.h (struct s390_processor):
7115         Share with s390-c.c, add arch_level field.
7116         * config/s390/s390.md:
7117         Remove occurrences in cpu attribute.
7118         * config/s390/s390.opt: Remove -march/-mtune support.
7119         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7120         * doc/invoke.texi: Remove deprecation warning.
7122 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7124         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7125         global.
7126         (qdf24xx_tunings): Set vector cost structure to
7127         qdf24xx_vector_cost.
7129         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7130         <register_sextend>: Set to 3.
7132 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7134         PR target/86838
7135         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7136         * config/aarch64/aarch64-simd.md
7137         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7138         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7139         (aarch64_frecpx<mode>): ...this new pattern.
7140         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7141         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7143 2018-08-07  Martin Liska  <mliska@suse.cz>
7145         PR middle-end/83023
7146         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7147         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7148         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7149         * doc/extend.texi: Document that malloc attribute adds
7150         hit to compiler.
7152 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7154         PR target/86785
7155         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7156         Define to speculation_safe_value_not_needed.
7158 2018-08-06  Jeff Law  <law@redhat.com>
7160         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7161         the vr_values instance to cprop_into_stmt.
7162         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7163         (cprop_operand): Also query EVRP to determine if OP is a constant.
7165 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7167         * diagnostic.c (diagnostic_report_current_module): Reroll
7168         included-at loop.  Translate text.
7170 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7172         * function-tests.c (selftest::test_expansion_to_rtl): Call
7173         free_after_compilation.
7175 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7177         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7179 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7181         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7182         loops with memory block operations from getting unrolled.
7184 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7186         PR target/86807
7187         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7188         Define to speculation_safe_value_not_needed.
7190 2018-08-06  Jeff Law  <law@redhat.com>
7192         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7193         assert.
7195 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7197         PR target/86662
7198         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7199         with all enabled __intN types.
7201         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7203 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7205         * alias.c (record_set): Check for clobber high.
7206         * cfgexpand.c (expand_gimple_stmt): Likewise.
7207         * combine-stack-adj.c (single_set_for_csa): Likewise.
7208         * combine.c (find_single_use_1): Likewise.
7209         (set_nonzero_bits_and_sign_copies): Likewise.
7210         (get_combine_src_dest): Likewise.
7211         (is_parallel_of_n_reg_sets): Likewise.
7212         (try_combine): Likewise.
7213         (record_dead_and_set_regs_1): Likewise.
7214         (reg_dead_at_p_1): Likewise.
7215         (reg_dead_at_p): Likewise.
7216         * dce.c (deletable_insn_p): Likewise.
7217         (mark_nonreg_stores_1): Likewise.
7218         (mark_nonreg_stores_2): Likewise.
7219         * df-scan.c (df_find_hard_reg_defs): Likewise.
7220         (df_uses_record): Likewise.
7221         (df_get_call_refs): Likewise.
7222         * dwarf2out.c (mem_loc_descriptor): Likewise.
7223         * haifa-sched.c (haifa_classify_rtx): Likewise.
7224         * ira-build.c (create_insn_allocnos): Likewise.
7225         * ira-costs.c (scan_one_insn): Likewise.
7226         * ira.c (equiv_init_movable_p): Likewise.
7227         (rtx_moveable_p): Likewise.
7228         (interesting_dest_for_shprep): Likewise.
7229         * jump.c (mark_jump_label_1): Likewise.
7230         * postreload-gcse.c (record_opr_changes): Likewise.
7231         * postreload.c (reload_cse_simplify): Likewise.
7232         (struct reg_use): Add source expr.
7233         (reload_combine): Check for clobber high.
7234         (reload_combine_note_use): Likewise.
7235         (reload_cse_move2add): Likewise.
7236         (move2add_note_store): Likewise.
7237         * print-rtl.c (print_pattern): Likewise.
7238         * recog.c (decode_asm_operands): Likewise.
7239         (store_data_bypass_p): Likewise.
7240         (if_test_bypass_p): Likewise.
7241         * regcprop.c (kill_clobbered_value): Likewise.
7242         (kill_set_value): Likewise.
7243         * reginfo.c (reg_scan_mark_refs): Likewise.
7244         * reload1.c (maybe_fix_stack_asms): Likewise.
7245         (eliminate_regs_1): Likewise.
7246         (elimination_effects): Likewise.
7247         (mark_not_eliminable): Likewise.
7248         (scan_paradoxical_subregs): Likewise.
7249         (forget_old_reloads_1): Likewise.
7250         * reorg.c (find_end_label): Likewise.
7251         (try_merge_delay_insns): Likewise.
7252         (redundant_insn): Likewise.
7253         (own_thread_p): Likewise.
7254         (fill_simple_delay_slots): Likewise.
7255         (fill_slots_from_thread): Likewise.
7256         (dbr_schedule): Likewise.
7257         * resource.c (update_live_status): Likewise.
7258         (mark_referenced_resources): Likewise.
7259         (mark_set_resources): Likewise.
7260         * rtl.c (copy_rtx): Likewise.
7261         * rtlanal.c (reg_referenced_p): Likewise.
7262         (single_set_2): Likewise.
7263         (noop_move_p): Likewise.
7264         (note_stores): Likewise.
7265         * sched-deps.c (sched_analyze_reg): Likewise.
7266         (sched_analyze_insn): Likewise.
7268 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7270         * cse.c (invalidate_reg): New function extracted from...
7271         (invalidate): ...here.
7272         (canonicalize_insn): Check for clobber high.
7273         (invalidate_from_clobbers): invalidate clobber highs.
7274         (invalidate_from_sets_and_clobbers): Likewise.
7275         (count_reg_usage): Check for clobber high.
7276         (insn_live_p): Likewise.
7277         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7278         (cselib_invalidate_regno): Check for clobber in setter.
7279         (cselib_invalidate_rtx): Pass through setter.
7280         (cselib_invalidate_rtx_note_stores):
7281         (cselib_process_insn): Check for clobber high.
7282         * cselib.h (cselib_invalidate_rtx): Add operand.
7284 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7286         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7287         (mark_not_eliminable): Likewise.
7288         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7289         * lra-lives.c (process_bb_lives): Check for clobber high.
7290         * lra.c (new_insn_reg): Remember clobber highs.
7291         (collect_non_operand_hard_regs): Check for clobber high.
7292         (lra_set_insn_recog_data): Likewise.
7293         (add_regs_to_insn_regno_info): Likewise.
7294         (lra_update_insn_regno_info): Likewise.
7296 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7298         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7299         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7301 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7303         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7304         (copy_insn_1): Likewise.
7305         (gen_hard_reg_clobber_high): New gen function.
7306         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7307         * genemit.c (gen_exp): Likewise.
7308         (gen_emit_seq): Pass through info.
7309         (gen_insn): Check for CLOBBER_HIGH.
7310         (gen_expand): Pass through info.
7311         (gen_split): Likewise.
7312         (output_add_clobbers): Likewise.
7313         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7314         (remove_clobbers): Likewise.
7315         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7317 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7319         * doc/rtl.texi (clobber_high): Add.
7320         (parallel): Add in clobber high
7321         * rtl.c (rtl_check_failed_code3): Add function.
7322         * rtl.def (CLOBBER_HIGH): Add expression.
7323         * rtl.h (RTL_CHECKC3): Add macro.
7324         (rtl_check_failed_code3): Add declaration.
7325         (XC3EXP): Add macro.
7327 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7329         PR target/86386
7330         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7331         cfun->machine->max_used_stack_alignment if needed.
7333 2018-08-04  Martin Sebor  <msebor@redhat.com>
7335         PR tree-optimization/86571
7336         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7337         NaN output to 4.
7339 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7341         PR target/86799
7342         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7343         Define.
7345 2018-08-03  Jeff Law  <law@redhat.com>
7347         PR target/86795
7348         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7349         Define to speculation_safe_value_not_needed.
7351 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7353         * doc/gcov.texi (-x): Remove duplicate "to".
7354         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7355         (-Wif-not-aligned): Remove duplicate "is".
7356         (-flto): Remove duplicate "the".
7357         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7358         duplicate "v5.00.b".
7359         (MSP430 Options): Remove duplicate "and" from the description
7360         of "-mgprel-sec=regexp".
7361         (x86 Options): Remove duplicate copies of "vmldLog102" and
7362         vmlsLog104 from description of "-mveclibabi=type".
7364 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7366         * internal-fn.h (first_commutative_argument): Declare.
7367         * internal-fn.c (first_commutative_argument): New function.
7368         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7369         restrictions for pattern statements.  Use first_commutative_argument
7370         to look for commutative operands in calls to internal functions.
7372 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7374         * Makefile.in (wide-int-range.o): New.
7375         * tree-vrp.c: Move all the wide_int_* functions to...
7376         * wide-int-range.cc: ...here.
7377         * tree-vrp.h: Move all the wide_int_* prototypes to...
7378         * wide-int-range.h: ...here.
7380 2018-08-03  Tom de Vries  <tdevries@suse.de>
7382         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7383         UI_NONE.
7384         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7385         * except.c (output_function_exception_table): Do early exit if
7386         targetm_common.except_unwind_info (&global_options) == UI_NONE.
7388 2018-08-03  Martin Liska  <mliska@suse.cz>
7390         * predict.c (dump_prediction): Change to 2 digits
7391         in fraction part.
7393 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7395         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7396         neon_dup_q to...
7397         (falkor_am_1_gtov_gtov): ... a new insn reservation.
7399 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7401         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7402         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7403         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7404         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7405         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7407 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7409         * diagnostic-show-locus.c (diagnostic_show_locus): Use
7410         pp_take_prefix when saving the existing prefix.
7411         * diagnostic.c (diagnostic_append_note): Likewise.
7412         * langhooks.c (lhd_print_error_function): Likewise.
7413         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7414         param's type.  Free the existing prefix.
7415         (pp_take_prefix): New function.
7416         (pretty_printer::pretty_printer): Drop the prefix parameter.
7417         Rename the length parameter to match the comment.
7418         (pretty_printer::~pretty_printer): Free the prefix.
7419         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7420         parameter.
7421         (struct pretty_printer): Drop the "const" from "prefix" field's
7422         type and clarify memory management.
7423         (pp_set_prefix): Drop the "const" from the 2nd param.
7424         (pp_take_prefix): New decl.
7426 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7428         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7429         for word_mode_ok here instead of passing as argument.
7430         (expand_block_compare): Change select_block_compare_mode() call.
7431         (expand_strncmp_gpr_sequence): New function.
7432         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7434 2018-08-02  Jeff Law  <law@redhat.com>
7436         PR target/86790
7437         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7438         Define to speculation_safe_value_not_needed.
7440         PR target/86784
7441         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7442         Define to speculation_safe_value_not_needed.
7444 2018-08-02  Tom de Vries  <tdevries@suse.de>
7446         PR target/86660
7447         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7448         function.  Return UI_TARGET unconditionally.
7449         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7450         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7452 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7454         * genemit.c (print_overload_test): Fix typo.
7456 2018-08-02  Richard Biener  <rguenther@suse.de>
7458         PR tree-optimization/86816
7459         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7460         which checks for value availability before querying it.
7461         (gvn_uses_equal): Use it.
7462         (same_succ_hash): Likewise.
7463         (gimple_equal_p): Likewise.
7465 2018-08-02  Nick Clifton  <nickc@redhat.com>
7467         PR target/86813
7468         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7469         Define to speculation_safe_value_not_needed.
7471         PR target/86810
7472         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7473         Define to speculation_safe_value_not_needed.
7475         PR target/86810
7476         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7477         Define to speculation_safe_value_not_needed.
7479         PR target/86803
7480         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7481         Define to speculation_safe_value_not_needed.
7483         PR target/86797
7484         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7485         Define to speculation_safe_value_not_needed.
7487         PR target/86791
7488         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7489         Define to speculation_safe_value_not_needed.
7491         PR target/86789
7492         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7493         Define to speculation_safe_value_not_needed.
7495         PR target/86787
7496         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7497         Define to speculation_safe_value_not_needed.
7499         PR target/86782
7500         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7501         speculation_safe_value_not_needed.
7503         PR target/86781
7504         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7505         to speculation_safe_value_not_needed.
7507 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7509         * doc/md.texi: Expand the documentation of instruction names
7510         to mention port-local uses.  Document '@' in pattern names.
7511         * read-md.h (overloaded_instance, overloaded_name): New structs.
7512         (mapping): Declare.
7513         (md_reader::handle_overloaded_name): New member function.
7514         (md_reader::get_overloads): Likewise.
7515         (md_reader::m_first_overload): New member variable.
7516         (md_reader::m_next_overload_ptr): Likewise.
7517         (md_reader::m_overloads_htab): Likewise.
7518         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
7519         m_next_overload_ptr and m_overloads_htab.
7520         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
7521         (get_mode_token, get_code_token, get_int_token): New functions.
7522         (map_attr_string): Add an optional argument that passes back
7523         the associated iterator.
7524         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
7525         (md_reader::handle_overloaded_name, add_overload_instance): New
7526         functions.
7527         (apply_iterators): Handle '@' names.  Report an error if '@'
7528         is used without iterators.
7529         (initialize_iterators): Initialize the new iterator_group fields.
7530         * genopinit.c (handle_overloaded_code_for)
7531         (handle_overloaded_gen): New functions.
7532         (main): Use them to print declarations of maybe_code_for_* and
7533         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
7534         * genemit.c (print_overload_arguments, print_overload_test)
7535         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
7536         (main): Use it to print definitions of maybe_code_for_* and
7537         maybe_gen_* functions.
7538         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
7539         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
7540         instead of explicit mode checks.
7541         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
7542         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
7543         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
7544         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
7545         (aarch64_expand_compare_and_swap): Likewise
7546         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
7547         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
7548         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
7549         (aarch64_constant_pool_reload_icode): Delete.
7550         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
7551         instead of aarch64_constant_pool_reload_icode.  Use
7552         code_for_aarch64_reload_mov instead of explicit mode checks.
7553         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
7554         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
7555         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
7556         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
7557         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
7558         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
7559         (aarch64_atomic_load_op_code): Delete.
7560         (aarch64_emit_atomic_load_op): Likewise.
7561         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
7562         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
7563         instead of aarch64_emit_atomic_load_op.
7564         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
7565         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
7566         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
7567         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
7568         character before the pattern name.
7569         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
7570         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
7571         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
7572         (aarch64_frecps<mode>): Likewise.
7573         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
7574         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
7575         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
7576         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
7577         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
7579 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7581         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
7582         Allow HFmode constants if TARGET_FP_F16INST.
7584 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
7586         PR target/86014
7587         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
7588         No longer check last store for clobber of address register.
7590 2018-08-02  Martin Liska  <mliska@suse.cz>
7592         PR gcov-profile/86817
7593         * gcov.c (process_all_functions): New function.
7594         (main): Call it.
7595         (process_file): Move functions processing to
7596         process_all_functions.
7598 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7600         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
7601         "const" to the "gimple *" and "rtx_insn *" parameters.
7602         * dumpfile.h (dump_user_location_t::dump_user_location_t):
7603         Likewise.
7604         (dump_location_t::dump_location_t): Likewise.
7606 2018-08-01  Martin Sebor  <msebor@redhat.com>
7608         PR tree-optimization/86650
7609         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
7610         rather than a "gcall *".  Directly pass the data of interest
7611         to percent_K_format, rather than building a temporary CALL_EXPR
7612         to hold it.
7613         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
7614         (gimple_fold_builtin_strncat): Adjust.
7615         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
7616         gcall* argument with gimple*.
7617         * gimple-ssa-warn-restrict.c (check_call): Same.
7618         (wrestrict_dom_walker::before_dom_children): Same.
7619         (builtin_access::builtin_access): Same.
7620         (check_bounds_or_overlap): Same
7621         (maybe_diag_overlap): Same.
7622         (maybe_diag_offset_bounds): Same.
7623         * tree-diagnostic.c (default_tree_printer): Move usage of
7624         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
7625         to this callsite.
7626         * tree-pretty-print.c (percent_K_format): Add argument.
7627         * tree-pretty-print.h: Add argument.
7628         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
7629         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
7630         (maybe_diag_stxncpy_trunc): Same.
7631         (handle_builtin_stxncpy): Same.
7632         (handle_builtin_strcat): Same.
7634 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7636         * match.pd: Optimise pointer range checks.
7638 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7640         PR tree-optimization/86758
7641         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
7642         to remove pattern statements.
7644 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7646         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
7647         result of dfs_enumerate_from when constructing stmt_vec_infos,
7648         instead of additionally calling get_loop_body.
7650 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7652         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
7653         parameter.
7654         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
7655         When creating an iv, assert that the step is not known to be zero.
7656         (vect_setup_realignment): Update call accordingly.
7657         * tree-vect-stmts.c (vectorizable_store): Likewise.
7658         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
7660 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7662         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
7663         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
7664         (vectorizable_reduction): Likewise.
7665         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7666         (vect_detect_hybrid_slp_stmts): Likewise.
7667         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7669 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
7671         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
7672         (wide_int_set_zero_nonzero_bits): ...this.
7673         (zero_nonzero_bits_from_vr): Rename to...
7674         (vrp_set_zero_nonzero_bits): ...this.
7675         (extract_range_from_multiplicative_op_1): Abstract wide int
7676         code...
7677         (wide_int_range_multiplicative_op): ...here.
7678         (extract_range_from_binary_expr_1): Extract wide int binary
7679         operations into their own functions.
7680         (wide_int_range_lshift): New.
7681         (wide_int_range_can_optimize_bit_op): New.
7682         (wide_int_range_shift_undefined_p): New.
7683         (wide_int_range_bit_xor): New.
7684         (wide_int_range_bit_ior): New.
7685         (wide_int_range_bit_and): New.
7686         (wide_int_range_trunc_mod): New.
7687         (extract_range_into_wide_ints): New.
7688         (vrp_shift_undefined_p): New.
7689         (extract_range_from_multiplicative_op): New.
7690         (vrp_can_optimize_bit_op): New.
7691         * tree-vrp.h (value_range::dump): New.
7692         (wide_int_range_multiplicative_op): New.
7693         (wide_int_range_lshift):New.
7694         (wide_int_range_shift_undefined_p): New.
7695         (wide_int_range_bit_xor): New.
7696         (wide_int_range_bit_ior): New.
7697         (wide_int_range_bit_and): New.
7698         (wide_int_range_trunc_mod): New.
7699         (zero_nonzero_bits_from_bounds): Rename to...
7700         (wide_int_set_zero_nonzero_bits): ...this.
7701         (zero_nonzero_bits_from_vr): Rename to...
7702         (vrp_set_zero_nonzero_bits): ...this.
7703         (range_easy_mask_min_max): Rename to...
7704         (wide_int_range_can_optimize_bit_op): this.
7705         * vr-values.c (simplify_bit_ops_using_ranges): Rename
7706         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
7708 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7710         * tree-vectorizer.h (vect_orig_stmt): New function.
7711         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
7712         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
7713         (vect_create_epilog_for_reduction): Likewise.
7714         (vectorizable_live_operation): Likewise.
7715         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7716         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
7717         * tree-vect-stmts.c (vectorizable_call): Likewise.
7718         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
7720 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7722         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
7723         argument.
7724         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
7725         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
7726         (vect_transform_loop): Likewise.
7727         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7729 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7731         * tree-vectorizer.h (vect_schedule_slp): Return void.
7732         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7733         (vect_schedule_slp): Likewise.
7735 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7737         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
7738         argument.
7739         (vect_transform_loop): Update calls accordingly.  Schedule SLP
7740         instances before the main loop, if any exist.
7742 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7744         PR tree-optimization/86749
7745         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
7746         If the lhs is used in a COND_EXPR, check that it is being used
7747         as the "then" or "else" value.
7749 2018-08-01  Tom de Vries  <tdevries@suse.de>
7751         PR target/86800
7752         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7753         speculation_safe_value_not_needed.
7755 2018-08-01  Richard Biener  <rguenther@suse.de>
7757         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
7758         as base and offset.
7760 2018-08-01  Martin Liska  <mliska@suse.cz>
7762         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
7763         format how successful transformation is dumped.
7764         (gimple_mod_pow2_value_transform): Likewise.
7765         (gimple_mod_subtract_transform): Likewise.
7766         (gimple_stringops_transform): Likewise.
7768 2018-08-01  Martin Liska  <mliska@suse.cz>
7770         PR value-prof/35543
7771         * value-prof.c (interesting_stringop_to_profile_p):
7772         Simplify the code and add BUILT_IN_MEMMOVE.
7773         (gimple_stringops_transform): Likewise.
7775 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
7777         * config/aarch64/aarch64-simd.md
7778         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
7779         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
7780         use GPI iterator instead of SI mode.
7782 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7784         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
7785         rs6000_speculation_barrier.
7786         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
7787         new barrier pattern name.
7789 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7791         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7792         (speculation_barrier): New insn.
7794 2018-08-01  Richard Biener  <rguenther@suse.de>
7796         PR bootstrap/86724
7797         * graphite.h: Include isl/id.h and isl/space.h to allow build
7798         with ISL 0.20.
7800 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
7802         PR target/86651
7803         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7804         mode for COFF targets.
7805         * defaults.h (TARGET_COFF): Define.
7806         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7807         TARGET_COFF): Define.
7808         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7809         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7810         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7812 2018-07-31  Alexandre Oliva <oliva@adacore.com>
7814         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7815         * gimple-streamer-out.c (output_bb): Save it.
7816         * lto-streamer-in.c (input_struct_function_base): Restore
7817         instance discriminator if available.  Create map on demand.
7818         * lto-streamer-out.c (output_struct_function_base): Save it if
7819         available.
7820         * final.c (decl_to_instance_map): Document LTO strategy.
7822 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
7823             Olivier Hainque  <hainque@adacore.com>
7825         * debug.h (decl_to_instance_map_t): New type.
7826         (decl_to_instance_map): Declare.
7827         (maybe_create_decl_to_instance_map): New inline function.
7828         * final.c (bb_discriminator, last_bb_discriminator): New statics,
7829         to track basic block discriminators.
7830         (final_start_function_1): Initialize them.
7831         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7832         bb_discriminator.
7833         (decl_to_instance_map): New variable.
7834         (map_decl_to_instance, maybe_set_discriminator): New functions.
7835         (notice_source_line): Set discriminator.
7837 2018-07-31  Ian Lance Taylor  <iant@golang.org>
7839         * targhooks.c (default_have_speculation_safe_value): Add
7840         ATTRIBUTE_UNUSED.
7842 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7844         * dump-context.h: Include "pretty-print.h".
7845         (dump_context::refresh_dumps_are_enabled): New decl.
7846         (dump_context::emit_item): New decl.
7847         (class dump_context): Add fields "m_test_pp" and
7848         "m_test_pp_flags".
7849         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7850         (temp_dump_context::get_dumped_text): New decl.
7851         (class temp_dump_context): Add field "m_pp".
7852         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7853         (dump_context::refresh_dumps_are_enabled): ...and add a test for
7854         m_test_pp.
7855         (set_dump_file): Update for above change.
7856         (set_alt_dump_file): Likewise.
7857         (dump_loc): New overload, taking a pretty_printer *.
7858         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
7859         to any test pretty-printer.
7860         (make_item_for_dump_gimple_stmt): New function, adapted from
7861         optinfo::add_gimple_stmt.
7862         (dump_context::dump_gimple_stmt): Call it, and use the result,
7863         eliminating the direct usage of dump_file and alt_dump_file in
7864         favor of indirectly using them via emit_item.
7865         (make_item_for_dump_gimple_expr): New function, adapted from
7866         optinfo::add_gimple_expr.
7867         (dump_context::dump_gimple_expr): Call it, and use the result,
7868         eliminating the direct usage of dump_file and alt_dump_file in
7869         favor of indirectly using them via emit_item.
7870         (make_item_for_dump_generic_expr): New function, adapted from
7871         optinfo::add_tree.
7872         (dump_context::dump_generic_expr): Call it, and use the result,
7873         eliminating the direct usage of dump_file and alt_dump_file in
7874         favor of indirectly using them via emit_item.
7875         (make_item_for_dump_printf_va): New function, adapted from
7876         optinfo::add_printf_va.
7877         (make_item_for_dump_printf): New function.
7878         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7879         and use the result, eliminating the direct usage of dump_file and
7880         alt_dump_file in favor of indirectly using them via emit_item.
7881         (make_item_for_dump_dec): New function.
7882         (dump_context::dump_dec): Call it, and use the result,
7883         eliminating the direct usage of dump_file and alt_dump_file in
7884         favor of indirectly using them via emit_item.
7885         (make_item_for_dump_symtab_node): New function, adapted from
7886         optinfo::add_symtab_node.
7887         (dump_context::dump_symtab_node): Call it, and use the result,
7888         eliminating the direct usage of dump_file and alt_dump_file in
7889         favor of indirectly using them via emit_item.
7890         (dump_context::begin_scope): Reimplement, avoiding direct usage
7891         of dump_file and alt_dump_file in favor of indirectly using them
7892         via emit_item.
7893         (dump_context::emit_item): New member function.
7894         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7895         Set up test pretty-printer on the underlying context.  Call
7896         refresh_dumps_are_enabled.
7897         (temp_dump_context::~temp_dump_context): Call
7898         refresh_dumps_are_enabled.
7899         (temp_dump_context::get_dumped_text): New member function.
7900         (selftest::verify_dumped_text): New function.
7901         (ASSERT_DUMPED_TEXT_EQ): New macro.
7902         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7903         and then without optinfo enabled.  Add uses of
7904         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7905         * dumpfile.h: Update comment for the dump_* API.
7906         * optinfo-emit-json.cc
7907         (selftest::test_building_json_from_dump_calls): Update for new
7908         param for temp_dump_context ctor.
7909         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7910         and "m_owned" field.
7911         (optinfo_item::~optinfo_item): Likewise.
7912         (optinfo::add_item): New member function.
7913         (optinfo::emit): Update comment.
7914         (optinfo::add_string): Delete.
7915         (optinfo::add_printf): Delete.
7916         (optinfo::add_printf_va): Delete.
7917         (optinfo::add_gimple_stmt): Delete.
7918         (optinfo::add_gimple_expr): Delete.
7919         (optinfo::add_tree): Delete.
7920         (optinfo::add_symtab_node): Delete.
7921         (optinfo::add_dec): Delete.
7922         * optinfo.h (class dump_context): New forward decl.
7923         (optinfo::add_item): New decl.
7924         (optinfo::add_string): Delete.
7925         (optinfo::add_printf): Delete.
7926         (optinfo::add_printf_va): Delete.
7927         (optinfo::add_gimple_stmt): Delete.
7928         (optinfo::add_gimple_expr): Delete.
7929         (optinfo::add_tree): Delete.
7930         (optinfo::add_symtab_node): Delete.
7931         (optinfo::add_dec): Delete.
7932         (optinfo::add_poly_int): Delete.
7933         (optinfo_item::optinfo_item): Remove "owned" param.
7934         (class optinfo_item): Remove field "m_owned".
7936 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7938         PR middle-end/86705
7939         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7940         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7941         requested variable alignment.
7942         (expand_one_ssa_partition): Likewise.
7943         (expand_one_var): Likewise.
7945 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7947         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7948         to speculation_safe_value_not_needed.
7950 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7952         * targhooks.h (speculation_safe_value_not_needed): New prototype.
7953         * targhooks.c (speculation_safe_value_not_needed): New function.
7954         * target.def (have_speculation_safe_value): Update documentation.
7955         * doc/tm.texi: Regenerated.
7957 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7959         * config/aarch64/iterators.md (ALLI_TI): New iterator.
7960         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7961         expand.
7962         (despeculate_copy<ALLI:mode>_insn): New insn.
7963         (despeculate_copyti_insn): New insn.
7964         (despeculate_simple<ALLI:mode>): New insn
7965         (despeculate_simpleti): New insn.
7966         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7967         function.
7968         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7969         aarch64_speculation_safe_value.
7970         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7972 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7974         * config/aarch64/aarch64-speculation.cc: New file.
7975         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7976         before pass_reorder_blocks.
7977         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7978         prototype.
7979         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7980         X14 and X15 when tracking speculation.
7981         * config/aarch64/aarch64.md (register name constants): Add
7982         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7983         (unspec): Add UNSPEC_SPECULATION_TRACKER.
7984         (speculation_barrier): New insn attribute.
7985         (cmp<mode>): Allow SP in comparisons.
7986         (speculation_tracker): New insn.
7987         (speculation_barrier): Add speculation_barrier attribute.
7988         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7989         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7990         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7992 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7994         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7995         aarch64_track_speculation is true.
7996         (tb<optab><mode>1): Likewise.
7997         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7998         generate CB[N]Z when tracking speculation.
7999         (aarch64_split_compare_and_swap): Likewise.
8000         (aarch64_split_atomic_op): Likewise.
8002 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8004         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8006 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8008         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8009         (speculation_barrier): New insn.
8011 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8013         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8014         * config/arm/arm.md (speculation_barrier): New expand.
8015         (speculation_barrier_insn): New pattern.
8017 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8019         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8020         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8021         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8022         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8023         list.
8024         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8025         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8026         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8027         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8028         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8029         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8030         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8031         * builtins.c (expand_speculation_safe_value): New function.
8032         (expand_builtin): Call it.
8033         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8034         * doc/extend.texi: Document __builtin_speculation_safe_value.
8035         * doc/md.texi: Document "speculation_barrier" pattern.
8036         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8037         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8038         * doc/tm.texi: Regenerated.
8039         * target.def (have_speculation_safe_value, speculation_safe_value): New
8040         hooks.
8041         * targhooks.c (default_have_speculation_safe_value): New function.
8042         (default_speculation_safe_value): New function.
8043         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8044         (default_speculation_safe_value): Add prototype.
8046 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8048         * dump-context.h (dump_context::dump_loc): New decl.
8049         * dumpfile.c (dump_context::dump_loc): New member function.
8050         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8051         and dump_gimple_stmt.
8052         (dump_context::dump_gimple_expr_loc): Likewise, using
8053         dump_gimple_expr.
8054         (dump_context::dump_generic_expr_loc): Likewise, using
8055         dump_generic_expr.
8056         (dump_context::dump_printf_loc_va): Likewise, using
8057         dump_printf_va.
8058         (dump_context::begin_scope): Explicitly using the global function
8059         "dump_loc", rather than the member function.
8061 2018-07-31  Martin Sebor  <msebor@redhat.com>
8063         PR tree-optimization/86741
8064         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8066 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8068         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8069         depend on whether prefetch instructions will be emitted or not.
8070         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8071         will be emitted or not.
8072         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8073         (TARGET_SETMEM_PFD): New macros.
8075 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8077         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8078         (NULL_STMT_VEC_INFO): Delete.
8079         (stmt_vec_info::operator*): Likewise.
8080         (stmt_vec_info::operator gimple *): Likewise.
8081         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8082         of NULL_STMT_VEC_INFO.
8083         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8084         (vect_reassociating_reduction_p): Likewise.
8085         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8086         (vectorizable_store): Likewise.
8087         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8088         (vec_info::free_stmt_vec_infos): Likewise.
8090 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8092         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8093         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8095 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8097         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8098         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8099         (vec_info::free_stmt_vec_info): New private member functions.
8100         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8101         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8102         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8103         set_stmt_vec_info_vec and free_stmt_vec_infos.
8104         * tree-vect-loop.c (_loop_vec_info): Remove call to
8105         set_stmt_vec_info_vec.
8106         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8107         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8108         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8109         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8110         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8111         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8112         with the clearing in the xcalloc.
8113         (stmt_vec_info_vec): Delete.
8114         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8115         (vectorize_loops): Likewise.
8116         (vec_info::~vec_info): Remove argument from call to
8117         free_stmt_vec_infos.
8118         (vec_info::add_stmt): Remove vinfo argument from call to
8119         new_stmt_vec_info.
8121 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8123         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8124         rather than a gimple stmt.
8125         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8126         information for pattern statements when passed the original
8127         statement; instead wait to be passed the pattern statement itself.
8128         Don't call set_vinfo_for_stmt here.
8129         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8130         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8131         stmt_vec_infos here.
8132         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8133         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8134         stmt_vec_infos entry.
8136 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8138         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8139         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8140         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8141         * tree-vect-stmts.c (vectorizable_call): Likewise.
8142         (vectorizable_simd_clone_call): Likewise.
8144 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8146         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8147         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8148         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8149         * tree-vect-loop.c (vect_transform_loop): Likewise.
8150         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8151         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8153 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8155         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8156         (vect_dr_stmt): Delete.
8157         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8158         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8159         of DR_VECT_AUX.
8160         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8161         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8162         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8163         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8164         (vect_analyze_data_refs): Likewise.
8165         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8166         argument.
8167         (vect_find_same_alignment_drs): Likewise.
8168         (vect_slp_analyze_node_dependences): Update calls accordingly.
8169         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8170         instead of DR_VECT_AUX.
8171         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8172         of a vector data references.  Use vec_info::lookup_dr instead of
8173         DR_VECT_AUX.
8174         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8175         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8176         instead of DR_VECT_AUX.
8178 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8180         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8181         dr_vec_info.
8182         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8183         accordingly.
8184         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8185         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8186         (vect_gen_prolog_loop_niters): Likewise.
8188 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8190         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8191         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8192         (vect_known_alignment_in_bytes, vect_dr_behavior)
8193         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8194         instead of data_references.  Update calls to other routines for
8195         which the same change has been made.
8196         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8197         dr_vec_infos instead of stmt_vec_infos.
8198         (vect_analyze_data_ref_dependence): Update call accordingly.
8199         (vect_slp_analyze_data_ref_dependence)
8200         (vect_record_base_alignments): Use DR_VECT_AUX.
8201         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8202         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8203         (vector_alignment_reachable_p, vect_get_data_access_cost)
8204         (vect_peeling_supportable, vect_analyze_group_access_1)
8205         (vect_analyze_group_access, vect_analyze_data_ref_access)
8206         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8207         (vect_compile_time_alias, vect_small_gap_p)
8208         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8209         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8210         instead of data_references.  Update calls to other routines for
8211         which the same change has been made.
8212         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8213         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8214         (vect_slp_analyze_and_verify_node_alignment)
8215         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8216         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8217         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8218         above changes.
8219         (_vect_peel_info::dr): Replace with...
8220         (_vect_peel_info::dr_info): ...this new field.
8221         (vect_peeling_hash_get_most_frequent)
8222         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8223         (vect_peeling_hash_get_lowest_cost):
8224         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8225         routines for which the same change has been made.
8226         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8227         data_reference.
8228         * tree-vect-loop-manip.c (get_misalign_in_elems)
8229         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8230         above changes.
8231         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8232         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8233         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8234         (get_group_load_store_type, get_negative_load_store_type)
8235         (vect_get_data_ptr_increment, vectorizable_store)
8236         (vectorizable_load): Likewise.
8237         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8238         Update calls to other routines for which the same change has been made.
8240 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8242         * tree-vectorizer.h (vec_info::move_dr): New member function.
8243         (dataref_aux): Rename to...
8244         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8245         (_stmt_vec_info::dr_aux): Update accordingly.
8246         (_stmt_vec_info::data_ref_info): Delete.
8247         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8248         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8249         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8250         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8251         of data_ref.
8252         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8253         (STMT_VINFO_DR_INFO): New macro.
8254         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8255         (set_dr_misalignment): Update after rename of dataref_aux.
8256         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8257         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8258         initialization of STMT_VINFO_DATA_REF.
8259         * tree-vectorizer.c (vec_info::move_dr): New function.
8260         * tree-vect-patterns.c (vect_recog_bool_pattern)
8261         (vect_recog_mask_conversion_pattern)
8262         (vect_recog_gather_scatter_pattern): Use it.
8263         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8264         the "dr" and "stmt" fields of dr_vec_info instead of
8265         STMT_VINFO_DATA_REF.
8267 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8269         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8270         (is_pattern_stmt_p): Use it.
8271         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8272         on pattern statements.
8274 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8276         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8277         original stmt as a stmt_vec_info rather than a gimple stmt.
8278         (vect_pattern_recog_1): Take the statement directly as a
8279         stmt_vec_info, rather than via a gimple_stmt_iterator.
8280         Update call to vect_mark_pattern_stmts.
8281         (vect_pattern_recog): Update calls accordingly.
8283 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8285         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8286         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8287         a vect_def_type for the first argument.
8288         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8289         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8290         operand if it isn't defined by a vectorized statement.
8291         (vect_build_gather_load_calls): Remove the mask_dt argument and
8292         update calls to vect_get_vec_def_for_stmt_copy.
8293         (vectorizable_bswap): Likewise the dt argument.
8294         (vectorizable_call): Update calls to vectorizable_bswap and
8295         vect_get_vec_def_for_stmt_copy.
8296         (vectorizable_simd_clone_call, vectorizable_assignment)
8297         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8298         (vectorizable_comparison): Update calls to
8299         vect_get_vec_def_for_stmt_copy.
8300         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8301         vect_is_simple_use.
8302         (vect_get_loop_based_defs): Remove dt argument and update call
8303         to vect_get_vec_def_for_stmt_copy.
8304         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8305         and vect_get_vec_def_for_stmt_copy.
8306         (vectorizable_load): Update calls to vect_build_gather_load_calls
8307         and vect_get_vec_def_for_stmt_copy.
8308         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8309         (vectorizable_reduction, vectorizable_live_operation): Update calls
8310         to vect_get_vec_def_for_stmt_copy.
8312 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8314         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8315         and gimple stmt arguments with a stmt_vec_info.
8316         (vect_record_base_alignments): Update calls accordingly.
8317         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8318         and gimple stmt arguments with a stmt_vec_info.
8319         (vect_build_slp_tree_1): Remove vinfo argument and update call
8320         to vect_record_max_nunits.
8321         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8322         and vect_record_max_nunits.
8324 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8326         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8327         file and take a stmt_vec_info instead of a gimple stmt.
8328         (supportable_widening_operation, vect_finish_replace_stmt)
8329         (vect_finish_stmt_generation, vect_get_store_rhs)
8330         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8331         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8332         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8333         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8334         (vect_create_data_ref_ptr, bump_vector_ptr)
8335         (vect_permute_store_chain, vect_setup_realignment)
8336         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8337         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8338         (vectorizable_reduction, vectorizable_induction)
8339         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8340         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8341         than gimple stmts as arguments.
8342         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8343         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8344         (can_group_stmts_p, vect_check_gather_scatter)
8345         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8346         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8347         (vect_permute_load_chain, vect_shift_permute_load_chain)
8348         (vect_transform_grouped_load)
8349         (vect_record_grouped_load_vectors): Likewise.
8350         * tree-vect-loop.c (vect_fixup_reduc_chain)
8351         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8352         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8353         (vectorizable_reduction, vectorizable_induction)
8354         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8355         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8356         (vect_get_load_store_mask): Likewise.
8357         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8358         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8359         * tree-vect-stmts.c (vect_mark_relevant)
8360         (is_simple_and_all_uses_invariant)
8361         (exist_non_indexing_operands_for_use_p, process_use)
8362         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8363         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8364         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8365         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8366         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8367         (get_negative_load_store_type, get_load_store_type)
8368         (vect_check_load_store_mask, vect_check_store_rhs)
8369         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8370         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8371         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8372         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8373         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8374         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8375         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8376         (supportable_widening_operation): Likewise.
8378 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8380         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8381         a stmt_vec_info instead of a gcall.
8382         (vect_check_gather_scatter): Update call accordingly.
8383         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8384         of a gphi.
8385         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8386         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8387         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8388         instead of a gimple stmt.
8389         (vect_transform_loop): Update calls accordingly.
8390         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8391         stmt_vec_infos instead of gimple stmts.
8392         (vect_analyze_slp_instance): Update use accordingly.
8393         * tree-vect-stmts.c (read_vector_array, write_vector_array)
8394         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8395         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8396         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8397         (vect_gen_widened_results_half, vect_get_loop_based_defs)
8398         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8399         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8400         down to subroutines.
8402 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8404         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8405         of the worklist from a vector of gimple stmts to a vector of
8406         stmt_vec_infos.
8407         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8408         (vect_mark_stmts_to_be_vectorized): Likewise
8410 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8412         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8413         statement before passing it to vect_analyze_stmt.
8414         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8415         the chain of phi vector definitions.  Track the exit phi via its
8416         stmt_vec_info.
8417         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8418         STMT_VINFO_REDUC_DEF.
8419         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8420         stmt_vec_infos to handle the statement chains.
8421         (vect_get_slp_defs): Record the first statement in the node
8422         using a stmt_vec_info.
8423         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8424         statements here and pass their stmt_vec_info down to subroutines.
8425         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8426         down to vect_finish_stmt_generation.
8427         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8428         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8429         stmt_vec_infos to subroutines.
8430         (vect_remove_stores): Use stmt_vec_infos to handle the statement
8431         chains.
8433 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8435         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8436         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8437         (vect_permute_store_chain, vect_setup_realignment)
8438         (vect_permute_load_chain, vect_shift_permute_load_chain)
8439         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8440         stmts internally, and when passing values to other vectorizer routines.
8441         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8442         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8443         (vect_analyze_loop_operations, get_initial_def_for_reduction)
8444         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8445         (vectorizable_reduction, vectorizable_induction)
8446         (vectorizable_live_operation, vect_transform_loop_stmt)
8447         (vect_transform_loop): Likewise.
8448         * tree-vect-patterns.c (vect_reassociating_reduction_p)
8449         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8450         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8451         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8452         (vect_slp_analyze_node_operations_1): Likewise.
8453         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8454         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8455         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8456         (vect_finish_stmt_generation_1, get_group_load_store_type)
8457         (get_load_store_type, vect_build_gather_load_calls)
8458         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8459         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8460         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8461         (vectorizable_store, vectorizable_load, vectorizable_condition)
8462         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8463         (supportable_widening_operation): Likewise.
8464         (vect_get_vector_types_for_stmt): Likewise.
8465         * tree-vectorizer.h (vect_dr_behavior): Likewise.
8467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8469         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8470         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8471         (vect_permute_store_chain, vect_permute_load_chain)
8472         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8473         repeated stmt_vec_info lookups.
8474         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8475         (vect_update_ivs_after_vectorizer): Likewise.
8476         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8477         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8478         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8479         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8480         (vect_bb_slp_scalar_cost): Likewise.
8481         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8483 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8485         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8486         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8487         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8488         to get gassigns and gcalls, rather than operating on generc gimple
8489         stmts.
8490         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8491         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8492         (vectorizable_load, vect_analyze_stmt): Likewise.
8493         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8495 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8497         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8498         return stmt_vec_infos rather than gimple stmts.  Do not accept
8499         null arguments.
8500         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8501         of a gimple stmt.
8502         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8503         Update use of get_later_stmt.
8504         (vect_get_constant_vectors): Update call accordingly.
8505         (vect_schedule_slp_instance): Likewise
8506         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
8507         (vect_slp_analyze_instance_dependence): Likewise.
8508         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
8510 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8512         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
8513         (stmt_info_for_cost::stmt_info): ...this new field.
8514         (add_stmt_costs): Update accordingly.
8515         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8516         (vect_get_known_peeling_cost): Likewise.
8517         (vect_estimate_min_profitable_iters): Likewise.
8518         * tree-vect-stmts.c (record_stmt_cost): Likewise.
8520 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8522         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
8523         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8524         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8525         accordingly.
8526         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
8528 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8530         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
8531         a gimple stmt to a stmt_vec_info.
8532         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8534 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8536         * tree-vectorizer.h (vec_info::grouped_stores): Change from
8537         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8538         (_loop_vec_info::reduction_chains): Likewise.
8539         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
8540         accordingly.
8541         * tree-vect-slp.c (vect_analyze_slp): Likewise.
8543 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8545         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
8546         a gimple stmt to a stmt_vec_info.
8547         (_stmt_vec_info::next_element): Likewise.
8548         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
8549         (vect_slp_analyze_and_verify_node_alignment)
8550         (vect_analyze_group_access_1, vect_analyze_group_access)
8551         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
8552         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
8553         (vect_supportable_dr_alignment): Update accordingly.
8554         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
8555         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
8556         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
8557         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8558         * tree-vect-slp.c (vect_build_slp_tree_1)
8559         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
8560         (vect_split_slp_store_group, vect_analyze_slp_instance)
8561         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
8562         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
8563         (get_group_load_store_type, get_load_store_type)
8564         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
8565         (vect_transform_stmt, vect_remove_stores): Likewise.
8567 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8569         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
8570         than a gimple stmt.
8571         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8572         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
8573         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8574         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
8575         (vector_alignment_reachable_p, vect_get_data_access_cost)
8576         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
8577         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
8578         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8579         (vect_analyze_group_access_1, vect_analyze_group_access)
8580         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
8581         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
8582         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
8583         result of vect_dr_stmt and use the stmt_vec_info instead of
8584         the associated gimple stmt.
8585         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8586         (vect_gen_prolog_loop_niters): Likewise.
8587         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8589 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8591         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
8592         to a vec<stmt_vec_info>.
8593         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
8594         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
8595         vec<stmt_vec_info>.
8596         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
8597         to a vec<stmt_vec_info>.
8598         (bst_traits::value_type, bst_traits::value_type): Likewise.
8599         (bst_traits::hash): Update accordingly.
8600         (vect_get_and_check_slp_defs): Change the stmts parameter from
8601         a vec<gimple *> to a vec<stmt_vec_info>.
8602         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
8603         (vect_build_slp_tree): Likewise.
8604         (vect_build_slp_tree_2): Likewise.  Update uses of
8605         SLP_TREE_SCALAR_STMTS.
8606         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
8607         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
8608         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
8609         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
8610         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
8611         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
8612         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
8613         (vect_get_constant_vectors, vect_get_slp_defs)
8614         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
8615         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
8616         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
8617         instead of gimple stmts.
8618         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
8619         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
8620         (vect_slp_analyze_instance_dependence): Update uses of
8621         SLP_TREE_SCALAR_STMTS.
8622         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8623         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
8624         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
8625         (get_initial_defs_for_reduction): Likewise.
8626         (vect_create_epilog_for_reduction): Likewise.
8627         (vectorize_fold_left_reduction): Likewise.
8628         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
8629         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
8630         (can_vectorize_live_stmts): Likewise.
8632 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8634         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
8635         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8636         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
8637         than gimple stmts.
8638         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
8639         of a gimple stmt.
8640         (gather_scalar_reductions): Update after above interface changes.
8641         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
8642         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
8643         than gimple stmts.
8644         (vect_force_simple_reduction): Likewise.
8645         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
8646         LOOP_VINFO_REDUCTIONS.
8647         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8649 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8651         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
8652         a gimple stmt to a stmt_vec_info.
8653         * tree-vect-loop.c (vect_active_double_reduction_p)
8654         (vect_force_simple_reduction, vectorizable_reduction): Update
8655         accordingly.
8657 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8659         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
8660         vec<gimple *> to a vec<stmt_vec_info>.
8661         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
8662         the reduction_phis argument from a vec<gimple *> to a
8663         vec<stmt_vec_info>.
8664         (vectorizable_reduction): Likewise the phis local variable that
8665         is passed to vect_create_epilog_for_reduction.  Update for new type
8666         of SLP_TREE_VEC_STMTS.
8667         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
8668         (vectorizable_live_operation): Likewise.
8669         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
8670         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
8672 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8674         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
8675         a gimple stmt to a stmt_vec_info.
8676         (vectorizable_condition, vectorizable_live_operation)
8677         (vectorizable_reduction, vectorizable_induction): Pass back the
8678         vectorized statement as a stmt_vec_info.
8679         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
8680         use of STMT_VINFO_VEC_STMT.
8681         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
8682         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
8683         as stmt_vec_infos rather than gimple stmts.
8684         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
8685         to a stmt_vec_info.
8686         (vectorizable_live_operation): Likewise.
8687         (vectorizable_reduction, vectorizable_induction): Likewise,
8688         updating use of STMT_VINFO_VEC_STMT.
8689         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
8690         of STMT_VINFO_VEC_STMT.
8691         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
8692         (vectorizable_simd_clone_call, vectorizable_conversion)
8693         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8694         (vectorizable_store, vectorizable_load, vectorizable_condition)
8695         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
8696         from a gimple stmt to a stmt_vec_info.
8697         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
8698         pointer to a stmt_vec_info to the vectorizable_* routines.
8700 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8702         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
8703         a gimple stmt to a stmt_vec_info.
8704         (is_pattern_stmt_p): Update accordingly.
8705         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
8706         (vect_record_grouped_load_vectors): Likewise.
8707         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8708         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
8709         (vect_model_reduction_cost): Likewise.
8710         (vect_create_epilog_for_reduction): Likewise.
8711         (vectorizable_reduction, vectorizable_induction): Likewise.
8712         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8713         Return the stmt_vec_info for the pattern statement.
8714         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
8715         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
8716         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8717         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
8718         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8719         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
8720         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
8721         (free_stmt_vec_info, vect_is_simple_use): Likewise.
8723 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8725         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
8726         (vect_finish_stmt_generation): Likewise.
8727         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8728         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
8729         (vect_build_gather_load_calls): Use the return value of the above
8730         functions instead of a separate call to vinfo_for_stmt.  Use narrow
8731         scopes for the input gimple stmt and wider scopes for the associated
8732         stmt_vec_info.  Use vec_info::lookup_def when setting these
8733         stmt_vec_infos from an SSA_NAME definition.
8734         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8735         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8736         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8737         (vectorizable_store, vectorizable_load, vectorizable_condition)
8738         (vectorizable_comparison): Likewise.
8739         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
8740         (vectorizable_reduction): Likewise.
8742 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8744         * tree-vectorizer.h (vect_is_simple_use): Add an optional
8745         stmt_vec_info * parameter before the optional gimple **.
8746         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8747         (process_use, vect_get_vec_def_for_operand_1): Update callers.
8748         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
8749         * tree-vect-loop.c (vectorizable_reduction): Likewise.
8750         (vectorizable_live_operation): Likewise.
8751         * tree-vect-patterns.c (type_conversion_p): Likewise.
8752         (vect_look_through_possible_promotion): Likewise.
8753         (vect_recog_rotate_pattern): Likewise.
8754         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8756 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8758         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
8759         a typedef to a wrapper class.
8760         (NULL_STMT_VEC_INFO): New macro.
8761         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
8762         (stmt_vec_info::operator*): New function.
8763         (stmt_vec_info::operator gimple *): Likewise.
8764         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
8765         (add_stmt_costs): Likewise.
8766         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
8767         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8768         (vect_get_known_peeling_cost): Likewise.
8769         (vect_estimate_min_profitable_iters): Likewise.
8770         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8771         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
8772         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8773         (vectorizable_store, free_stmt_vec_infos): Likewise.
8774         (new_stmt_vec_info): Change return type of xcalloc to
8775         _stmt_vec_info *.
8777 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8779         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
8780         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
8781         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
8782         a single_imm_use-based sequence.
8783         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
8785 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8787         * tree-vectorizer.h (vec_info::lookup_def): Declare.
8788         * tree-vectorizer.c (vec_info::lookup_def): New function.
8789         * tree-vect-patterns.c (vect_get_internal_def): Use it.
8790         (vect_widened_op_tree): Likewise.
8791         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8792         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8793         (vectorizable_reduction): Likewise.
8794         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8795         of a gimple *.
8796         (vect_is_slp_reduction): Update calls accordingly.  Use
8797         vec_info::lookup_def.
8798         (vect_is_simple_reduction): Likewise
8799         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8801 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8803         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8804         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8805         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8806         of vinfo_for_stmt.
8807         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8808         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8809         (vect_update_vf_for_slp, vect_analyze_loop_operations)
8810         (vect_is_slp_reduction, vectorizable_induction)
8811         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8812         * tree-vect-patterns.c (vect_init_pattern_stmt):
8813         (vect_determine_min_output_precision_1, vect_determine_precisions)
8814         (vect_pattern_recog): Likewise.
8815         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8816         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8817         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8818         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8819         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8820         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
8821         info field from a loop to a loop_vec_info.
8823 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8825         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8826         (vec_info::add_stmt): Declare.
8827         * tree-vectorizer.c (vec_info::add_stmt): New function.
8828         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8829         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8830         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8831         (vectorizable_induction): Likewise.
8832         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8833         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8834         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8835         (vectorizable_load): Likewise.
8836         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8837         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8838         (vect_recog_gather_scatter_pattern): Likewise.
8839         (append_pattern_def_seq): Likewise.  Remove a check that is
8840         performed by add_stmt itself.
8842 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8844         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8845         which make_ssa_name was called with new_stmt before new_stmt
8846         had been created.
8848 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8850         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8851         split out from...
8852         (vect_is_slp_reduction): ...here...
8853         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
8854         that are already known to be false.
8856 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8858         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8859         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
8860         STMT_VINFO_NUM_SLP_USES when it's true.
8861         (vect_free_slp_instance): Add a final_p parameter and pass it to
8862         vect_free_slp_tree.
8863         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8864         (vect_analyze_slp_instance): Likewise.
8865         (vect_slp_analyze_operations): Likewise.
8866         (vect_slp_analyze_bb_1): Likewise.
8867         * tree-vectorizer.c (vec_info): Likewise.
8868         * tree-vect-loop.c (vect_transform_loop): Likewise.
8870 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8872         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8873         function is not called for second and subsequent members of
8874         a reduction group.
8876 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8878         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8879         cases for nested loops from here to ...
8880         (vect_create_epilog_for_reduction): ...here.  Only call
8881         vect_is_simple_use for inner-loop reductions.
8883 2018-07-31  Martin Liska  <mliska@suse.cz>
8885         PR gcov-profile/85338
8886         PR gcov-profile/85350
8887         PR gcov-profile/85372
8888         * profile.c (struct location_triplet): New.
8889         (struct location_triplet_hash): Likewise.
8890         (output_location): Do not output a BB that
8891         is already recorded for a line.
8892         (branch_prob): Use streamed_locations.
8894 2018-07-31  Martin Liska  <mliska@suse.cz>
8896         PR gcov-profile/85370
8897         * coverage.c (coverage_begin_function): Do not mark target
8898         clones as artificial functions.
8900 2018-07-31  Martin Liska  <mliska@suse.cz>
8902         PR gcov-profile/83813
8903         PR gcov-profile/84758
8904         PR gcov-profile/85217
8905         PR gcov-profile/85332
8906         * profile.c (branch_prob): Do not record GOTO expressions
8907         for GIMPLE statements which locations are already streamed.
8909 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8911         * gcc.c (handle_spec_function): Accept a soft_matched_part
8912         argument, as do_spec_1.  Pass it down to ...
8913         (eval_spec_function): Accept a soft_matched_part argument,
8914         and pass it down to ...
8915         (do_spec_2): Accept a soft_matched_part argument, and pass
8916         it down to do_spec_1.
8917         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8918         (handle_braces): Update call to handle_spec_function.
8919         (driver::set_up_specs): Update calls to do_spec_2.
8920         (compare_debug_dump_opt_spec_function): Likewise.
8921         (compare_debug_self_opt_spec_function): Likewise.
8923 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8925         * common.opt (nolibc): New option.
8926         * doc/invoke.texi (Link Options): Document it.
8927         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8928         * config/alpha/linux.h: Likewise.
8929         * config/arc/elf.h: Likewise.
8930         * config/arm/uclinux-elf.h: Likewise.
8931         * config/arm/unknown-elf.h: Likewise.
8932         * config/avr/avrlibc.h: Likewise.
8933         * config/bfin/bfin.h: Likewise.
8934         * config/bfin/linux.h: Likewise.
8935         * config/bfin/uclinux.h: Likewise.
8936         * config/darwin.h: Likewise.
8937         * config/darwin10.h: Likewise.
8938         * config/darwin12.h: Likewise.
8939         * config/gnu-user.h: Likewise.
8940         * config/lm32/uclinux-elf.h: Likewise.
8941         * config/pa/pa-hpux11.h: Likewise.
8942         * config/pa/pa64-hpux.h: Likewise.
8943         * config/sparc/sparc.h: Likewise.
8945 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8947         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8948         undefined variables.
8950 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8952         PR target/86640
8953         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8954         instead of GEN_INT.
8956 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8958         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8959         terminated string literal.
8961 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8963         PR rtl-optimization/85160
8964         * combine.c (is_just_move): New function.
8965         (try_combine): Allow combining two instructions into two if neither of
8966         the original instructions was a move.
8968 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
8970         PR target/86673
8971         * doc/extend.texi (Global Register Variables): Discourage use of type
8972         qualifiers.
8973         (Local Register Variables): Likewise.
8975 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
8977         PR tree-optimization/86506
8978         * hwint.h (ceil_log2): Resync with hwint.c implementation.
8980 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
8982         PR target/86547
8983         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8984         hard_regno, make sure no insn between `from` and `to` clobbers it.
8986 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
8987             Tom de Vries  <tdevries@suse.de>
8989         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8990         (PTX_DEFAULT_RUNTIME_DIM): ... this.
8991         (nvptx_goacc_validate_dims): Set default worker and gang dims to
8992         PTX_DEFAULT_RUNTIME_DIM.
8993         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8995 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
8997         * config/pa/pa.c (pa_output_addr_vec): Align address table.
8998         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8999         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9001 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9003         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9004         used.
9005         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9006         predicate to reflect toc fusion has been deleted.
9007         (toc_fusion_mem_raw): Delete, no longer used.
9008         (toc_fusion_mem_wrapped): Likewise.
9009         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9010         fusion mask bit.
9011         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9012         Delete, no longer used.
9013         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9014         meant to be used for toc fusion.
9015         (rs6000_debug_print_mode): Delete toc fusion debugging.
9016         (rs6000_debug_reg_global): Likewise.
9017         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9018         fusion and secondary reload support that were never used.
9019         (rs6000_option_override_internal): Delete TOC fusion, that was only
9020         partially defined, and it did not work unless you also used the
9021         -mcmodel= switch.
9022         (rs6000_legitimate_address_p): Delete TOC fusion support.
9023         (rs6000_opt_masks): Likewise.
9024         (fusion_wrap_memory_address): Delete function, no longer used.
9025         (fusion_split_address); Delete TOC fusion support.
9026         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9027         longer used with toc fusion being deleted.
9028         (TARGET_TOC_FUSION_FP): Likewise.
9029         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9030         UNSPEC.
9031         (toc fusion spliter): Delete TOC fusion support.
9032         (toc_fusionload_<mode>): Likewise.
9033         (toc_fusionload_di): Likewise.
9034         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9035         longer needs to be named.  Rename predicate to delete TOC fusion.
9036         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9037         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9038         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9039         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9040         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9042 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9044         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9045         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9046         __int128 in built-in function prototypes.
9047         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9048         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9050 2018-07-27  Martin Sebor  <msebor@redhat.com>
9052         PR tree-optimization/86696
9053         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9054         types, including enums.
9055         (handle_char_store): Be prepared for the above function to fail.
9057 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9059         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9060         when optimization level is lower than 2 or optimize for size.
9062 2018-07-26  Martin Sebor  <msebor@redhat.com>
9064         PR tree-optimization/86043
9065         PR tree-optimization/86042
9066         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9067         (get_string_cst_length): Rename...
9068         (get_min_string_length): ...to this.  Add argument.
9069         (handle_char_store): Extend to handle multi-character stores by
9070         MEM_REF.
9071         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9072         * tree.h (initializer_zerop): Add argument.
9074 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9076         PR middle-end/86660
9077         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9078         declare target to variables if they have always,{to,from,tofrom} map
9079         kinds.
9081 2018-07-26  Martin Liska  <mliska@suse.cz>
9083         PR lto/86548
9084         * lto-wrapper.c: Add linker_output as prefix
9085         for ltrans_output_file.
9087 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9089         PR rtl-optimization/85805
9090         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9091         value for hard registers if that was written in the same mode.
9093 2018-07-26  Martin Liska  <mliska@suse.cz>
9095         PR gcov-profile/86536
9096         * gcov.c (format_gcov): Use printf format %.*f directly
9097         and do not handle special values.
9099 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9101         * common/config/arc/arc-common.c (arc_option_optimization_table):
9102         Update default optimizations for size.
9104 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9106         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9107         * config/arc/constraints.md (CfZ): New constraint.
9108         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9109         (subsf3_fpu): Likewise.
9110         (cmpsf_fpu): Likewise.
9111         (cmpsf_fpu_uneq): Likewise.
9113 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9115         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9116         accesses as well.
9117         (arc_is_uncached_mem_p): uncached applies to both the variable and
9118         the pointer.
9120 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9122         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9123         register names.
9125 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9127         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9128         field "m_scopes" from vec to auto_vec.
9130 2018-07-25  Martin Liska  <mliska@suse.cz>
9132         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9133         return type.
9135 2018-07-25  Richard Biener  <rguenther@suse.de>
9137         PR debug/86654
9138         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9139         special wrt context_die late.
9140         (gen_subprogram_die): Re-use DIEs in local scope.
9142 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9144         PR tree-optimization/86644
9145         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9147 2018-07-25  Martin Liska  <mliska@suse.cz>
9149         PR middle-end/86645
9150         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9151         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9153 2018-07-25  Martin Liska  <mliska@suse.cz>
9155         PR sanitizer/79635
9156         * params.def: Explain ASan abbreviation and provide
9157         a documentation link.
9159 2018-07-24  Martin Sebor  <msebor@redhat.com>
9161         PR tree-optimization/86622
9162         PR tree-optimization/86532
9163         * builtins.h (string_length): Declare.
9164         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9165         (check_access): Be prepared for non-constant length ranges.
9166         (string_length): Make extern.
9167         * expr.c (string_constant): Only handle the minor non-constant
9168         array index.  Use string_constant to compute the length of
9169         a generic string constant.
9171 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9173         PR tree-optimization/86618
9174         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9175         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9177 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9179         PR tree-optimization/86636
9180         * json.cc (json::object::set): Fix comment.  Add assertions.
9181         (json::array::append): Move here from json.h.  Add comment and an
9182         assertion.
9183         (json::string::string): Likewise.
9184         * json.h (json::array::append): Move to json.cc.
9185         (json::string::string): Likewise.
9186         * optinfo-emit-json.cc
9187         (optrecord_json_writer::impl_location_to_json): Assert that we
9188         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9189         wrapper around it.  Expand the location once, rather than three
9190         times.
9191         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9192         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9193         wrappers.
9194         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9195         places.  Fix some overlong lines.
9197 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9199         * config/aarch64/aarch64-simd.md
9200         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9201         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9202         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9203         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9204         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9205         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9206         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9207         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9208         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9210 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9212         PR middle-end/86627
9213         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9214         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9215         and abs_d == d, do the power of two handling if profitable.
9217 2018-07-24  Richard Biener  <rguenther@suse.de>
9219         * match.pd: Add BIT_FIELD_REF canonicalizations.
9221 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9223         PR c/86617
9224         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9226 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9228         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9229         terminated STRING_CST object.
9231 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9233         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9235 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9237         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9238         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9239         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9240         the elements into a register.
9241         (rs6000_split_v4si_init_di_reg): Delete.
9242         (rs6000_split_v4si_init): Delete.
9243         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9244         (vsx_init_v4si): Rewrite as a define_expand.
9246 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9248         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9249         zero_extend argument from memory): New.
9251 2018-07-22  Martin Sebor  <msebor@redhat.com>
9253         PR bootstrap/86621
9254         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9255         diagnosing calls with unknown arguments unless -Walloca-larger-than
9256         is restricted to less than PTRDIFF_MAX bytes.
9258 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9260         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9262 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9264         * pretty-print.c (text_info::set_location): Remove redundant
9265         "line_table" parameter from call to rich_location::set_range.
9267 2018-07-20  Martin Sebor  <msebor@redhat.com>
9269         PR middle-end/82063
9270         * builtins.c (expand_builtin_alloca): Adjust.
9271         * calls.c (alloc_max_size): Simplify.
9272         * cgraphunit.c (cgraph_node::expand): Adjust.
9273         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9274         variables.
9275         (frame_larger_than_size): Same.
9276         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9277         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9278         Initialize.
9279         * doc/invoke.texi (GCC Command Options): Document option arguments.
9280         Explain byte-size arguments and suffixes.
9281         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9282         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9283         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9284         * doc/options.texi (UInteger): Expand.
9285         (Host_Wide_Int, ByteSize): Document new properties.
9286         * final.c (final_start_function_1): Include sizes in an error message.
9287         * function.c (frame_offset_overflow): Same.
9288         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9289         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9290         Diagnose unbounded alloca calls only for limits of less than
9291         PTRDIFF_MAX.
9292         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9293         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9294         for alloca(0).
9295         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9296         only for limits of less than PTRDIFF_MAX.
9297         * langhooks-def.h (lhd_handle_option): Change function argument
9298         to HOST_WIDE_INT.
9299         * langhooks.c (lhd_handle_option): Same.
9300         * langhooks.h (handle_option): Same.
9301         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9302         ByteSize flags.
9303         (var_type, var_type_struct): Same.
9304         (var_set): Handle ByteSize flag.
9305         * optc-gen.awk: Add comments to output to ease debugging.  Make
9306         use of HOST_WIDE_INT where appropriate.
9307         * opts-gen-save.awk:  Use %lx to format unsigned long.
9308         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9309         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9310         arguments.  Parse bytes-size suffixes.
9311         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9312         (enum_value_to_arg): Same.
9313         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9314         (handle_option): Adjust.
9315         (generate_option): Change function argument to HOST_WIDE_INT.
9316         (cmdline_handle_error): Adjust.
9317         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9318         (set_option): Change function argument to HOST_WIDE_INT.
9319         (option_enabled): Handle cl_host_wide_int.
9320         (get_option_state): Handle CLVC_SIZE.
9321         (control_warning_option): Same.
9322         * opts.c (common_handle_option): Change function argument to
9323         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9324         OPT_Wvla_larger_than_.
9325         * opts.h (enum cl_var_type): Add an enumerator.
9326         * stor-layout.c (layout_decl): Print a more meaningful warning.
9327         * toplev.c (output_stack_usage): Adjust.
9329 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9331         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9332         call to inline_expand_builtin_string_cmp.
9333         (expand_builtin_strcmp): Likewise.
9334         (expand_builtin_strncmp): Likewise.
9335         (inline_string_cmp): Delete the last parameter, change char_type_node
9336         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9337         two operands.
9338         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9339         the inlining expansion on target where the type of the call has same or
9340         narrower precision than unsigned char.
9342 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9344         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9345         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9346         * common.opt (fsave-optimization-record): New option.
9347         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9348         * doc/invoke.texi (-fsave-optimization-record): New option.
9349         * dumpfile.c: Include "optinfo-emit-json.h".
9350         (struct kv_pair): Move to coretypes.h.
9351         (optgroup_options): Make non-static.
9352         (dump_context::end_scope): Call
9353         optimization_records_maybe_pop_dump_scope.
9354         * dumpfile.h (optgroup_options): New decl.
9355         * json.cc: New file.
9356         * json.h: New file.
9357         * optinfo-emit-json.cc: New file.
9358         * optinfo-emit-json.h: New file.
9359         * optinfo.cc: Include "optinfo-emit-json.h".
9360         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9361         (optinfo_enabled_p): Check optimization_records_enabled_p.
9362         (optinfo_wants_inlining_info_p): Likewise.
9363         * optinfo.h: Update comment.
9364         * profile-count.c (profile_quality_as_string): New function.
9365         * profile-count.h (profile_quality_as_string): New decl.
9366         (profile_count::quality): New accessor.
9367         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9368         and optinfo_emit_json_cc_tests.
9369         * selftest.h (selftest::json_cc_tests): New decl.
9370         (selftest::optinfo_emit_json_cc_tests): New decl.
9371         * toplev.c: Include "optinfo-emit-json.h".
9372         (compile_file): Call optimization_records_finish.
9373         (do_compile): Call optimization_records_start.
9374         * tree-ssa-live.c: Include optinfo.h.
9375         (remove_unused_scope_block_p): Retain inlining information if
9376         optinfo_wants_inlining_info_p returns true.
9378 2018-07-20  Richard Biener  <rguenther@suse.de>
9380         PR debug/86585
9381         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9382         to cover -flto-partition=none.
9384 2018-07-20  Martin Liska  <mliska@suse.cz>
9386         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9387         (get_decl_source_range): Remove unused function.
9389 2018-07-20  Richard Biener  <rguenther@suse.de>
9391         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9392         (struct vn_phi_s): Likewise.
9393         (struct vn_reference_s): Likewise.
9394         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9395         for searching the slot of an entry known to be in the hash itself.
9396         (vn_phi_hasher::equal): Likewise.
9397         (vn_reference_hasher::equal): Likewise.
9398         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9399         globals.
9400         (optimistic_info, current_info): Remove, keeping only valid_info.
9401         (vn_reference_lookup_1): Remove fallback lookup.
9402         (vn_reference_lookup_2): Likewise.
9403         (vn_nary_op_lookup_1): Likewise.
9404         (vn_phi_lookup): Likewise.
9405         (vn_nary_build_or_lookup_1): Make sure to not chain the built
9406         hash element.
9407         (vn_reference_insert): Adjust, chain the inserted hash element
9408         at last_inserted_ref.
9409         (vn_reference_insert_pieces): Likewise.
9410         (visit_reference_op_call): Likewise.
9411         (vn_nary_op_insert_into): Chain the inserted hash element at
9412         last_inserted_nary.
9413         (vn_nary_op_insert_pieces): Adjust.
9414         (vn_nary_op_insert): Likewise.
9415         (vn_nary_op_insert_stmt): Likewise.
9416         (vn_phi_insert): Adjust, chain the inserted hash element at
9417         last_inserted_phi.
9418         (process_scc): Remove clearing and copying the optimistic
9419         table.  Instead remove elements inserted during an optimistic
9420         iteration from the single table we maintain.
9421         (init_scc_vn): Adjust.
9422         (free_scc_vn): Likewise.
9423         (sccvn_dom_walker::record_cond): Likewise.
9424         (sccvn_dom_walker::after_dom_children): Likewise.
9426 2018-07-19  Martin Sebor  <msebor@redhat.com>
9428         PR tree-optimization/84047
9429         PR tree-optimization/83776
9430         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9431         (check_array_bounds): Call it.
9433 2018-07-19  Martin Sebor  <msebor@redhat.com>
9435         * align.h (align_flags): Use member initialization.
9437 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
9439         * Makefile.in (OBJS): Add optinfo.o.
9440         * coretypes.h (class symtab_node): New forward decl.
9441         (struct cgraph_node): New forward decl.
9442         (class varpool_node): New forward decl.
9443         * dump-context.h: New file.
9444         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9445         "tree-pass.h".
9446         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9447         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9448         (set_alt_dump_file): Likewise.
9449         (dump_context::~dump_context): New dtor.
9450         (dump_gimple_stmt): Move implementation to...
9451         (dump_context::dump_gimple_stmt): ...this new member function.
9452         Add the stmt to any pending optinfo, creating one if need be.
9453         (dump_gimple_stmt_loc): Move implementation to...
9454         (dump_context::dump_gimple_stmt_loc): ...this new member function.
9455         Start a new optinfo and add the stmt to it.
9456         (dump_gimple_expr): Move implementation to...
9457         (dump_context::dump_gimple_expr): ...this new member function.
9458         Add the stmt to any pending optinfo, creating one if need be.
9459         (dump_gimple_expr_loc): Move implementation to...
9460         (dump_context::dump_gimple_expr_loc): ...this new member function.
9461         Start a new optinfo and add the stmt to it.
9462         (dump_generic_expr): Move implementation to...
9463         (dump_context::dump_generic_expr): ...this new member function.
9464         Add the tree to any pending optinfo, creating one if need be.
9465         (dump_generic_expr_loc): Move implementation to...
9466         (dump_context::dump_generic_expr_loc): ...this new member
9467         function.  Add the tree to any pending optinfo, creating one if
9468         need be.
9469         (dump_printf): Move implementation to...
9470         (dump_context::dump_printf_va): ...this new member function.  Add
9471         the text to any pending optinfo, creating one if need be.
9472         (dump_printf_loc): Move implementation to...
9473         (dump_context::dump_printf_loc_va): ...this new member function.
9474         Start a new optinfo and add the stmt to it.
9475         (dump_dec): Move implementation to...
9476         (dump_context::dump_dec): ...this new member function.  Add the
9477         value to any pending optinfo, creating one if need be.
9478         (dump_context::dump_symtab_node): New member function.
9479         (dump_context::get_scope_depth): New member function.
9480         (dump_context::begin_scope): New member function.
9481         (dump_context::end_scope): New member function.
9482         (dump_context::ensure_pending_optinfo): New member function.
9483         (dump_context::begin_next_optinfo): New member function.
9484         (dump_context::end_any_optinfo): New member function.
9485         (dump_context::s_current): New global.
9486         (dump_context::s_default): New global.
9487         (dump_scope_depth): Delete global.
9488         (dumpfile_ensure_any_optinfo_are_flushed): New function.
9489         (dump_symtab_node): New function.
9490         (get_dump_scope_depth): Reimplement in terms of dump_context.
9491         (dump_begin_scope): Likewise.
9492         (dump_end_scope): Likewise.
9493         (selftest::temp_dump_context::temp_dump_context): New ctor.
9494         (selftest::temp_dump_context::~temp_dump_context): New dtor.
9495         (selftest::verify_item): New function.
9496         (ASSERT_IS_TEXT): New macro.
9497         (ASSERT_IS_TREE): New macro.
9498         (ASSERT_IS_GIMPLE): New macro.
9499         (selftest::test_capture_of_dump_calls): New test.
9500         (selftest::dumpfile_c_tests): Call it.
9501         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9502         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9503         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9504         descriptive comment.
9505         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9506         (dump_node, dump_bb): Move these unrelated decls.
9507         (class dump_manager): Add leading comment.
9508         * optinfo.cc: New file.
9509         * optinfo.h: New file.
9511 2018-07-19  Michael Collison  <michael.collison@arm.com>
9512             Richard Henderson <rth@redhat.com>
9514         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
9515         (subti): Handle op1 zero.
9516         (subvti4, usub4ti4): New.
9517         (*sub<GPI>3_compare1_imm): New.
9518         (sub<GPI>3_carryinCV): New.
9519         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
9520         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
9522 2018-07-19  Michael Collison  <michael.collison@arm.com>
9523             Richard Henderson <rth@redhat.com>
9525         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
9526         (addti3): Create simpler code if low part is already known to be 0.
9527         (addvti4, uaddvti4): New.
9528         (*add<GPI>3_compareC_cconly_imm): New.
9529         (*add<GPI>3_compareC_cconly): New.
9530         (*add<GPI>3_compareC_imm): New.
9531         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
9532         handle constants within this pattern..
9533         (*add<GPI>3_compareV_cconly_imm): New.
9534         (*add<GPI>3_compareV_cconly): New.
9535         (*add<GPI>3_compareV_imm): New.
9536         (add<GPI>3_compareV): New.
9537         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
9538         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
9539         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
9540         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
9541         with 'comparison' operator.
9542         (*add<GPI>3_compareV_cconly_imm): Ditto.
9543         (*add<GPI>3_compareV_cconly): Ditto.
9544         (*add<GPI>3_compareV_imm): Ditto.
9545         (add<GPI>3_compareV): Ditto.
9546         (add<mode>3_carryinC): Ditto.
9547         (*add<mode>3_carryinC_zero): Ditto.
9548         (*add<mode>3_carryinC): Ditto.
9549         (add<mode>3_carryinV): Ditto.
9550         (*add<mode>3_carryinV_zero): Ditto.
9551         (*add<mode>3_carryinV): Ditto.
9553 2018-07-19  Michael Collison  <michael.collison@arm.com>
9554             Richard Henderson <rth@redhat.com>
9556         * config/aarch64/aarch64-modes.def (CC_V): New.
9557         * config/aarch64/aarch64-protos.h
9558         (aarch64_addti_scratch_regs): Declare
9559         (aarch64_subvti_scratch_regs): Declare.
9560         (aarch64_expand_subvti): Declare.
9561         (aarch64_gen_unlikely_cbranch): Declare
9562         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
9563         for signed overflow using CC_Vmode.
9564         (aarch64_get_condition_code_1): Handle CC_Vmode.
9565         (aarch64_gen_unlikely_cbranch): New function.
9566         (aarch64_addti_scratch_regs): New function.
9567         (aarch64_subvti_scratch_regs): New function.
9568         (aarch64_expand_subvti): New function.
9570 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9572         * config/aarch64/aarch64-option-extensions.def: New entry for profile
9573         extension.
9574         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
9575         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
9576         extension.
9578 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9580         PR target/83009
9581         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9582         address check not strict.
9584 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9586         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
9587         Umq with Umn.
9588         (store_pair_lanes<mode>): Likewise.
9589         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
9590         enum value 'ADDR_QUERY_LDP_STP_N'.
9591         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
9592         (aarch64_print_address_internal): Add declaration.
9593         (aarch64_print_ldpstp_address): Remove.
9594         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
9595         (aarch64_print_operand): Change printing of 'y'.
9596         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
9597         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
9598         'true' rather than '1'.
9599         * config/aarch64/constraints.md (Uml): Likewise.
9600         (Uml): Rename to Umn.
9601         (Umq): Remove.
9603 2018-07-19  Richard Biener  <rguenther@suse.de>
9605         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
9606         a trailing array.
9607         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
9608         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
9609         (vn_reference_hasher): Likewise.
9610         (struct vn_tables_s): Remove obstack and alloc-pool members.
9611         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
9612         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
9613         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
9614         (vn_reference_insert_pieces): Likewise.
9615         (alloc_vn_nary_op_noinit): Adjust.
9616         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
9617         (vn_phi_eq): Adjust.
9618         (shared_lookup_phiargs): Remove.
9619         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
9620         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
9621         (visit_reference_op_call): Likewise.
9622         (copy_nary, copy_phi, copy_reference): Remove.
9623         (process_scc): Rewind the obstack when iterating.  Do not
9624         copy the elements to valid_info but just move them from one
9625         hashtable to the other.
9626         (allocate_vn_table): Adjust.
9627         (free_vn_table): Likewise.
9628         (init_scc_vn): Likewise.
9629         (free_scc_vn): Likewise.
9631 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
9633         PR target/86560
9634         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
9635         indirect_return as function type attribute.
9636         (ix86_attribute_table): Change indirect_return to function
9637         type attribute.
9638         * doc/extend.texi: Update indirect_return attribute.
9640 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
9642         * wide-int.h (widest2_int): New.
9643         * gimple-fold.c (arith_overflowed_p): Use it.
9644         * tree.h (widest2_int_cst): New.
9645         * tree-vrp.c (wide_int_binop_overflow): Rename from
9646         vrp_int_const_binop.
9647         Rewrite to work on trees.
9648         (extract_range_from_multiplicative_op_1): Abstract code to...
9649         (wide_int_range_min_max): ...here.
9650         (wide_int_range_cross_product): ...and here.
9651         (extract_range_from_binary_expr_1): Abstract overflow code to...
9652         (wide_int_range_mult_wrapping): ...here.
9653         * tree-vrp.h (wide_int_range_cross_product): New.
9654         (wide_int_range_mult_wrapping): New.
9656 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
9657             Julia Koval  <julia.koval@intel.com>
9659         * config/i386/x86-tune-costs.h (skylake_memcpy,
9660         skylake_memset): Replace rep_prefix with unrolling for size 512.
9662 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
9664         PR middle-end/86544
9665         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
9666         comparision with EQ_EXPR in last stmt.
9668 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9670         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
9671         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
9672         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
9673         previously known as "PowerPC AltiVec Built-in Functions".  Move
9674         some material to new subsubsections "PowerPC AltiVec Built-in
9675         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
9676         ISA 2.07".
9677         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
9678         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
9679         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
9680         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
9682 2018-07-18  Richard Biener  <rguenther@suse.de>
9684         PR tree-optimization/86557
9685         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
9686         EXACT_DIV_EXPR.
9688 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
9690         * config/s390/s390.c (s390_function_profiler): Generate CFI.
9692 2018-07-17  Jeff Law  <law@redhat.com>
9694         * config/arm/arm.c (get_label_padding): Update for recent
9695         changes to label_to_alignment.
9697         PR tree-optimization/86010
9698         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
9700         * config/mips/mips.c (vr4130_align_insns): Update for recent
9701         changes to label_to_alignment.
9703         * config/frv/frv.c (frv_label_align): Update for recent changes
9704         to label_to_alignment.
9706         * config/nios2/nios2.c (nios2_label_align): Update for recent
9707         changes which dropped ALIGN_LABELS_LOG.
9709 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
9711         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
9712         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
9714 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
9716         * config/arc/arc.c (arc_label_align): Use align_labels instead of
9717         deprecated align_labels_log.
9719 2018-07-17  Richard Biener  <rguenther@suse.de>
9721         PR lto/86456
9722         * dwarf2out.c (init_sections_and_labels): Always generate
9723         a debug_line_str_section for early LTO debug.
9724         (dwarf2out_finish): Reset debug_line_str_hash output early.
9725         Bump counter for extra dwarf5 .debug_loc labels to not conflict
9726         with fat LTO part.
9727         (dwarf2out_early_finish): Output debug_line_str.
9729 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9731         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
9732         index register on z196 or later.
9734 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9736         * config/s390/s390.c (s390_default_align): Set default function
9737         alignment to 16.
9738         (s390_override_options_after_change): Call s390_default align.
9739         (s390_option_override_internal): Call s390_default_align.
9740         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
9742 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9744         PR middle-end/86542
9745         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
9746         to _looptemp_ clauses, other than the first two.
9748 2018-07-17  Martin Liska  <mliska@suse.cz>
9750         * opts.c: Do not enable OPT_falign_* for -Os.
9752 2018-07-17  Martin Liska  <mliska@suse.cz>
9754         * align.h (MAX_CODE_ALIGN): New.
9755         (MAX_CODE_ALIGN_VALUE): New.
9756         * common/config/i386/i386-common.c (ix86_handle_option):
9757         (MAX_CODE_ALIGN): Moved to align.h.
9758         * final.c (MAX_CODE_ALIGN): Likewise.
9759         * opts.c (parse_and_check_align_values):
9760         (MAX_CODE_ALIGN): Likewise.
9761         (MAX_CODE_ALIGN_VALUE): Likewise.
9763 2018-07-17  Martin Liska  <mliska@suse.cz>
9765         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
9766         in order to fulfil coding style.
9767         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
9768         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9769         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9770         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
9771         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
9772         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
9773         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
9774         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
9775         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9777 2018-07-17  Martin Liska  <mliska@suse.cz>
9779         * align.h: New file.
9780         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
9781         directly.
9782         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
9783         align_flags of label_to_alignment.
9784         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
9785         align_flags class.
9786         * config/m68k/m68k.c: Do not use removed align_labels_value and
9787         align_loops_value.
9788         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9789         (LOOP_ALIGN): Likewise.
9790         (LABEL_ALIGN): Likewise.
9791         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9792         Remove not used macro.
9793         (rs6000_loop_align): Change return type to align_flags.
9794         (rs6000_loop_align_max_skip): Remove.
9795         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9796         Change return type to align_flags.
9797         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9798         Remove not used macro.
9799         (rs6000_loop_align):  Change return type to align_flags.
9800         (rs6000_loop_align_max_skip): Remove.
9801         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9802         * config/rx/rx-protos.h (rx_align_for_label): Make it
9803         static function.
9804         * config/rx/rx.c (rx_align_for_label): Change return type
9805         to align_flags.
9806         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9807         macro definitions.
9808         into align_flags class.
9809         (LABEL_ALIGN): Likewise.
9810         (LOOP_ALIGN): Likewise.
9811         * config/s390/s390.c (s390_label_align): Use align_flags
9812         class member.
9813         (s390_asm_output_function_label): Likewise.
9814         * config/sh/sh.c (sh_override_options_after_change):
9815         Use align_flags class directly without macros.
9816         (find_barrier): Likewise.
9817         (barrier_align): Likewise.
9818         (sh_loop_align): Likewise.
9819         * config/spu/spu.c (spu_option_override):
9820         Use align_flags_tuple::get_value instead of removed macros.
9821         (spu_sched_init): Likewise.
9822         * config/spu/spu.h (GTY): Likewise.
9823         * config/visium/visium.c (visium_option_override):
9824         Set "8" as default secondary alignment.
9825         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9826         in order to guarantee secondary alignment of 8.
9827         * coretypes.h: Include align.h header file.
9828         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9829         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9830         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9831         * doc/tm.texi.in: Likewise.
9832         * final.c (struct label_alignment): Remove not used structure.
9833         (LABEL_ALIGN): Change type to align_flags.
9834         (LOOP_ALIGN): Likewise.
9835         (JUMP_ALIGN): Likewise.
9836         (default_loop_align_max_skip): Remove.
9837         (default_label_align_max_skip): Likewise.
9838         (default_jump_align_max_skip): Likewise.
9839         (default_label_align_after_barrier_max_skip):
9840         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9841         (LABEL_TO_MAX_SKIP): Remove.
9842         (label_to_alignment): Return align_flags type instead of integer.
9843         (label_to_max_skip): Remove.
9844         (align_fuzz): Use align_flags type.
9845         (compute_alignments): Use align_flags type and use align_flags::max
9846         to combine multiple alignments.
9847         (grow_label_align): Grow vec instead of C array.
9848         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9849         (shorten_branches):  Use align_flags type and use align_flags::max
9850         to combine multiple alignments.
9851         (final_scan_insn_1): Remove usage of secondary alignment that comes
9852         from label alignment, but instead use proper secondary alignment
9853         which is computed in grow_label_align.
9854         * flags.h (struct align_flags_tuple): Move to align.h.
9855         (struct align_flags): Likewise.
9856         (state_align_loops): Rename to align_loops.
9857         (state_align_jumps): Rename to align_jumps.
9858         (state_align_labels): Rename to align_labels.
9859         (state_align_functions): Rename to align_functions.
9860         (align_loops_log): Remove.
9861         (align_jumps_log): Remove.
9862         (align_labels_log): Remove.
9863         (align_functions_log): Remove.
9864         (align_loops_max_skip): Remove.
9865         (align_jumps_max_skip): Remove.
9866         (align_labels_max_skip): Remove.
9867         (align_functions_max_skip): Remove.
9868         (align_loops_value): Remove.
9869         (align_jumps_value): Remove.
9870         (align_labels_value): Remove.
9871         (align_functions_value): Remove.
9872         * output.h (label_to_alignment): Change return type to align_flags.
9873         (label_to_max_skip): Remove.
9874         * target.def: Remove loop_align_max_skip, label_align_max_skip,
9875         jump_align_max_skip macros.
9876         * targhooks.h (default_loop_align_max_skip): Remove.
9877         (default_label_align_max_skip): Likewise.
9878         (default_jump_align_max_skip): Likewise.
9879         (default_label_align_after_barrier_max_skip): Remove.
9880         * toplev.c (read_log_maxskip): Use ::normalize function.
9881         (parse_N_M): Remove not used argument and also call ::normalize.
9882         (parse_alignment_opts): Do not pass unused arguments.
9883         * varasm.c (assemble_start_function): Use directly align_functions
9884         instead of removed macros.
9885         * system.h: Do not poison removed macros.
9887 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9889         PR middle-end/86539
9890         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9891         and cond temporaries don't have reference type if iterator has
9892         pointer type.  For init use &for_pre_body instead of pre_p if
9893         for_pre_body is non-empty.
9895 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9897         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9898         double-double modes to SFmode directly directly.
9899         (trunc<mode>sf2_fprs): Delete.
9901 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9903         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9904         for conversions between IFmode and the decimal floating point modes.
9905         (init_float128_ieee): Use the correct names for conversions between
9906         KFmode and the decimal floating point modes.
9908 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9910         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9911         for the conversions between TDmode and IFmode.
9912         (init_float128_ieee): Use more correct names for the conversions
9913         between TDmode and KFmode.
9915 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9917         PR tree-optimization/86526
9918         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9919         (inline_expand_builtin_string_cmp): Likewise.
9920         (inline_string_cmp): Likewise.  Use c_readstr instead of
9921         builtin_memcpy_read_str.  Add unit_mode temporary.
9923 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9925         PR middle-end/86528
9926         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9927         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9929 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9931         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9932         Alphabetize prototypes of built-in functions, separating out
9933         built-in functions that are listed in this section but should be
9934         described elsewhere.
9936 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9938         PR target/86511
9939         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9940         conditional move when trapping comparison was split to a
9941         non-trapping one (and vice versa).
9943 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9945         * config/s390/s390.c (s390_function_profiler): Generate nops
9946         instead of profiler call sequences.
9947         * config/s390/s390.opt: Add the new option.
9949 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9951         * config/s390/s390.c (s390_function_profiler): Generate
9952         __mcount_loc section.
9953         * config/s390/s390.opt: Add the new option.
9955 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9957         * common.opt: Add the new warning.
9958         * config/s390/s390.c (s390_function_profiler): Emit "brasl
9959         %r0,__fentry__" when -mfentry is specified.
9960         (s390_option_override_internal): Disallow -mfentry for 31-bit
9961         CPUs.
9962         * config/s390/s390.opt: Add the new option.
9964 2018-07-16  Richard Biener  <rguenther@suse.de>
9966         PR lto/86523
9967         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9968         for function-local FUNCTION_DECL and RESULT_DECL immediately.
9970 2018-07-16  Martin Liska  <mliska@suse.cz>
9972         PR ipa/86529
9973         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9974         to ::get_create.
9976 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9978         * config/arc/arcHS.md: Update ARCHS scheduling rules.
9980 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9982         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9983         for ARCHS4x.
9984         * config/arc/arc-cpus.def (hs4x): New cpu.
9985         (hs4xd): Likewise.
9986         * config/arc/arc-tables.opt: Regenerate.
9987         * config/arc/arc.c (arc_sched_issue_rate): New function.
9988         (TARGET_SCHED_ISSUE_RATE): Define.
9989         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9990         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9991         fpu_cvt.
9992         (attr tune): Add ARCHS4x tune values.
9993         (attr tune_dspmpy): Define.
9994         (*tst): Correct instruction type.
9995         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9996         * config/arc/arcHS4x.md: New file.
9997         * config/arc/fpu.md: Update instruction type attributes.
9998         * config/arc/t-multilib: Regenerate.
10000 2018-07-16  Tom de Vries  <tdevries@suse.de>
10002         PR debug/86455
10003         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10005 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10007         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10008         early clobber.
10010 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10012         PR tree-optimization/86514
10013         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10014         conversion to a boolean type from a type with greater precision.
10016 2018-07-16  Tom de Vries  <tdevries@suse.de>
10018         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10019         dump_flags request TDF_SLIM.
10021 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10023         * fold-const.c (int_const_binop_1): Abstract...
10024         (wide_int_binop): ...wide int code here.
10025         (poly_int_binop): ...poly int code here.
10026         Abstract the rest of int_const_binop_1 into int_const_binop.
10027         * fold-const.h (wide_int_binop): New.
10028         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10029         Remove useless PLUS/MINUS_EXPR case.
10030         (zero_nonzero_bits_from_vr): Move wide int code...
10031         (zero_nonzero_bits_from_bounds): ...here.
10032         (extract_range_from_binary_expr_1): Move mask optimization code...
10033         (range_easy_mask_min_max): ...here.
10034         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10035         (range_easy_mask_min_max): New.
10037 2018-07-15  Jeff Law  <law@redhat.com>
10039         PR target/85993
10040         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10041         block.
10043 2018-07-14  Jim Wilson  <jimw@sifive.com>
10045         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10047 2018-07-14  Paul Koning  <ni1d@arrl.net>
10049         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10051 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10053         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10054         copying section.
10056 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10057             Steve Munroe  <munroesj52@gmail.com>
10059         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10060         (_mm_andnot_si128): Likewise.
10061         (_mm_or_si128): Likewise.
10062         (_mm_xor_si128): Likewise.
10064 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10066         PR middle-end/78809
10067         * builtins.c (expand_builtin_memcmp): Inline the calls first
10068         when result_eq is false.
10069         (expand_builtin_strcmp): Inline the calls first.
10070         (expand_builtin_strncmp): Likewise.
10071         (inline_string_cmp): New routine. Expand a string compare
10072         call by using a sequence of char comparison.
10073         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10074         a call to str(n)cmp/memcmp.
10075         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10076         New option.
10077         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10079 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10081         * config/arm/driver-arm.c: Include arm-native.h.
10082         (host_detect_local_cpu): Use auto-generated data tables.
10083         (vendors, arm_cpu_table): Delete.  Move part information to ...
10084         * config/arm/arm-cpus.in: ... here.
10085         * config/arm/parsecpu.awk (gen_native): New function.
10086         (vendor, part): New CPU fields.
10087         (END): Add support for building the native CPU detection tables.
10088         * config/arm/t-arm (arm-native.h): Add build rule.
10089         (driver-arm.o): Add dependency on arm-native.h.
10091 2018-07-13  Richard Biener  <rguenther@suse.de>
10093         PR middle-end/85974
10094         * match.pd (addr1 - addr2): Allow either of the operand to
10095         have a conversion.
10097 2018-07-13  Tom de Vries  <tdevries@suse.de>
10099         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10100         in remap_ssa_name.
10102 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10104         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10105         arrays instead of numbered variables.
10107 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10109         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10110         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10112 2018-07-13  Richard Biener  <rguenther@suse.de>
10114         PR debug/86452
10115         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10116         instead of get_context_die.
10118 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10119             Richard Biener  <rguenther@suse.de>
10121         PR middle-end/86489
10122         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10123         that the loop latch destination where phi is defined.
10125 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10127         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10128         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10129         (riscv_set_current_function): Extract getting interrupt type to new
10130         function.
10131         (riscv_get_interrupt_type): New function.
10132         (riscv_merge_decl_attributes): New function, checking interrupt type is
10133         same.
10134         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10136 2018-07-12  Paul Koning  <ni1d@arrl.net>
10138         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10139         directive.
10141 2018-07-12  Paul Koning  <ni1d@arrl.net>
10143         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10144         zero reference, add doloop_end instead.
10145         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10146         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10147         detail for doloop_end.
10149 2018-07-12  Martin Sebor  <msebor@redhat.com>
10151         PR c/86453
10152         * attribs.c (decl_attributes): Reject conflicting attributes before
10153         calling attribute handlers.
10155 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10157         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10158          parameter.
10159         (gcc::dump_manager::get_dump_file_name): likewise.
10160         (dump_begin): Likewise.
10161         * dumpfile.h (dump_begin): Update prototype.
10162         (gcc::dump_manager::get_dump_file_name,
10163         gcc::dump_manager::get_dump_file_name): Update prototype.
10165 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10167         * internal-fn.h (vectorizable_internal_fn_p): New function.
10168         * tree-vect-slp.c (compatible_calls_p): Likewise.
10169         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10170         to internal functions.
10171         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10173 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10175         * fold-const.h (inverse_conditions_p): Declare.
10176         * fold-const.c (inverse_conditions_p): New function.
10177         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10178         that test the inverse condition of a conditional internal function.
10179         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10180         * internal-fn.c (internal_fn_mask_index): Handle conditional
10181         internal functions.
10182         (vectorized_internal_fn_supported_p): New function.
10183         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10184         (any_pred_load_store): Replace with...
10185         (need_to_predicate): ...this new variable.
10186         (redundant_ssa_names): New variable.
10187         (ifcvt_can_use_mask_load_store): Move initial checks to...
10188         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10189         for which a conditional internal function exists.
10190         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10191         instead of ifcvt_can_use_mask_load_store.  Update after variable
10192         name change.
10193         (predicate_load_or_store): New function, split out from
10194         predicate_mem_writes.
10195         (check_redundant_cond_expr): New function.
10196         (value_available_p): Likewise.
10197         (predicate_rhs_code): Likewise.
10198         (predicate_mem_writes): Rename to...
10199         (predicate_statements): ...this.  Use predicate_load_or_store
10200         and predicate_rhs_code.
10201         (combine_blocks, tree_if_conversion): Update after above name changes.
10202         (ifcvt_local_dce): Handle redundant_ssa_names.
10203         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10204         general conditional functions.
10205         * tree-vect-stmts.c (vectorizable_call): Likewise.
10207 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10208             Alan Hayward  <alan.hayward@arm.com>
10209             David Sherwood  <david.sherwood@arm.com>
10211         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10212         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10213         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10214         plus and minus and convert them into IFN_COND_FMA-based sequences.
10215         (convert_mult_to_fma): Handle conditional plus and minus.
10217 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10219         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10220         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10221         (cond_fnms_optab): New optabs.
10222         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10223         internal functions.
10224         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10225         * internal-fn.h (get_conditional_internal_fn): Declare.
10226         (get_unconditional_internal_fn): Likewise.
10227         * internal-fn.c (cond_ternary_direct): New macro.
10228         (expand_cond_ternary_optab_fn): Likewise.
10229         (direct_cond_ternary_optab_supported_p): Likewise.
10230         (FOR_EACH_COND_FN_PAIR): Likewise.
10231         (get_conditional_internal_fn): New function.
10232         (get_unconditional_internal_fn): Likewise.
10233         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10234         (gimple_match_op::gimple_match_op): Add a new overload for 5
10235         operands.
10236         (gimple_match_op::set_op): Likewise.
10237         (gimple_resimplify5): Declare.
10238         * genmatch.c (decision_tree::gen): Generate simplifications for
10239         5 operands.
10240         * gimple-match-head.c (gimple_simplify): Define an overload for
10241         5 operands.  Handle calls with 5 arguments in the top-level overload.
10242         (convert_conditional_op): Handle conversions from unconditional
10243         internal functions to conditional ones.
10244         (gimple_resimplify5): New function.
10245         (build_call_internal): Pass a fifth operand.
10246         (maybe_push_res_to_seq): Likewise.
10247         (try_conditional_simplification): Try converting conditional
10248         internal functions to unconditional internal functions.
10249         Handle 3-operand unconditional forms.
10250         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10251         Define ternary equivalents of the current rules for binary conditional
10252         internal functions.
10253         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10254         ternary operations.
10255         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10256         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10257         (optab): Handle them.
10258         (SVE_COND_FP_TERNARY): New int iterator.
10259         (sve_fmla_op, sve_fmad_op): New int attributes.
10260         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10261         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10262         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10264 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10266         * target.def (preferred_else_value): New target hook.
10267         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10268         * doc/tm.texi: Regenerate.
10269         * targhooks.h (default_preferred_else_value): Declare.
10270         * targhooks.c (default_preferred_else_value): New function.
10271         * internal-fn.h (conditional_internal_fn_code): Declare.
10272         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10273         (get_conditional_internal_fn): Use it.
10274         (conditional_internal_fn_code): New function.
10275         * gimple-match.h (gimple_match_cond): New struct.
10276         (gimple_match_op): Add a cond member function.
10277         (gimple_match_op::gimple_match_op): Update all forms to take a
10278         gimple_match_cond.
10279         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10280         for the suboperation, but don't specify a particular else_value.
10281         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10282         (visit_nary_op, visit_reference_op_load): Pass
10283         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10284         * gimple-match-head.c: Include tree-eh.h
10285         (convert_conditional_op): New function.
10286         (maybe_resimplify_conditional_op): Likewise.
10287         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10288         (gimple_resimplify2): Likewise.
10289         (gimple_resimplify3): Likewise.
10290         (gimple_resimplify4): Likewise.
10291         (maybe_push_res_to_seq): Return null for conditional operations.
10292         (try_conditional_simplification): New function.
10293         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10294         constructor.
10295         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10296         IFN_COND_* call.
10297         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10298         function.
10299         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10301 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10303         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10304         DECL_FCONTEXT
10305         (hash_tree): Do not hash DECL_FCONTEXT
10306         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10307         Do not stream DECL_FCONTEXT.
10308         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10309         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10311 2018-07-12  Richard Biener  <rguenther@suse.de>
10313         PR debug/86462
10314         * dwarf2out.c (gen_block_die): Only output blocks when they have
10315         at least one !DECL_IGNORED_P variable.
10317 2018-07-12  Richard Biener  <rguenther@suse.de>
10319         PR target/84829
10320         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10321         Remove -mieee-fp handling.
10323 2018-07-12  Richard Biener  <rguenther@suse.de>
10325         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10326         left-over from last patch.
10328 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10330         PR tree-optimization/86492
10331         * gimple-ssa-store-merging.c
10332         (imm_store_chain_info::coalesce_immediate_stores): Call
10333         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10335 2018-07-12  Richard Biener  <rguenther@suse.de>
10337         PR middle-end/86479
10338         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10339         move possibly trapping operations into the conditional.
10341 2018-07-12  Richard Biener  <rguenther@suse.de>
10343         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10344         (vn_lookup_simplify_result): Remove recursion limit applied
10345         here.
10346         (vn_nary_build_or_lookup_1): Adjust.
10347         (try_to_simplify): Likewise.
10348         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10349         here.
10350         (gimple_resimplify2): Likewise.
10351         (gimple_resimplify3): Likewise.
10352         (gimple_resimplify4): Likewise.
10354 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10356         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10357         Use __mmask64 type instead of __mmask8 for __M argument.
10358         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10359         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10360         __U argument.
10361         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10362         __mmask16 for __M argument.
10363         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10364         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10365         to __mmask16 instead of __mmask8.
10366         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10367         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10368         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10369         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10370         instead of __mmask16 for __U argument.
10371         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10372         __mmask16 instead of __mmask8 for __U argument.
10373         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10374         __U argument.
10375         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10376         __mmask16.
10377         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10378         argument.
10379         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10380         __U argument.
10381         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10382         __mmask16.
10383         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10384         of __mmask16.
10385         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10386         __U argument.
10387         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10388         __U argument.
10389         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10390         __U argument.
10391         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10392         __U argument.
10393         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10394         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10395         return type as well as __M argument type and all casts from __mmask8
10396         to __mmask32.
10397         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10398         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10399         return type as well as __M argument type and all casts from __mmask8
10400         to __mmask16.
10401         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10402         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10403         return type as well as __M argument type and all casts from __mmask8
10404         to __mmask32.
10405         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10406         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10407         return type as well as __M argument type and all casts from __mmask8
10408         to __mmask16.
10409         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10410         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10411         __mmask16.
10413 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
10415         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10416         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10417         for __U argument.
10419 2018-07-11  Paul Koning  <ni1d@arrl.net>
10421         * doc/md.texi (define_subst): Document how multiple occurrences of
10422         the same argument in the replacement pattern are handled.
10424 2018-07-11  Paul Koning  <ni1d@arrl.net>
10426         * doc/extend.texi (Common Variable Attributes): Move "mode" into
10427         alphabetical order.
10428         (Common Type Attributes): Add "mode" attribute.
10430 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
10432         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10433         stream DECL_ORIGINAL_TYPE.
10434         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10435         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10436         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10437         Do not walk original type.
10438         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10439         external decls.
10440         (write_ts_decl_non_common_tree_pointers): Do not stream
10441         DECL_ORIGINAL_TYPE
10442         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10443         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10445 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10447         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10448         thread twice from the same starting edge.
10450 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10452         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10453         * gimple.c (gimple_call_nonnull_result_p): ...here...
10454         (gimple_call_nonnull_arg): ...and here.
10455         * gimple.h (gimple_call_nonnull_result_p): New.
10456         (gimple_call_nonnull_arg): New.
10458 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
10460         * config/arm/arm-cpus.in: Move information from fpu field of each
10461         cpu definition to the isa field.
10462         * config/arm/parsecpu.awk (fpu): Delete match rule.
10463         (gen_comm_data): Don't add bits from the CPU's FPU entry.
10465 2018-07-11  Richard Biener  <rguenther@suse.de>
10467         PR debug/86457
10468         * dwarf2out.c (init_sections_and_labels): Use
10469         output_asm_line_debug_info consistently.
10470         (dwarf2out_early_finish): Likewise.
10471         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10472         type units.
10474 2018-07-11  Richard Biener  <rguenther@suse.de>
10476         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10477         Rework father_bb setting in a way to avoid propagating constants
10478         multiple times on a loop body.
10480 2018-07-10  Mark Wielaard  <mark@klomp.org>
10482         PR debug/86459
10483         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10485 2018-07-10  Richard Biener  <rguenther@suse.de>
10487         * hash-map.h (hash_map::iterator::operator*): Return
10488         references to key and value.
10490 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
10492         PR c++/86443
10493         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10494         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10495         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10496         (gimplify_omp_for): For composite loops, move outer
10497         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10498         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
10499         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
10500         TREE_LIST for both the original class iterator and the "last" helper
10501         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
10502         loop, remember has_decl_expr from outer composite loops for the
10503         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10505 2018-07-09  Martin Sebor  <msebor@redhat.com>
10507         PR middle-end/77357
10508         PR middle-end/86428
10509         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
10510         accessing implicitly initialized array elements.
10511         * expr.c (string_constant): Handle string initializers of
10512         character arrays within aggregates.
10513         * gimple-fold.c (fold_array_ctor_reference): Add argument.
10514         Store element offset.  As a special case, handle zero size.
10515         (fold_nonarray_ctor_reference): Same.
10516         (fold_ctor_reference): Add argument.  Store subobject offset.
10517         * gimple-fold.h (fold_ctor_reference): Add argument.
10519 2018-07-09  Paul Koning  <ni1d@arrl.net>
10521         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
10522         (pdp11_insn_cost): New function.
10523         (pdp11_md_asm_adjust): New function.
10524         (TARGET_INVALID_WITHIN_DOLOOP): Define.
10525         (pdp11_rtx_costs): Update to match machine better.
10526         (output_addr_const_pdp11): Correct format mismatch warnings.
10527         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
10528         * config/pdp11/pdp11.md: General change to add base_cost and/or
10529         length attributes for use by new pdp11_insn_cost function.
10530         (MIN_BRANCH): Correct definition.
10531         (MIN_SOB): Ditto.
10532         (doloop_end): Use standard pattern name for looping pattern.
10533         (doloop_end_nocc): New.
10534         (movsf): Add another constraint alternative.
10535         (zero_extendqihi2): Add constraint alternatives for not in place
10536         extend.
10537         (zero_extendhisi2): Remove.
10538         (shift patterns): Add CC handling variants.
10539         (bswaphi2): New.
10540         (bswapsi2): New.
10541         (rothi3): New.
10542         (define_peephole2): New peephole to recognize mov that sets CC for
10543         subsequent test.
10545 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10547         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
10548         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
10549         wi::add.
10551 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
10553         PR c/86420
10554         * real.c (real_nextafter): Return true if result is denormal.
10556 2018-07-09  Martin Liska  <mliska@suse.cz>
10558         * common.opt: Add back wrongly removed attribute.
10560 2018-07-09  Richard Biener  <rguenther@suse.de>
10562         PR debug/86413
10563         * dwarf2out.c (gen_block_die): For an early generated DIE
10564         always output high/low PC attributes.
10566 2018-07-09  Tom de Vries  <tdevries@suse.de>
10568         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
10569         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
10570         onto VAR_DECL with abstract origin.
10572 2018-07-07  Jim Wilson  <jimw@sifive.com>
10574         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
10576 2018-07-07  Tom de Vries  <tdevries@suse.de>
10578         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
10580 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
10582         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
10583         overflow_type.
10584         (combine_bound): Use wide-int overflow calculation instead of
10585         rolling our own.
10586         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
10587         overflow_type.
10588         * fold-const.c (int_const_binop_2): Same.
10589         (extract_muldiv_1): Same.
10590         (fold_div_compare): Same.
10591         (fold_abs_const): Same.
10592         * match.pd: Same.
10593         * poly-int.h (add): Same.
10594         (sub): Same.
10595         (neg): Same.
10596         (mul): Same.
10597         * predict.c (predict_iv_comparison): Same.
10598         * profile-count.c (slow_safe_scale_64bit): Same.
10599         * simplify-rtx.c (simplify_const_binary_operation): Same.
10600         * tree-chrec.c (tree_fold_binomial): Same.
10601         * tree-data-ref.c (split_constant_offset_1): Same.
10602         * tree-if-conv.c (idx_within_array_bound): Same.
10603         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
10604         * tree-ssa-phiopt.c (minmax_replacement): Same.
10605         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
10606         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
10607         * vr-values.c (vr_values::adjust_range_with_scev): Same.
10608         * wide-int.cc (wi::add_large): Same.
10609         (wi::mul_internal): Same.
10610         (wi::sub_large): Same.
10611         (wi::divmod_internal): Same.
10612         * wide-int.h: Change overflow type to overflow_type for neg, add,
10613         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
10614         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
10615         mul_internal, divmod_internal.
10616         (overflow_type): New enum.
10617         (accumulate_overflow): New.
10619 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
10621         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
10622         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
10624 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10626         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
10627         argument is checked for zero before entering loop, avoid checking again.
10629 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10631         * gimplify.h (generic_expr_could_trap_p): Set as global function.
10632         * gimplify.h (generic_expr_could_trap_p): Likwise.
10633         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
10635 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
10637         PR tree-optimization/86401
10638         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
10639         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
10640         (fold_bit_and_mask): ... here.  New helper function for match.pd.
10641         * fold-const.h (fold_bit_and_mask): Declare.
10642         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
10644 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
10646         PR target/86324
10647         * target.def (translate_mode_attribute): New hook.
10648         * targhooks.h (default_translate_mode_attribute): Declare.
10649         * targhooks.c (default_translate_mode_attribute): New function.
10650         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
10651         * doc/tm.texi: Regenerate.
10652         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
10653         (rs6000_translate_mode_attribute): New function.
10655 2018-07-06  Paul Koning  <ni1d@arrl.net>
10657         * doc/md.texi (define_split): Document DONE and FAIL.
10658         (define_peephole2): Ditto.
10660 2018-07-05  Jeff Law  <law@redhat.com>
10662         PR tree-optimization/86010
10663         * tree-ssa-dse.c (compute_trims): More aggressively trim at
10664         both the head and tail of mem* and str* calls.
10666 2018-07-05  Jim Wilson  <jimw@sifive.com>
10668         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
10670 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
10672         * config/aarch64/aarch64-simd.md: correct flags text for
10673         MIN_EXPR replacement.
10675 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
10677         * configure: Regenerated.
10679 2018-07-05  Carl Love  <cel@us.ibm.com>
10681         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
10682         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
10683         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
10684         VSX_BUILTIN_DOUBLEL_V4SF.
10686 2018-07-05  Martin Sebor  <msebor@redhat.com>
10688         PR c++/86400
10689         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
10690         than its domain to compute its the upper bound of a char array.
10692 2018-07-05  Nathan Sidwell  <nathan@acm.org>
10694         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
10695         * doc/cpp.texi: Update comment.
10696         * doc/tm.texi: Rebuilt.
10697         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
10698         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
10699         * doc/extend.texi (Backwards Compatibility): Clarify it is system
10700         headers affected by extern "C".
10701         * system.h: Poison NO_IMPLICIT_EXTERN_C.
10702         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
10703         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
10704         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
10705         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
10706         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
10707         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
10708         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
10709         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
10710         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
10711         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
10712         config/sparc/sp64-elf.h, config/spu/spu.h,
10713         config/stormy16/stormy16.h, config/v850/v850.h,
10714         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
10715         define NO_IMPLICIT_EXTERN_C.
10716         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
10718 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10720         PR target/84711
10721         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10722         instead of GET_MODE_SIZE when comparing Units.
10724 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10726         PR target/84711
10727         * rtlanal.c (set_noop_p): Constrain on mode change,
10728         include hard-reg-set.h
10730 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10732         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
10734 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
10736         Revert
10737         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10739         PR sanitizer/84250
10740         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10741         libasan.
10742         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10744 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10746         PR sanitizer/84250
10747         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10748         libasan.
10749         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10751 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
10753         PR middle-end/86380
10754         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
10756 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
10758         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
10759         neg_*_op* variables.
10761 2018-07-04  Martin Liska  <mliska@suse.cz>
10763         * tree-switch-conversion.c: Define
10764         max_ratio_for_speed and max_ratio_for_size constants.
10766 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10767             Martin Liska  <mliska@suse.cz>
10769         PR middle-end/66240
10770         PR target/45996
10771         PR c/84100
10772         * common.opt: Rename align options with 'str_' prefix.
10773         * common/config/i386/i386-common.c (set_malign_value): New
10774         function.
10775         (ix86_handle_option): Use it to set -falign-* options/
10776         * config/aarch64/aarch64-protos.h (struct tune_params): Change
10777         type from int to string.
10778         * config/aarch64/aarch64.c: Update default values from int
10779         to string.
10780         * config/alpha/alpha.c (alpha_override_options_after_change):
10781         Likewise.
10782         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
10783         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10784         max skip conditionally.
10785         * config/i386/freebsd.h (SUBALIGN_LOG): New.
10786         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10787         max skip conditionally.
10788         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10789         max skip conditionally.
10790         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10791         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10792         max skip conditionally.
10793         * config/i386/i386.c (struct ptt): Change type from int to
10794         string.
10795         (ix86_default_align): Set default values.
10796         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10797         max skip conditionally.
10798         * config/i386/iamcu.h (SUBALIGN_LOG): New.
10799         (ASM_OUTPUT_MAX_SKIP_ALIGN):
10800         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10801         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10802         max skip conditionally.
10803         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10804         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10805         * config/i386/x86-64.h (SUBALIGN_LOG): New.
10806         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10807         max skip conditionally.
10808         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10809         * config/ia64/ia64.c (ia64_option_override): Set default values
10810         for alignment options.
10811         * config/m68k/m68k.c: Handle new str_align_* options.
10812         * config/mips/mips.c (mips_set_compression_mode): Change
10813         type of constants.
10814         (mips_option_override): Set default values for options.
10815         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10816         Likewise.
10817         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10818         Likewise.
10819         * config/rx/rx.c (rx_option_override): Likewise.
10820         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10821         (LABEL_ALIGN): Use align_labels_log.
10822         (LOOP_ALIGN): Use align_loops_align.
10823         * config/s390/s390.c (s390_asm_output_function_label): Use new
10824         macros.
10825         * config/sh/sh.c (sh_override_options_after_change):
10826         Change type of constants.
10827         * config/spu/spu.c (spu_sched_init): Likewise.
10828         * config/sparc/sparc.c (sparc_option_override): Set default
10829         values for options.
10830         * config/visium/visium.c (visium_option_override): Likewise.
10831         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10832         emit p2align format with last argument if it's not needed.
10833         * doc/invoke.texi: Document extended format of -falign-*.
10834         * final.c: Use align_labels alignment.
10835         * flags.h (struct target_flag_state): Change type to use
10836         align_flags.
10837         (struct align_flags_tuple): New.
10838         (struct align_flags): Likewise.
10839         (align_loops_log): Redefine macro to use new types.
10840         (align_loops_max_skip): Redefine macro to use new types.
10841         (align_jumps_log): Redefine macro to use new types.
10842         (align_jumps_max_skip): Redefine macro to use new types.
10843         (align_labels_log): Redefine macro to use new types.
10844         (align_labels_max_skip): Redefine macro to use new types.
10845         (align_functions_log): Redefine macro to use new types.
10846         (align_loops): Redefine macro to use new types.
10847         (align_jumps): Redefine macro to use new types.
10848         (align_labels): Redefine macro to use new types.
10849         (align_functions): Redefine macro to use new types.
10850         (align_functions_max_skip): Redefine macro to use new types.
10851         (align_loops_value): New macro.
10852         (align_jumps_value): New macro.
10853         (align_labels_value): New macro.
10854         (align_functions_value): New macro.
10855         * function.c (invoke_set_current_function_hook): Propagate
10856         alignment values from flags to global variables default in
10857         topleev.h.
10858         * ipa-icf.c (sem_function::equals_wpa): Use
10859         cl_optimization_option_eq instead of memcmp.
10860         * lto-streamer.h (cl_optimization_stream_out): Support streaming
10861         of string types.
10862         (cl_optimization_stream_in): Likewise.
10863         * optc-save-gen.awk: Support strings in cl_optimization.
10864         * opth-gen.awk: Likewise.
10865         * opts.c (finish_options): Remove error checking of invalid
10866         value ranges.
10867         (MAX_CODE_ALIGN): Remove.
10868         (MAX_CODE_ALIGN_VALUE): Likewise.
10869         (parse_and_check_align_values): New function.
10870         (check_alignment_argument): Likewise.
10871         (common_handle_option): Use check_alignment_argument.
10872         * opts.h (parse_and_check_align_values): Declare.
10873         * toplev.c (init_alignments): Remove.
10874         (read_log_maxskip): New.
10875         (parse_N_M): Likewise.
10876         (parse_alignment_opts): Likewise.
10877         (backend_init_target): Remove usage of init_alignments.
10878         * toplev.h (parse_alignment_opts): Declare.
10879         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10880         argument.
10881         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10882         * tree.c (cl_option_hasher::equal): New.
10883         * varasm.c: Use new global macros.
10885 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10887         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10888         Use a simpler align directive also if MAXSKIP = ALIGN-1.
10889         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10890         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10891         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10892         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10893         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10894         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10895         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10896         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10897         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10898         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10899         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10901 2018-07-04  Martin Liska  <mliska@suse.cz>
10902             Jonathan Wakely  <jwakely@redhat.com>
10904         * coverage.c: Use correct type.
10905         * doc/invoke.texi: Language correction.
10907 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10909         PR target/85620
10910         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10911         ENDBRANCH for non-tail call which may return via indirect branch.
10912         * doc/extend.texi: Document indirect_return attribute.
10914 2018-07-03  Martin Sebor  <msebor@redhat.com>
10916         PR tree-optimization/86274
10917         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10918         precondition.
10919         (format_floating): Correct handling of infinities and NaNs.
10921 2018-07-03  Martin Sebor  <msebor@redhat.com>
10923         * print-tree.c (print_real_cst): New function.
10924         (print_node_brief): Call it.
10925         (print_node): Ditto.
10927 2018-07-03  Jeff Law  <law@redhat.com>
10929         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10930         into a single pattern.
10932         * config/h8300/h8300.md (ors code_iterator): New.
10933         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10934         a single pattern and single splitter.
10935         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10936         (iorqi3_1, xorqi3_1): Likewise.
10937         (iorqi3, xorqi3 expanders): Similarly.
10939         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10940         (movmd_internal) into a single pattern using the P mode iterator.
10941         (movmd splitters): Similarly.
10942         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10943         (movsd splitters): Similarly.
10945         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10946         ADDB, ADDW and ADDL into a single ADD attribute which selects the
10947         right table based on the size of the operand.
10948         * config/h8300/h8300.md (length_table): Corresponding changes. All
10949         references to "addb", "addw" and "addl" changed to "add".
10950         (btst patterns): Merge two variants into a single pattern.
10951         (tstqi, tsthi): Likewise.
10952         (addhi3_incdec, addsi3_incdec): Likewise.
10953         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10954         (mulhi3, mulsi3): Likewise.
10955         (udivhi3, udivsi3): Likewise.
10956         (divhi3, divsi3): Likewise.
10957         (andorqi3, andorhi3, andorsi3): Likewise.
10959 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
10961         PR target/85694
10962         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10963         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10965 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10967         PR tree-optimization/85694
10968         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10969         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10970         UNSPEC_URHADD.
10971         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10972         (<u>avg<mode>3_ceil): New patterns.
10974 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
10976         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10977         scan-tree-dump directive.
10978         * gcc.dg/vect/slp-perm-2.c: Likewise.
10979         * gcc.dg/vect/slp-perm-3.c: Likewise.
10980         * gcc.dg/vect/slp-perm-5.c: Likewise.
10981         * gcc.dg/vect/slp-perm-6.c: Likewise.
10982         * gcc.dg/vect/slp-perm-7.c: Likewise.
10983         * gcc.dg/vect/slp-perm-8.c: Likewise.
10985 2018-07-03  Marek Polacek  <polacek@redhat.com>
10987         PR middle-end/86202
10988         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10990 2018-07-03  Richard Biener  <rguenther@suse.de>
10992         PR ipa/86389
10993         * tree-ssa-structalias.c (find_func_clobbers): Properly
10994         handle indirect calls.
10996 2018-07-03  Jeff Law  <law@redhat.com>
10998         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10999         (shifts): New code iterator.
11000         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11001         expander.  Fix HImode handling on H8/SX.
11002         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11003         (subqi3, subhi3, subsi3 expanders): Likewise.
11004         (andqi3, andhi3, andsi3 expanders): Likewise.
11005         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11006         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11007         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11008         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11009         (zero_extendqihi2, zero_extendqisi2): Likewise.
11010         (extendqihi2, extendqisi2): Likewise.
11011         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11012         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11013         (rotlqi3_1, rotlhi3_1): Likewise.
11014         (logicalhi3_sn, logicalsi3_sn): Likewise.
11015         (logicalhi3, logicalsi3): Likewise.
11017 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11019         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11020         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11021         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11022         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11023         type to append_pattern_def_seq instead of creating a stmt_vec_info
11024         directly.
11025         (build_mask_conversion): Likewise.  Remove vinfo argument.
11026         (vect_add_conversion_to_patterm): Likewise, renaming to...
11027         (vect_add_conversion_to_pattern): ...this.
11028         (vect_recog_mask_conversion_pattern): Update call to
11029         build_mask_conversion.  Pass the vector type to
11030         append_pattern_def_seq here too.
11031         (vect_recog_gather_scatter_pattern): Update call to
11032         vect_add_conversion_to_pattern.
11034 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11036         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11037         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11038         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11039         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11040         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11041         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11042         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11043         append_pattern_def_seq instead of new_pattern_def_seq.
11044         (vect_recog_divmod_pattern): Do both of the above.
11045         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11046         is null.
11048 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11050         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11051         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11052         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11053         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11054         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11055         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11056         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11057         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11058         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11059         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11060         parameter with a single stmt_vec_info.
11061         (vect_recog_func_ptr): Likewise.
11062         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11063         (vect_try_gather_scatter_pattern): ...this.
11064         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11065         the stmt_vec_info of the statement to be matched.  Don't clear
11066         STMT_VINFO_RELATED_STMT.
11067         (vect_pattern_recog): Update call accordingly.
11069 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11071         PR tree-optimization/85694
11072         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11073         (uavgM3_ceil): Document new optabs.
11074         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11075         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11076         functions.
11077         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11078         (savg_ceil_optab): New optabs.
11079         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11080         (vect_vect_recog_func_ptrs): Add it.
11081         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11082         constant directly from the associated lhs.
11084 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11086         * tree-vect-patterns.c (vect_split_statement): New function.
11087         (vect_convert_input): Use it to try to split an existing cast.
11089 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11091         * poly-int.h (print_hex): New function.
11092         * dumpfile.h (dump_dec, dump_hex): Declare.
11093         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11094         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11095         min_input_precision, operation_precision and operation_sign.
11096         * tree-vect-patterns.c (vect_get_range_info): New function.
11097         (vect_same_loop_or_bb_p, vect_single_imm_use)
11098         (vect_operation_fits_smaller_type): Delete.
11099         (vect_look_through_possible_promotion): Add an optional
11100         single_use_p parameter.
11101         (vect_recog_over_widening_pattern): Rewrite to use new
11102         stmt_vec_info infomration.  Handle one operation at a time.
11103         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11104         (vect_truncatable_operation_p, vect_set_operation_type)
11105         (vect_set_min_input_precision): New functions.
11106         (vect_determine_min_output_precision_1): Likewise.
11107         (vect_determine_min_output_precision): Likewise.
11108         (vect_determine_precisions_from_range): Likewise.
11109         (vect_determine_precisions_from_users): Likewise.
11110         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11111         (vect_vect_recog_func_ptrs): Put over_widening first.
11112         Add cast_forwprop.
11113         (vect_pattern_recog): Call vect_determine_precisions.
11115 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11117         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11118         statements that have been replaced by further pattern statements.
11119         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11121 2018-07-03  Richard Biener  <rguenther@suse.de>
11123         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11124         always set *dt.  Dump vectype in vectype overload.
11125         * dumpfile.h (dump_gimple_expr): New function.
11126         (dump_gimple_expr_loc): Likewise.
11127         * dumpfile.c (dump_gimple_expr): New function.
11128         (dump_gimple_expr_loc): Likewise.
11130 2018-07-02  Jeff Law  <law@redhat.com>
11132         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11133         the H8/300, H8/300H and H8/S variants into a single pattern.
11134         (movhi_h8300, movqi_h8300hs): Similarly.
11135         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11136         (QHI mode iterator): New.
11138         * config/h8300/h8300.md: Remove trailing whitespace.
11140 2018-07-02  Jim Wilson  <jimw@sifive.com>
11142         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11143         instead of emit_insn for interrupt returns.
11144         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11145         (riscv_sret, riscv_uret): Likewise.
11147 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11149         * pretty-print.c (selftest::test_pp_format): Move save and restore
11150         of quotes to class auto_fix_quotes, and add an instance.
11151         * selftest.c: Include "intl.h".
11152         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11153         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11154         * selftest.h (selftest::auto_fix_quotes): New class.
11156 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11158         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11159         (aarch64_sve_prepare_conditional_op): Remove.
11160         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11161         Allow aarch64_simd_reg_or_zero as select operand; remove
11162         the aarch64_sve_prepare_conditional_op call.
11163         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11164         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11165         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11166         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11167         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11168         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11169         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11170         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11171         and a splitters to match all of the *_any patterns.
11172         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11174         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11175         (SVE_COND_FP_BINARY_REV): Remove.
11176         (sve_int_op_rev, sve_fp_op_rev): New.
11177         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11178         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11179         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11180         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11181         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11182         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11183         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11184         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11185         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11187         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11188         Remove match_dup 1 from the inner unspec.
11189         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11191         * config/aarch64/aarch64.md (movprfx): New attr.
11192         (length): Default movprfx to 8.
11193         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11194         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11195         (*<su>mul<SVE_I>3_highpart): Likewise.
11196         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11197         (*v<ASHIFT><SVE_I>3): Likewise.
11198         (*<su><MAXMIN><SVE_I>3): Likewise.
11199         (*<su><MAXMIN><SVE_F>3): Likewise.
11200         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11201         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11202         (*div<SVE_F>4): Likewise.
11204 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11206         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11207         in dump string.
11209 2018-07-02  Richard Biener  <rguenther@suse.de>
11211         PR tree-optimization/86363
11212         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11213         memset argument refers to a non-variable address.
11215 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11217         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11218         {PLUS,MINUS}_EXPR code to...
11219         (adjust_symbolic_bound): ...here,
11220         (combine_bound): ...here,
11221         (set_value_range_with_overflow): ...and here.
11223 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11225         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11226         code...
11227         (extract_range_from_abs_expr): ...here.
11229 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11231         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11232         -fno-omit-frame-pointer when not optimizing.
11234 2018-07-02  Martin Liska  <mliska@suse.cz>
11236         PR ipa/86279
11237         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11238         (propagate_nothrow): Likewise.
11240 2018-07-02  Martin Liska  <mliska@suse.cz>
11242         PR ipa/86323
11243         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11245 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11247         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11248         function in r262149, changing "loc" param from source_location to
11249         const dump_location_t &.
11250         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11251         declaration, as above.
11253 2018-07-01  Paul Koning  <ni1d@arrl.net>
11255         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11256         -munit-asm, -mgnu-asm, -mdec-asm.
11257         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11258         (pdp11_output_labelref): New.
11259         (pdp11_output_def): New.
11260         (pdp11_output_addr_vec_elt): New.
11261         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11262         %# and %@ format codes.
11263         (pdp11_option_override): New.
11264         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11265         (pdp11_output_ident): New.
11266         (pdp11_asm_named_section): New.
11267         (pdp11_asm_init_sections): New.
11268         (pdp11_file_start): New.
11269         (pdp11_file_end): New.
11270         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11271         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11272         %o, like %c but octal.
11273         (pdp11_option_override): New.
11274         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11275         -mdec-asm.
11276         (DATA_SECTION_ASM_OP): Ditto.
11277         (READONLY_DATA_SECTION_ASM_OP): New.
11278         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11279         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11280         (ASM_OUTPUT_LABELREF): Ditto.
11281         (ASM_OUTPUT_DEF): Ditto.
11282         (ASM_OUTPUT_EXTERNAL): New.
11283         (ASM_OUTPUT_SOURCE_FILENAME): New.
11284         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11285         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11286         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11287         %# and %@ format codes.
11288         * config/pdp11/pdp11.opt (mgnu-asm): New.
11289         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11290         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11291         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11293 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11295         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11296         dereferencing path[] beyond its length.
11297         (debug_path): New.
11298         (debug_all_paths): New.
11299         (rewire_first_differing_edge): New.
11300         (adjust_paths_after_duplication): New.
11301         (duplicate_thread_path): Call adjust_paths_after_duplication.
11302         Add new argument.
11303         (thread_through_all_blocks): Add new argument to
11304         duplicate_thread_path.
11306 2018-06-30  Jim Wilson  <jimw@sifive.com>
11308         * config/riscv/predicates.md (p2m1_shift_operand): New.
11309         (high_mask_shift_operand): New.
11310         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11311         pattern using p2m1_shift_operand.
11312         (lshsi3_zero_extend_3+2): New combiner pattern using
11313         high_mask_shift_operand.
11315 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11317         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11318         split out from...
11319         (vect_recog_rotate_pattern): ...here.
11320         (vect_convert_input): Try to insert casts of invariants in the
11321         preheader.
11322         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11323         preheader to be empty.
11325 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11327         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11328         vector type.  If given, install it in the new statement's
11329         STMT_VINFO_VECTYPE.
11330         (vect_element_precision): New function.
11331         (vect_unpromoted_value): New struct.
11332         (vect_unpromoted_value::vect_unpromoted_value): New function.
11333         (vect_unpromoted_value::set_op): Likewise.
11334         (vect_look_through_possible_promotion): Likewise.
11335         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11336         (vect_widened_op_tree, vect_convert_input): Likewise.
11337         (vect_convert_inputs, vect_convert_output): Likewise.
11338         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11339         to handle the optional cast of the multiplication result and
11340         vect_widened_op_tree to detect the widened multiplication itself.
11341         Do not require the input and output of promotion casts to have
11342         the same sign, but base the signedness of the operation on the
11343         input rather than the result.  If the pattern includes two
11344         promotions, check that those promotions have the same sign.
11345         Do not restrict the MULT_EXPR handling to a double-width result;
11346         handle quadruple-width results and wider.  Use vect_convert_inputs
11347         to convert the inputs to the common type.
11348         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11349         to handle the optional cast of the ABS result.  Also allow a sign
11350         change or a sign extension between the ABS and MINUS.
11351         Use vect_widened_op_tree to detect the widened subtraction and use
11352         vect_convert_inputs to convert the inputs to the common type.
11353         (vect_handle_widen_op_by_const): Delete.
11354         (vect_recog_widen_op_pattern): New function.
11355         (vect_recog_widen_mult_pattern): Use it.
11356         (vect_recog_widen_shift_pattern): Likewise.
11357         (vect_recog_widen_sum_pattern): Use
11358         vect_look_through_possible_promotion to handle the promoted
11359         PLUS_EXPR operand.
11361 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11363         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11364         the containing gimple_seq *.
11365         * gimple-iterator.h (gsi_for_stmt): Declare it.
11366         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11367         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11368         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11369         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11370         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11371         checks.
11372         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11373         split out from...
11374         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11375         statement being replaced is part of an existing pattern
11376         definition sequence, inserting the new pattern statements before
11377         the original one.
11378         (vect_pattern_recog_1): Don't return a bool.  If the statement
11379         is already part of a pattern, instead apply pattern matching
11380         to the pattern definition statements.  Don't clear the
11381         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11382         (vect_pattern_recog): Don't break after the first match;
11383         continue processing the pattern definition statements instead.
11384         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11386 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11388         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11389         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11390         (vect_recog_widen_sum_pattern): Use it.
11392 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11394         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11395         phi is not a pattern statement and has not been replaced by
11396         a pattern statement.
11397         * tree-vect-patterns.c (type_conversion_p): Don't check
11398         STMT_VINFO_IN_PATTERN_P.
11399         (vect_recog_vector_vector_shift_pattern): Likewise.
11400         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11401         the pattern statement rather than the original statement; check
11402         directly for a WIDEN_MULT_EXPR here.
11403         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11404         vect_is_simple_use to return the pattern statement rather
11405         than the original statement; use is_pattern_stmt_p to check
11406         for such a pattern statement.
11407         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11408         to return the pattern statement rather than the original statement;
11409         don't do the same transformation here.
11410         (vect_is_simple_use): If the defining statement has been replaced
11411         by a pattern statement, return the pattern statement instead.
11412         Remove the corresponding (local) transformation from the vectype
11413         overload.
11415 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11417         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11418         end and default to null.
11419         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11420         (vectorizable_reduction): Update calls accordingly, dropping the
11421         gimple ** argument if the passed-back statement isn't needed.
11422         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11423         (vect_recog_rotate_pattern): Likewise.
11424         (vect_recog_mask_conversion_pattern): Likewise.
11425         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11426         (vect_mask_constant_operand_p): Likewise.
11427         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11428         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11429         (get_group_load_store_type, get_load_store_type): Likewise.
11430         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11431         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11432         (vectorizable_conversion, vectorizable_assignment): Likewise.
11433         (vectorizable_shift, vectorizable_operation): Likewise.
11434         (vectorizable_store, vect_is_simple_cond): Likewise.
11435         (vectorizable_condition, vectorizable_comparison): Likewise.
11436         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11437         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11438         and move it to the end.  Cope with null def_stmt_outs.
11440 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11442         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11444 2018-06-29  Jeff Law  <law@redhat.com>
11446         * config/v850/v850.c (v850_legitimate_address_p): Handle large
11447         displacements for TARGET_V850E2V3 and newer.
11448         (TARGET_LRA_P): Remove.  Defaults to LRA now.
11449         * config/v850/v850.md (sign23byte_load): Remove.
11450         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11451         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11453 2018-06-29  Martin Liska  <mliska@suse.cz>
11455         PR lto/85759
11456         * coverage.c (coverage_init): Mangle full path name.
11457         * doc/invoke.texi: Document the change.
11458         * gcov-io.c (mangle_path): New.
11459         * gcov-io.h (mangle_path): Likewise.
11460         * gcov.c (mangle_name): Use mangle_path for path mangling.
11462 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11464         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11465         if starting source register is not even.
11467 2018-06-29  Martin Liska  <mliska@suse.cz>
11469         PR tree-optimization/86263
11470         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11471         Make edge redirection.
11473 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
11475         * dumpfile.c (dump_loc): Add indentation based on scope depth.
11476         (dump_scope_depth): New variable.
11477         (get_dump_scope_depth): New function.
11478         (dump_begin_scope): New function.
11479         (dump_end_scope): New function.
11480         * dumpfile.h (get_dump_scope_depth): New declaration.
11481         (dump_begin_scope): New declaration.
11482         (dump_end_scope): New declaration.
11483         (class auto_dump_scope): New class.
11484         (AUTO_DUMP_SCOPE): New macro.
11485         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11486         AUTO_DUMP_SCOPE.
11488 2018-06-29  Richard Biener  <rguenther@suse.de>
11490         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11491         compute_all_dependences succeeds.
11492         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11493         exceed --param loop-max-datarefs-for-datadeps.
11495 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
11497         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11499 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
11501         PR target/86348
11502         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11503         alternative 0 in preferred_for_speed attribute.
11505 2018-06-28  Paul Koning  <ni1d@arrl.net>
11507         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
11508         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
11509         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
11510         * config/pdp11/pdp11.md: Correct "length" attribute calculation
11511         for shift insn patterns.
11513 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
11515         * cgraph.c (cgraph_node::get_body): Replace assignments to
11516         "dump_file" with calls to set_dump_file.
11517         * dumpfile.c (alt_dump_file): Make static, and group with...
11518         (alt_flags): ...this definition.
11519         (dumps_are_enabled): New variable.
11520         (refresh_dumps_are_enabled): New function.
11521         (set_dump_file): New function.
11522         (set_alt_dump_file): New function.
11523         (gcc::dump_manager::dump_start): Replace assignments to
11524         "dump_file" and "alt_dump_file" with calls to set_dump_file and
11525         set_alt_dump_file.
11526         (gcc::dump_manager::dump_finish): Likewise.
11527         * dumpfile.h (alt_dump_file): Delete decl.
11528         (dumps_are_enabled): New variable decl.
11529         (set_dump_file): New function decl.
11530         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
11531         global.
11532         * tree-nested.c (lower_nested_functions): Replace assignments to
11533         "dump_file" with calls to set_dump_file.
11535 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11537         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
11538         goto_locus of each outgoing edge of each basic block.
11540 2018-06-28  Richard Biener  <rguenther@suse.de>
11542         * dwarf2out.c (decl_scope_table): Remove.
11543         (push_decl_scope): Likewise.
11544         (pop_decl_scope): Likewise.
11545         (gen_type_die_for_member): Do not call push/pop_decl_scope.
11546         (gen_struct_or_union_type_die): Likewise.
11547         (gen_tagged_type_die): Likewise.
11548         (dwarf2out_init): Do not initialize decl_scope_table.
11549         (dwarf2out_c_finalize): Do not free it.
11551 2018-06-28  Richard Biener  <rguenther@suse.de>
11553         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
11554         deciding whether to not re-use a DIE.
11556 2018-06-28  Richard Biener  <rguenther@suse.de>
11558         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
11559         DW_AT_abstract_origin attribute.
11561 2018-06-28  Martin Liska  <mliska@suse.cz>
11563         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11564         Use newly introduced constants.
11565         * tree-switch-conversion.h (struct jump_table_cluster):
11566         Define max_ratio_for_size and max_ratio_for_speed.
11568 2018-06-28  Martin Liska  <mliska@suse.cz>
11570         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11571         Add new checking assert to catch invalid state.
11572         (jump_table_cluster::can_be_handled): Handle single case
11573         clusters.
11574         (jump_table_cluster::is_beneficial): Bail out for such case.
11575         (bit_test_cluster::find_bit_tests):
11576         Add new checking assert to catch invalid state.
11577         (bit_test_cluster::can_be_handled): Handle single case
11578         clusters.
11579         (bit_test_cluster::is_beneficial): Bail out for such case.
11580         (switch_decision_tree::analyze_switch_statement):
11581         Fix comment.
11583 2018-06-28  Martin Liska  <mliska@suse.cz>
11585         * common.opt: Introduce -completion option.
11586         * gcc.c (driver_handle_option): Handle it.
11587         (driver::main): Print completions if completion
11588         is set.
11589         * opt-suggestions.c (option_proposer::get_completions):
11590         New function.
11591         (option_proposer::suggest_completion): Likewise.
11592         (option_proposer::find_param_completions): Likewise.
11593         (verify_autocompletions): Likewise.
11594         (test_completion_valid_options): Likewise.
11595         (test_completion_valid_params): Likewise.
11596         (in_completion_p): Likewise.
11597         (empty_completion_p): Likewise.
11598         (test_completion_partial_match): Likewise.
11599         (test_completion_garbage): Likewise.
11600         (opt_proposer_c_tests): Likewise.
11601         * opt-suggestions.h: Declare new functions.
11602         * opts.c (common_handle_option): Handle OPT__completion_.
11603         * selftest-run-tests.c (selftest::run_tests): Add
11604         opt_proposer_c_tests.
11605         * selftest.c (assert_str_startswith): New.
11606         * selftest.h (assert_str_startswith): Likewise.
11607         (opt_proposer_c_tests): New.
11608         (ASSERT_STR_STARTSWITH): Likewise.
11610 2018-06-28  Martin Liska  <mliska@suse.cz>
11612         * Makefile.in: Add opt-suggestions.o.
11613         * gcc-main.c: Include opt-suggestions.h.
11614         * gcc.c (driver::driver): Likewise.
11615         (driver::~driver): Remove m_option_suggestions.
11616         (driver::build_option_suggestions): Moved to option_proposer.
11617         (driver::suggest_option): Likewise.
11618         (driver::handle_unrecognized_options): Use option_proposer.
11619         * gcc.h (class driver): Add new memver m_option_proposer.
11620         * opt-suggestions.c: New file.
11621         * opt-suggestions.h: New file.
11623 2018-06-28  Martin Liska  <mliska@suse.cz>
11625         * vec.h (class auto_string_vec): New (moved from auto_argvec).
11626         (auto_string_vec::~auto_string_vec): Likewise.
11628 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11630         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
11631         prevent_decl_creation_for_types fields up and add reset_location field.
11632         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
11633         statement if id->reset_location is true.
11634         (copy_edges_for_bb): Do not set goto_locus on the new edges if
11635         id->reset_location is true.
11636         (copy_phis_for_bb): Force input_location on the arguments if
11637         id->reset_location is true.
11638         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
11639         is set on the function to be inlined.
11641 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
11643         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
11645 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
11647         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
11648         registers for Pmode.
11649         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
11650         hard registers for the clobbered pseudo.
11652 2018-06-27  Paul Koning  <ni1d@arrl.net>
11654         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11655         mutually exclusive options.
11656         * config/pdp11/constraints.md (h): New constraint.
11657         (O): Update definition to match shift code generation.
11658         (D): New constraint.
11659         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
11660         (CCFP): Remove.
11661         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
11662         function.
11663         (output_jump): Change arguments.
11664         (pdp11_fixed_cc_regs): New function.
11665         (pdp11_cc_mode): Ditto.
11666         (pdp11_expand_shift): Ditto.
11667         (pdp11_assemble_shift): Ditto.
11668         (pdp11_small_shift): Ditto.
11669         (pdp11_branch_cost): Remove.
11670         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
11671         from output.
11672         (pdp11_register_move_cost): Update for CC registers.
11673         (pdp11_rtx_costs): Add case for LSHIFTRT.
11674         (pdp11_output_jump): Add CCNZ mode conditional branches.
11675         (notice_update_cc_on_set): Remove.
11676         (pdp11_cc_mode): New function.
11677         (simple_memory_operand): Correct pre/post decrement case.
11678         (no_side_effect_operand): New function.
11679         (pdp11_regno_reg_class): Add CC_REGS class.
11680         (pdp11_fixed_cc_regs): New function.
11681         (pdp11_small_shift): New function.
11682         (pdp11_expand_shift): New function to expand shift insns.
11683         (pdp11_assemble_shift): New function to output shifts.
11684         (pdp11_branch_cost): Remove.
11685         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
11686         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
11687         (WCHAR_TYPE): Ditto.
11688         (PTRDIFF_TYPE): Ditto.
11689         (ADJUST_INSN_LENGTH): New macro.
11690         (FIXED_REGISTERS): Add CC registers.
11691         (CALL_USED_REGISTERS): Ditto.
11692         (reg_class): Ditto.
11693         (REG_CLASS_NAMES): Ditto.
11694         (REG_CLASS_CONTENTS): Ditto.
11695         (SELECT_CC_MODE): Use new function.
11696         (TARGET_FLAGS_REGNUM): New macro.
11697         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
11698         (cc0_reg_rtx): Remove.
11699         (CC_STATUS_MDEP): Remove.
11700         (CC_STATUS_MDEFP_INIT): Remove.
11701         (CC_IN_FPU): Remove.
11702         (NOTICE_UPDATE_CC): Remove.
11703         (REGISTER_NAMES): Add CC registers.
11704         (BRANCH_COST): Change to constant 1.
11705         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
11706         handling.
11707         * config/pdp11/pdp11.opt (mbcopy): Remove.
11708         (mbcopy-builtin): Remove.
11709         (mbranch-cheap): Remove.
11710         (mbranch-expensive): Remove.
11711         * config/pdp11/predicates.md (expand_shift_operand): Update to
11712         match shift code generation.
11713         (ccnz_operator): New predicate.
11714         * doc/invoke.texi (PDP-11 Options): Remove deleted options
11715         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
11716         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
11717         exclusive options.
11718         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
11719         description of O constraint.
11721 2018-06-27  Jeff Law  <law@redhat.com>
11722             Austin Law  <austinklaw@gmail.com>
11724         * config/v850/v850.md (addsi3_set_flags): New pattern.
11725         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
11726         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
11727         (zero_extendhisi2_v850_set_flags): Likewise.
11728         (zero_extendqisi2_v850_set_flags): Likewise.
11729         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
11730         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
11731         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
11733         * config/v850/v850-protos.h (notice_update_cc): Remove.
11734         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
11735         (v850_print_operand): Handle 'D' and "d".
11736         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
11737         Add handling of arithmetic/logical operations compared against zero.
11738         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
11739         Do not look at v850_compare_op, instead get mode from last argument.
11740         (v850_gen_compare): Remove
11741         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
11742         after reload for prologue insns.
11743         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
11744         patterns.
11745         (construct_save_jarl): Likewise.
11746         (TARGET_FLAGS_REGNUM): Define.
11747         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
11748         (NOTICE_UPDATE_CC): Remove.
11749         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
11750         than cc0.  Conditionalize on reload_completed.
11751         (cmpsi_insn, setfcc_insn): Likewise.
11752         (tst1 splitter): Turn into define_and_split which sets the flags
11753         after reload.
11754         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
11755         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
11756         (cstoresf4, cstoredf4): Clobber the flags.
11757         (cmpsi, cmpsf, cmpdf): Remove expanders.
11758         (setf_insn): Remove pattern.
11759         (addsi3): Turn into define_and_split which clobbers the flags after
11760         reload and a suitable pattern (addsi3_clobber_flags) for use after
11761         reload.
11762         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
11763         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
11764         (ashrsi3, ashrsi3_v850e2): Likewise.
11765         (bins): Clobber the flags.
11766         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
11767         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
11768         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
11769         (call_value_internal_short, call_value_internal_long): Likewise.
11770         (callt_save_interrupt, callt_return_interrupt): Likewise.
11771         (save_interrupt, return_interrupt): Likewise.
11772         (callt_save_all_interrupt, save_all_interrupt): Likewise.
11773         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
11774         (restore_all_interrupt, _restore_all_interrupt): Likewise.
11775         (All FP comparisons): Only allow after reload has completed.
11776         (trfsr): Likewise.
11777         (divh, divhu): Tweak output template.
11778         (branch_z_normal, branch_z_invert): Remove
11779         (branch_nz_normal, branch_nz_invert): Likewise.
11780         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
11782         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
11783         * config/v850/v850.c (notice_update_cc): Remove.
11784         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
11785         (CC_NO_CARRY): Likewise.
11786         (NOTICE_UPDATE_CC): Define to nothing.
11787         * config/v850/v850.md: Remove block comment on cc0 handling
11788         Remove "cc" attribute from all patterns.  Remove cc_status handling
11789         from all patterns.  Minor formatting fixes.
11791 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11793         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11794         (cortex-a76.cortex-a55): Likewise.
11795         * config/aarch64/aarch64-tune.md: Regenerate.
11796         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11797         cortex-a76.cortex-a55.
11799 2018-06-27  Jeff Law  <law@redhat.com>
11801         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11802         (MULTILIB_DIRNAMES): Similarly.
11804 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
11806         * gimple.h (gimple_return_retbnd): Delete.
11807         (gimple_return_set_retbnd): Likewise.
11808         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11809         gimple_return_set_retbnd.
11810         * gimple-pretty-print.c (dump_gimple_return): Remove call to
11811         gimple_return_retbnd and adjust.
11812         * tree-inline.h (struct copy_body_data): Remove retbnd field.
11813         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11814         Explicitly return NULL in a couple more cases.  Move assertion
11815         on debug statements and remove unreachable code.
11816         (reset_debug_binding): Do not test id->retbnd.
11817         (expand_call_inline): Do not set it.
11819 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11821         * configure.ac: Add --disable-gcov option.
11822         * configure: Regenerate.
11823         * Makefile.in: Honour @enable_gcov@.
11824         * doc/install.texi: Document --disable-gcov.
11826 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11828         * config/arm/arm-cpus.in (cortex-a76): New entry.
11829         (cortex-a76.cortex-a55): Likewise.
11830         * config/arm/arm-tables.opt: Regenerate.
11831         * config/arm/arm-tune.md: Likewise.
11832         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11833         * doc/invoke.texi (ARM Options): Document cortex-a76 and
11834         cortex-a76.cortex-a55.
11836 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
11838         PR target/85769
11839         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11841 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11843         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11844         comment.
11845         (EPILOGUE_USES): Likewise.
11847 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
11849         * tree-inline.c (remap_location): New function extracted from...
11850         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
11851         (copy_phis_for_bb): ...here.  Call remap_location.
11852         (copy_cfg_body): Adjust call to copy_edges_for_bb.
11854 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
11856         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11857         unaligned vsx for 16B memset.
11859 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
11861         PR target/86285
11862         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11863         ieee128_float_type_node to long_double_type_node unless
11864         TARGET_LONG_DOUBLE_128 is set.
11866 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
11868         * cfgloop.c (get_loop_location): Convert return type from
11869         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11870         by implicit construction from rtx_insn *, and using
11871         dump_user_location_t::from_function_decl for the fallback case.
11872         * cfgloop.h (get_loop_location): Convert return type from
11873         location_t to dump_user_location_t.
11874         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11875         dump_printf_loc to pass in a dump_location_t rather than a
11876         location_t, via the gimple stmt.
11877         * coverage.c (get_coverage_counts): Update calls to
11878         dump_printf_loc to pass in dump_location_t rather than a
11879         location_t.
11880         * doc/optinfo.texi (Dump types): Convert example of
11881         dump_printf_loc from taking "locus" to taking "insn".  Update
11882         description of the "_loc" calls to cover dump_location_t.
11883         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11884         "selftest.h".
11885         (dump_user_location_t::dump_user_location_t): New constructors,
11886         from gimple *stmt and rtx_insn *.
11887         (dump_user_location_t::from_function_decl): New function.
11888         (dump_loc): Make static.
11889         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11890         const dump_location_t &.
11891         (dump_generic_expr_loc): Delete.
11892         (dump_printf_loc): Convert param "loc" from location_t to
11893         const dump_location_t &.
11894         (selftest::test_impl_location): New function.
11895         (selftest::dumpfile_c_tests): New function.
11896         * dumpfile.h: Include "profile-count.h".
11897         (class dump_user_location_t): New class.
11898         (struct dump_impl_location_t): New struct.
11899         (class dump_location_t): New class.
11900         (dump_printf_loc): Convert 2nd param from source_location to
11901         const dump_location_t &.
11902         (dump_generic_expr_loc): Delete.
11903         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11904         const dump_location_t &.
11905         * gimple-fold.c (fold_gimple_assign): Update call to
11906         dump_printf_loc to pass in a dump_location_t rather than a
11907         location_t, via the gimple stmt.
11908         (gimple_fold_call): Likewise.
11909         * gimple-loop-interchange.cc
11910         (loop_cand::analyze_iloop_reduction_var): Update for change to
11911         check_reduction_path.
11912         (tree_loop_interchange::interchange): Update for change to
11913         find_loop_location.
11914         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11915         change in return-type of find_loop_location.
11916         (graphite_regenerate_ast_isl): Likewise.
11917         * graphite-optimize-isl.c (optimize_isl): Likewise.
11918         * graphite.c (graphite_transform_loops): Likewise.
11919         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11920         pass in a dump_location_t rather than a location_t, via the
11921         gimple stmt.
11922         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11923         * ipa.c (walk_polymorphic_call_targets): Likewise.
11924         * loop-unroll.c (report_unroll): Convert "locus" param from
11925         location_t to dump_location_t.
11926         (decide_unrolling): Update for change to get_loop_location's
11927         return type.
11928         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11929         location_t to dump_user_location_t.
11930         (grid_find_single_omp_among_assignments_1): Updates calls to
11931         dump_printf_loc to pass in a dump_location_t rather than a
11932         location_t, via the gimple stmt.
11933         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11934         location_t to dump_location_t.  Updates calls to dump_printf_loc
11935         to pass in a dump_location_t rather than a location_t, via the
11936         gimple stmt.
11937         (grid_inner_loop_gridifiable_p): Likewise.
11938         (grid_dist_follows_simple_pattern): Likewise.
11939         (grid_gfor_follows_tiling_pattern): Likewise.
11940         (grid_target_follows_gridifiable_pattern): Likewise.
11941         (grid_attempt_target_gridification): Convert initialization
11942         of local "grid" from memset to zero-initialization; FIXME: does
11943         this require C++11?  Update call to dump_printf_loc to pass in a
11944         optinfo_location rather than a location_t, via the gimple stmt.
11945         * profile.c (read_profile_edge_counts): Updates call to
11946         dump_printf_loc to pass in a dump_location_t rather than a
11947         location_t
11948         (compute_branch_probabilities): Likewise.
11949         * selftest-run-tests.c (selftest::run_tests): Call
11950         dumpfile_c_tests.
11951         * selftest.h (dumpfile_c_tests): New decl.
11952         * tree-loop-distribution.c (pass_loop_distribution::execute):
11953         Update for change in return type of find_loop_location.
11954         * tree-parloops.c (parallelize_loops): Likewise.
11955         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11956         "locus" from location_t to dump_user_location_t.
11957         (canonicalize_loop_induction_variables): Likewise.
11958         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11959         for change in return type of find_loop_location.
11960         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11961         to dump_printf_loc to pass in a dump_location_t rather than a
11962         location_t, via the stmt.
11963         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11964         Likewise.
11965         * tree-vect-loop-manip.c (find_loop_location): Convert return
11966         type from source_location to dump_user_location_t.
11967         (vect_do_peeling): Update for above change.
11968         (vect_loop_versioning): Update for change in type of
11969         vect_location.
11970         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11971         from location_t to dump_user_location_t.
11972         (vect_estimate_min_profitable_iters): Update for change in type
11973         of vect_location.
11974         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11975         location_t to dump_location_t.
11976         (vect_slp_bb): Update for change in type of vect_location.
11977         * tree-vectorizer.c (vect_location): Convert from source_location
11978         to dump_user_location_t.
11979         (try_vectorize_loop_1): Update for change in vect_location's type.
11980         (vectorize_loops): Likewise.
11981         (increase_alignment): Likewise.
11982         * tree-vectorizer.h (vect_location): Convert from source_location
11983         to dump_user_location_t.
11984         (find_loop_location): Convert return type from source_location to
11985         dump_user_location_t.
11986         (check_reduction_path): Convert 1st param from location_t to
11987         dump_user_location_t.
11988         * value-prof.c (check_counter): Update call to dump_printf_loc to
11989         pass in a dump_user_location_t rather than a location_t; update
11990         call to error_at for change in type of "locus".
11991         (check_ic_target): Update call to dump_printf_loc to
11992         pass in a dump_user_location_t rather than a location_t, via the
11993         call_stmt.
11995 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11997         * config/s390/s390.h (enum processor_flags): Do not use
11998         default tune parameter when -march was specified.
12000 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12002         PR target/86314
12003         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12004         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12005         operands.
12007 2018-06-26  Richard Biener  <rguenther@suse.de>
12009         PR tree-optimization/86287
12010         PR bootstrap/86316
12011         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12012         (vect_analyze_loop): Initialize n_stmts.
12014 2018-06-26  Richard Biener  <rguenther@suse.de>
12016         PR middle-end/86271
12017         * fold-const.c (fold_convertible_p): Pointer extension
12018         isn't valid.
12020 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12022         PR debug/86064
12023         * dwarf2out.c (loc_list_has_views): Adjust comments.
12024         (dw_loc_list): Split single cross-partition range with
12025         nonzero locview.
12027 2018-06-25  Jeff Law  <law@redhat.com>
12029         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12030         on -mbig-switch by default.
12032         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12033         test.
12034         (const_float_0_operand): Remove unused predicate.
12035         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12036         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12037         (recipsf2): New expander.  Original pattern now called
12038         (recipsf2_insn).
12039         (recipdf2, recipdf2_insn): Similarly.
12040         (rsqrtsf2, rsqrtsf2_insn): Similarly
12041         (rsqrtdf2, rsqrtdf2_insn): Similarly
12043 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12045         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12046         Simplify logic for FreeBSD (twice).
12048 2018-06-25  Martin Sebor  <msebor@redhat.com>
12050         PR tree-optimization/86204
12051         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12052         a strnlen result if it's less than the length of the string.
12054 2018-06-25  Martin Sebor  <msebor@redhat.com>
12056         PR tree-optimization/85700
12057         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12058         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12059         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12061 2018-06-25  Martin Sebor  <msebor@redhat.com>
12063         * doc/extend.texi (Zero-length arrays): Update and clarify.
12065 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12067         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12068         added IEEE/IBM long double multilib support on PowerPC little
12069         endian Linux systems.
12070         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12071         (MULTILIB_DEFAULTS): Likewise.
12072         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12073         Likewise.
12074         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12075         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12076         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12078 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12080         PR middle-end/86311
12081         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12082         (REORDER_45): Likewise.
12084 2018-06-25  Jeff Law  <law@redhat.com>
12086         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12087         dividend to 32 bits.  Adjust length.
12088         (udivmodhi4): Cleanup output template.  Fix length.
12090 2018-06-25  Carl Love  <cel@us.ibm.com>
12092         * config/rs6000/vsx.md: Change word selector to prefered location.
12094 2018-06-25  Richard Biener  <rguenther@suse.de>
12096         PR tree-optimization/86304
12097         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12098         epilogue-if-converted loops as well.
12100 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12102         * lto-section-out.c (lto_begin_section): Do not print section
12103         name for noaddr and unnumbered dumps.
12105 2018-06-25  Richard Biener  <rguenther@suse.de>
12107         * tree-vectorizer.h (struct vec_info_shared): New structure
12108         with parts split out from struct vec_info and loop_nest from
12109         struct _loop_vec_info.
12110         (struct vec_info): Adjust accordingly.
12111         (struct _loop_vec_info): Likewise.
12112         (LOOP_VINFO_LOOP_NEST): Adjust.
12113         (LOOP_VINFO_DATAREFS): Likewise.
12114         (LOOP_VINFO_DDRS): Likewise.
12115         (struct _bb_vec_info): Likewise.
12116         (BB_VINFO_DATAREFS): Likewise.
12117         (BB_VINFO_DDRS): Likewise.
12118         (struct _stmt_vec_info): Add dr_aux member.
12119         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12120         (DR_MISALIGNMENT_UNINITIALIZED): New.
12121         (set_dr_misalignment): Adjust.
12122         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12123         (vect_analyze_loop): Adjust prototype.
12124         (vect_analyze_loop_form): Likewise.
12125         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12126         Compute dependences lazily.
12127         (vect_record_base_alignments): Use shared datarefs/ddrs.
12128         (vect_verify_datarefs_alignment): Likewise.
12129         (vect_analyze_data_refs_alignment): Likewise.
12130         (vect_analyze_data_ref_accesses): Likewise.
12131         (vect_analyze_data_refs): Likewise.
12132         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12133         constructor parameter for shared part.
12134         (vect_analyze_loop_form): Pass in shared part and adjust.
12135         (vect_analyze_loop_2): Pass in storage for the number of
12136         stmts.  Move loop nest finding to the caller.  Compute
12137         datarefs lazily.
12138         (vect_analyze_loop): Pass in shared part.
12139         (vect_transform_loop): Verify shared datarefs are unchanged.
12140         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12141         constructor parameter for shared part.
12142         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12143         (vect_slp_bb): Verify shared datarefs are unchanged before
12144         transform.
12145         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12146         change.
12147         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12148         DR_MISALIGNMENT_UNINITIALIZED.
12149         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12150         parameter for shared part.
12151         (vec_info::~vec_info): Adjust.
12152         (vec_info_shared::vec_info_shared): New.
12153         (vec_info_shared::~vec_info_shared): Likewise.
12154         (vec_info_shared::save_datarefs): Likewise.
12155         (vec_info_shared::check_datarefs): Likewise.
12156         (try_vectorize_loop_1): Construct shared part live for analyses
12157         of a single loop for multiple vector sizes.
12158         * tree-parloops.c (gather_scalar_reductions): Adjust.
12160 2018-06-25  Richard Biener  <rguenther@suse.de>
12162         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12163         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12164         (vect_analyze_data_refs): Remove similar code from here and
12165         simplify accordingly.
12167 2018-06-25  Richard Biener  <rguenther@suse.de>
12169         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12170         for reverse storage order accesses rather than asserting
12171         they cannot happen here.
12173 2018-06-25  Tom de Vries  <tdevries@suse.de>
12175         PR debug/86257
12176         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12177         Use data16 instead of .byte for insn prefix.
12179 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12181         PR C++/86082
12182         * parser.c (make_char_string_pack): Pass this literal chars
12183         through cpp_interpret_string.
12184         (cp_parser_userdef_numeric_literal): Check the result of
12185         make_char_string_pack.
12187 2018-06-24  Maya Rashish  <coypu@sdf.org>
12189         * ginclude/stddef.h: Simplify conditions around avoiding
12190         re-definition of __size_t.
12192 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12194         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12195         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12197 2018-06-22  Maya Rashish  <coypu@sdf.org>
12199         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12200         made to OpenBSD and NetBSD through the years.
12202 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12204         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12205         behavior of vec_pack (vector double, vector double) to match
12206         behavior of vec_float2 (vector double, vector double).
12208 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12210         * gimplify.c (gimplify_function_tree): Prevent creation
12211         of a trampoline for the address of the current function
12212         passed to entry/exit instrumentation hooks.
12214 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12216         PR target/86222
12217         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12218         correctly.
12220 2018-06-22  Martin Liska  <mliska@suse.cz>
12222         PR tree-optimization/86263
12223         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12224         Bail out if is_enabled is false.
12225         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12226         New declaration.
12227         (jump_table_cluster::is_enabled): New function.
12229 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12231         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12232         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12233         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12234         (lto_input_ts_binfo_tree_pointers): Likewise.
12235         * tree-streamer-out.c (streamer_write_tree_bitfields,
12236         write_ts_binfo_tree_pointers): Likewise.
12237         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12239 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12241         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12243 2018-06-22  Martin Liska  <mliska@suse.cz>
12245         * symbol-summary.h (get): Make it pure and inline move
12246         functionality from ::get function.
12247         (get): Remove and inline into ::get and ::get_create.
12248         (get_create): Move code from ::get function.
12250 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12252         PR target/85994
12253         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12254         -x assembler-with-cpp.
12256 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12258         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12259         _FILE_OFFSET_BITS=64 for C++.
12261 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12263         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12264         conversion insn that shows up when pr85657-3.c is compiled using
12265         IEEE 128-bit long double.
12266         (neg<mode>2_internal): Use the correct mode to check whether the
12267         mode is IBM extended.
12268         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12269         multiply and divide external functions from being created more
12270         than once.
12272 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12274         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12275         functions.
12276         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12277         the edge can be forwarded.
12278         (cfg_layout_merge_blocks): Likewise.
12280 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12282         * except.c (finish_eh_generation): Commit edge insertions only after
12283         the EH edges have been redirected from post-landing to landing pads.
12285 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12287         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12288         create_tmp_var_for to create the FRAME decl.
12289         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12291 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12293         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12294         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12295         debug statement when resetting its value.
12296         (expand_call_inline): Copy the locus of the call onto the assignment
12297         of the return value, if any.  Use local variable in more cases.
12299 2018-06-21  Martin Liska  <mliska@suse.cz>
12301         * ipa-pure-const.c (propagate_nothrow): Use
12302         funct_state_summaries->get.
12303         (dump_malloc_lattice): Likewise.
12304         (propagate_malloc): Likewise.
12306 2018-06-21  Richard Biener  <rguenther@suse.de>
12308         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12309         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12310         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12311         comment.
12312         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12313         BLOCK_ABSTRACT_ORIGIN unconditionally.
12315 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12317         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12318         deleting it.
12319         * ipa-reference.c (ipa_reference_c_finalize): Delete
12320         ipa_ref_opt_sum_summaries and set it to NULL.
12322 2018-06-21  Tom de Vries  <tdevries@suse.de>
12324         PR tree-optimization/85859
12325         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12326         test with comment from bb_no_side_effects_p.
12328 2018-06-21  Richard Biener  <rguenther@suse.de>
12330         PR tree-optimization/86232
12331         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12332         max for constant niter.
12334 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12336         * config/aarch64/aarch64-simd.md
12337         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12339 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12341         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12342         Make opernads of the unspec commutative.
12344 2018-06-21  Richard Biener  <rguenther@suse.de>
12346         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12347         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12348         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12349         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12350         NULL DR_STEP.
12351         (vect_record_base_alignments): Do not record base alignment
12352         for gathers or scatters.
12353         (vect_compute_data_ref_alignment): Drop return value that is always
12354         true.  Bail out early for gathers or scatters.
12355         (vect_enhance_data_refs_alignment): Bail out early for gathers
12356         or scatters.
12357         (vect_find_same_alignment_drs): Likewise.
12358         (vect_analyze_data_refs_alignment): Remove dead code.
12359         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12360         (vect_analyze_data_refs): For possible gathers or scatters do
12361         not create an alternate DR, just check their possible validity
12362         and mark them.  Adjust DECL_NONALIASED handling to not rely
12363         on DR_BASE_ADDRESS.
12364         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12365         update inits of gathers or scatters.
12366         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12367         Also copy gather/scatter flag to pattern vinfo.
12369 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12371         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12372         behavior of vec_packsu (vector unsigned long long, vector unsigned
12373         long long) to match behavior of vec_packs with same signature.
12375 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12376             Thomas Schwinge <thomas@codesourcery.com>
12377             Cesar Philippidis  <cesar@codesourcery.com>
12379         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12380         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12381         (gimplify_adjust_omp_clauses): Likewise.
12382         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12383         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12384         (gimplify_omp_target_update): Update handling of acc update and
12385         enter/exit data.
12386         * omp-low.c (install_var_field): Remove unused parameter
12387         base_pointers_restrict.
12388         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12389         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12390         FINALIZE}
12391         (omp_target_base_pointers_restrict_p): Delete.
12392         (scan_omp_target): Update call to scan_sharing_clauses.
12393         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12394         FINALIZE}.
12395         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12396         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12397         (convert_local_omp_clauses): Likewise.
12398         * tree-pretty-print.c (dump_omp_clause): Likewise.
12399         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
12400         FINALIZE}.
12401         (omp_clause_code_name): Likewise.
12403 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12405         PR debug/86194
12406         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12407         be narrowed.
12409         PR tree-optimization/86231
12410         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
12411         anti-range don't overwrite *vr0min before using it to compute *vr0max.
12413 2018-06-20  Tom de Vries  <tdevries@suse.de>
12415         PR tree-optimization/86097
12416         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12417         iv type if signedness of iv type is not the same as that of *nit.
12419 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12421         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
12422         EDGE_EH edges, verify they are all EDGE_EH.
12424 2018-06-20  Maya Rashish  <coypu@sdf.org>
12426         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12428 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12430         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12431         * config/aarch64/aarch64.c (xgene1_tunings): Add
12432         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12433         (aarch64_mode_valid_for_sched_fusion_p):
12434         Allow 16-byte modes.
12435         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12436         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12437         128-bit modes.
12438         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12439         New pattern.
12440         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12441         * config/aarch64/iterators.md (VQ2): New mode iterator.
12443 2018-06-20  Martin Liska  <mliska@suse.cz>
12445         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12446         Change default ratio from 10 to 8.
12448 2018-06-20  Martin Liska  <mliska@suse.cz>
12450         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12451         New.
12452         (bit_test_cluster::find_bit_tests): Likewise.
12453         (switch_decision_tree::analyze_switch_statement): Find clusters.
12454         * tree-switch-conversion.h (struct jump_table_cluster): Document
12455         hierarchy.
12457 2018-06-20  Martin Liska  <mliska@suse.cz>
12459         * tree-switch-conversion.c (switch_conversion::collect):
12460         Record m_uniq property.
12461         (switch_conversion::expand): Bail out for special conditions.
12462         (group_cluster::~group_cluster): New.
12463         (group_cluster::group_cluster): Likewise.
12464         (group_cluster::dump): Likewise.
12465         (jump_table_cluster::emit): New.
12466         (switch_decision_tree::fix_phi_operands_for_edges): New.
12467         (struct case_node): Remove struct.
12468         (jump_table_cluster::can_be_handled): New.
12469         (case_values_threshold): Moved to header.
12470         (reset_out_edges_aux): Likewise.
12471         (jump_table_cluster::is_beneficial): New.
12472         (bit_test_cluster::can_be_handled): Likewise.
12473         (add_case_node): Remove.
12474         (bit_test_cluster::is_beneficial): New.
12475         (case_bit_test::cmp): New.
12476         (bit_test_cluster::emit): New.
12477         (expand_switch_as_decision_tree_p): Remove.
12478         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12479         (fix_phi_operands_for_edge): Likewise.
12480         (switch_decision_tree::analyze_switch_statement): New.
12481         (compute_cases_per_edge): Move ...
12482         (switch_decision_tree::compute_cases_per_edge): ... here.
12483         (try_switch_expansion): Likewise.
12484         (switch_decision_tree::try_switch_expansion): Likewise.
12485         (record_phi_operand_mapping): Likewise.
12486         (switch_decision_tree::record_phi_operand_mapping): Likewise.
12487         (emit_case_decision_tree): Likewise.
12488         (switch_decision_tree::emit): Likewise.
12489         (balance_case_nodes): Likewise.
12490         (switch_decision_tree::balance_case_nodes): Likewise.
12491         (dump_case_nodes): Likewise.
12492         (switch_decision_tree::dump_case_nodes): Likewise.
12493         (emit_jump): Likewise.
12494         (switch_decision_tree::emit_jump): Likewise.
12495         (emit_cmp_and_jump_insns): Likewise.
12496         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12497         (emit_case_nodes): Likewise.
12498         (switch_decision_tree::emit_case_nodes): Likewise.
12499         (conditional_probability): Remove.
12500         * tree-switch-conversion.h (enum cluster_type): New.
12501         (PRINT_CASE): New.
12502         (struct cluster): Likewise.
12503         (cluster::cluster): Likewise.
12504         (struct simple_cluster): Likewise.
12505         (simple_cluster::simple_cluster): Likewise.
12506         (struct group_cluster): Likewise.
12507         (struct jump_table_cluster): Likewise.
12508         (struct bit_test_cluster): Likewise.
12509         (struct min_cluster_item): Likewise.
12510         (struct case_tree_node): Likewise.
12511         (case_tree_node::case_tree_node): Likewise.
12512         (jump_table_cluster::case_values_threshold): Likewise.
12513         (struct case_bit_test): Likewise.
12514         (struct switch_decision_tree): Likewise.
12515         (struct switch_conversion): Likewise.
12516         (switch_decision_tree::reset_out_edges_aux): Likewise.
12518 2018-06-20  Martin Liska  <mliska@suse.cz>
12520         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
12521         (hoist_edge_and_branch_if_true): Likewise.
12522         (expand_switch_using_bit_tests_p): Likewise.
12523         (struct case_bit_test): Likewise.
12524         (case_bit_test_cmp): Likewise.
12525         (emit_case_bit_tests): Likewise.
12526         (switch_conversion::switch_conversion): New class.
12527         (struct switch_conv_info): Remove old struct.
12528         (collect_switch_conv_info): More to ...
12529         (switch_conversion::collect): ... this.
12530         (check_range): Likewise.
12531         (switch_conversion::check_range): Likewise.
12532         (check_all_empty_except_final): Likewise.
12533         (switch_conversion::check_all_empty_except_final): Likewise.
12534         (check_final_bb): Likewise.
12535         (switch_conversion::check_final_bb): Likewise.
12536         (create_temp_arrays): Likewise.
12537         (switch_conversion::create_temp_arrays): Likewise.
12538         (free_temp_arrays): Likewise.
12539         (gather_default_values): Likewise.
12540         (switch_conversion::gather_default_values): Likewise.
12541         (build_constructors): Likewise.
12542         (switch_conversion::build_constructors): Likewise.
12543         (constructor_contains_same_values_p): Likewise.
12544         (switch_conversion::contains_same_values_p): Likewise.
12545         (array_value_type): Likewise.
12546         (switch_conversion::array_value_type): Likewise.
12547         (build_one_array): Likewise.
12548         (switch_conversion::build_one_array): Likewise.
12549         (build_arrays): Likewise.
12550         (switch_conversion::build_arrays): Likewise.
12551         (gen_def_assigns): Likewise.
12552         (switch_conversion::gen_def_assigns): Likewise.
12553         (prune_bbs): Likewise.
12554         (switch_conversion::prune_bbs): Likewise.
12555         (fix_phi_nodes): Likewise.
12556         (switch_conversion::fix_phi_nodes): Likewise.
12557         (gen_inbound_check): Likewise.
12558         (switch_conversion::gen_inbound_check): Likewise.
12559         (process_switch): Use the newly created class.
12560         (switch_conversion::expand): New.
12561         (switch_conversion::~switch_conversion): New.
12562         * tree-switch-conversion.h: New file.
12564 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12566         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
12567         tree-vect-patterns.c.
12568         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
12569         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
12570         (vect_recog_sad_pattern): Likewise.
12571         (vect_recog_widen_sum_pattern): Likewise.
12572         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
12573         (vect_recog_widen_shift_pattern): Remove the type_in argument.
12574         (vect_recog_rotate_pattern): Likewise.
12575         (vect_recog_mult_pattern): Likewise.
12576         (vect_recog_vector_vector_shift_pattern): Likewise.
12577         (vect_recog_divmod_pattern): Likewise.
12578         (vect_recog_mixed_size_cond_pattern): Likewise.
12579         (vect_recog_bool_pattern): Likewise.
12580         (vect_recog_mask_conversion_pattern): Likewise.
12581         (vect_try_gather_scatter_pattern): Likewise.
12582         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
12583         (vect_recog_over_widening_pattern): Likewise.
12584         (vect_recog_gather_scatter_pattern): Likewise.
12585         (vect_recog_func_ptr): Move from tree-vectorizer.h
12586         (vect_vect_recog_func_ptrs): Move further down the file.
12587         (vect_recog_func): Likewise.  Remove the third argument.
12588         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
12589         (vect_pattern_recog_1): Expect the pattern function to do any
12590         necessary target tests.  Also expect it to provide a vector type.
12591         Remove the type_in handling.
12593 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12595         * tree-vect-patterns.c (vect_pattern_detected): New function.
12596         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
12597         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
12598         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
12599         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
12600         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12601         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12602         (vect_recog_mask_conversion_pattern)
12603         (vect_try_gather_scatter_pattern): Likewise.
12605 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12607         * tree-vect-patterns.c (vect_get_internal_def): New function.
12608         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12609         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
12610         (search_type_for_mask_1): Use it.
12612 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12614         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12615         redundant WIDEN_SUM_EXPR handling.
12616         (vect_recog_sad_pattern): Likewise.
12618 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12620         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12621         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
12622         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
12623         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
12624         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
12626 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12628         * tree-vect-stmts.c (vectorizable_call): Make sure that we
12629         use the stmt_vec_info of the original bb statement for the
12630         new zero assignment, even if the call is part of a pattern.
12632 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12634         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
12635         that the sequence is attached to the original statement rather
12636         than the pattern statement.
12637         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
12638         PATTERN_DEF_SEQ from the original statement rather than
12639         the main pattern statement.
12640         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
12641         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
12642         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
12644 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12646         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
12647         definition statements before the early exit for statements that aren't
12648         live or relevant.
12649         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
12650         split out from...
12651         (vect_transform_loop): ...here.  Process pattern definition
12652         statements without first checking whether the main pattern
12653         statement is live or relevant.
12655 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12657         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
12658         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
12660 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
12662         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
12663         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
12664         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
12665         (expand_block_compare): Change select_block_compare_mode call.
12666         (expand_strncmp_align_check): Use new functions, fix comment.
12667         (emit_final_str_compare_gpr): New function.
12668         (expand_strn_compare): Refactor and clean up code.
12669         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
12671 2018-06-19  Tony Reix  <tony.reix@atos.com>
12672             Damien Bergamini  <damien.bergamini@atos.com>
12673             David Edelsohn  <dje.gcc@gmail.com>
12675         * collect2.c (static_obj): New variable.
12676         (static_libs): New variable.
12677         (is_in_list): Uncomment declaration.
12678         (main): Track AIX libraries linked statically.
12679         (is_in_list): Uncomment definition.
12680         (scan_prog_file): Don't add AIX shared libraries initializer
12681         to constructor list if linking statically.
12683 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
12685         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
12686         constant.
12687         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
12689 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
12691         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
12692         blocks.
12694 2018-06-19  Martin Liska  <mliska@suse.cz>
12696         * config/i386/i386.c (ix86_can_inline_p): Do not use
12697         ipa_fn_summaries::get_create.
12698         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
12699         get.
12700         (devirtualization_time_bonus): Likewise.
12701         (ipcp_propagate_stage): Likewise.
12702         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12703         (edge_set_predicate): Likewise.
12704         (evaluate_conditions_for_known_args): Likewise.
12705         (evaluate_properties_for_edge): Likewise.
12706         (ipa_call_summary::reset): Tranform to ...
12707         (ipa_call_summary::~ipa_call_summary): ... this.
12708         (ipa_fn_summary::reset): Transform to ...
12709         (ipa_fn_summary::~ipa_fn_summary): ... this.
12710         (ipa_fn_summary_t::remove): Rename to ...
12711         (ipa_fn_summary_t::remove_callees): ... this.
12712         (ipa_fn_summary_t::duplicate): Use placement new
12713         instead of memory copy.
12714         (ipa_call_summary_t::duplicate): Likewise.
12715         (ipa_call_summary_t::remove): Remove.
12716         (dump_ipa_call_summary): Change get_create to get.
12717         (ipa_dump_fn_summary): Dump only when summary exists.
12718         (analyze_function_body): Use symbol_summary::get instead
12719         of get_create.
12720         (compute_fn_summary): Likewise.
12721         (estimate_edge_devirt_benefit): Likewise.
12722         (estimate_edge_size_and_time): Likewise.
12723         (inline_update_callee_summaries): Likewise.
12724         (remap_edge_change_prob): Likewise.
12725         (remap_edge_summaries): Likewise.
12726         (ipa_merge_fn_summary_after_inlining): Likewise.
12727         (write_ipa_call_summary): Likewise.
12728         (ipa_fn_summary_write): Likewise.
12729         (ipa_free_fn_summary): Likewise.
12730         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
12731         (struct ipa_call_summary): Likewise.
12732         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
12733         of get_create.
12734         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
12735         (estimate_size_after_inlining): Likewise.
12736         (estimate_growth): Likewise.
12737         (growth_likely_positive): Likewise.
12738         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12739         (inline_call): Likewise.
12740         * ipa-inline.c (caller_growth_limits): Likewise.
12741         (can_inline_edge_p): Likewise.
12742         (can_inline_edge_by_limits_p): Likewise.
12743         (compute_uninlined_call_time): Likewise.
12744         (compute_inlined_call_time): Likewise.
12745         (want_inline_small_function_p): Likewise.
12746         (edge_badness): Likewise.
12747         (update_caller_keys): Likewise.
12748         (update_callee_keys): Likewise.
12749         (inline_small_functions): Likewise.
12750         (inline_to_all_callers_1): Likewise.
12751         (dump_overall_stats): Likewise.
12752         (early_inline_small_functions): Likewise.
12753         (early_inliner): Likewise.
12754         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12755         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12756         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12757         * ipa-split.c (execute_split_functions): Likewise.
12758         * symbol-summary.h: Likewise.
12759         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12761 2018-06-19  Richard Biener  <rguenther@suse.de>
12763         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
12764         (vectorize_loops): ... here.  Fix dbgcnt handling.
12765         (try_vectorize_loop): Wrap try_vectorize_loop_1.
12767 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
12769         PR target/86197
12770         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
12771         ieee128 argument takes up only one (vector) register, not two (floating
12772         point) registers.
12774 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12776         * gimplify.c (gimplify_init_constructor): Really never clear for an
12777         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
12779 2018-06-19  Richard Biener  <rguenther@suse.de>
12781         PR tree-optimization/86179
12782         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
12783         after failed recognition.
12785 2018-06-18  Martin Sebor  <msebor@redhat.com>
12787         PR middle-end/85602
12788         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12789         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12790         Handle integer subtraction.
12791         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12792         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12794 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12796         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12797         param from rtx to rtx_insn *.
12798         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12799         param.
12800         (frv_ifcvt_modify_insn): Likwise.
12801         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12802         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
12803         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12804         as_a <rtx_insn *> cast to local "unprotected_region" once
12805         it's been established that it's not NULL or pc_rtx.
12806         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12807         param "sethi" from rtx to rtx_insn *.
12808         (nds32_group_float_insns): Likewise for param "insn".
12809         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12810         param.
12811         (vax_output_int_subtract): Likewise.
12812         * config/vax/vax.c (vax_output_int_add): Likewise for param
12813         "insn".
12814         (vax_output_int_subtract): Likewise.
12815         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12816         (emit_pattern_after): Likewise for param "after".
12817         (emit_insn_after): Likewise.
12818         (emit_jump_insn_after): Likewise.
12819         (emit_call_insn_after): Likewise.
12820         (emit_debug_insn_after): Likewise.
12821         (emit_pattern_before): Likewise for param "before".
12822         (emit_insn_before): Likewise.
12823         (emit_jump_insn_before): Likewise.
12824         * final.c (get_insn_template): Likewise for param "insn", removing
12825         a cast.
12826         * output.h (get_insn_template): Likewise for 2nd param.
12827         * rtl.h (emit_insn_before): Likewise.
12828         (emit_jump_insn_before): Likewise.
12829         (emit_debug_insn_before_noloc): Likewise.
12830         (emit_insn_after): Likewise.
12831         (emit_jump_insn_after): Likewise.
12832         (emit_call_insn_after): Likewise.
12833         (emit_debug_insn_after): Likewise.
12834         (set_insn_deleted): Likewise for param.
12836 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
12838         PR target/85358
12839         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12840         floating point modes, so that IFmode is numerically greater than
12841         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12842         to declare the ordering.  This prevents IFmode from being
12843         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12844         machine.  Include rs6000-modes.h to share the fractional values
12845         between genmodes* and the rest of the compiler.
12846         (IFmode): Likewise.
12847         (KFmode): Likewise.
12848         (TFmode): Likewise.
12849         * config/rs6000/rs6000-modes.h: New file.
12850         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12851         meaning of rs6000_long_double_size so that 126..128 selects an
12852         appropriate 128-bit floating point type.
12853         (rs6000_option_override_internal): Likewise.
12854         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12855         (TARGET_LONG_DOUBLE_128): Change the meaning of
12856         rs6000_long_double_size so that 126..128 selects an appropriate
12857         128-bit floating point type.
12858         (LONG_DOUBLE_TYPE_SIZE): Update comment.
12859         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12860         source and destination to match the standard usage.
12861         (truncifkf2): Likewise.
12862         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12863         ISA 2.07 to use an explicit clobber, instead of passing in a
12864         temporary.
12865         (copysign<mode>3_soft): Likewise.
12867 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12869         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12870         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12871         (vect_slp_analyze_instance_dependence): Likewise.
12872         (vect_enhance_data_refs_alignment): Likewise.
12873         (vect_analyze_data_refs_alignment): Likewise.
12874         (vect_slp_analyze_and_verify_instance_alignment
12875         (vect_analyze_data_ref_accesses): Likewise.
12876         (vect_prune_runtime_alias_test_list): Likewise.
12877         (vect_analyze_data_refs): Likewise.
12878         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12879         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12880         (vect_analyze_scalar_cycles_1): Likewise.
12881         (vect_get_loop_niters): Likewise.
12882         (vect_analyze_loop_form_1): Likewise.
12883         (vect_update_vf_for_slp): Likewise.
12884         (vect_analyze_loop_operations): Likewise.
12885         (vect_analyze_loop): Likewise.
12886         (vectorizable_induction): Likewise.
12887         (vect_transform_loop): Likewise.
12888         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12889         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12890         (vect_make_slp_decision): Likewise.
12891         (vect_detect_hybrid_slp): Likewise.
12892         (vect_slp_analyze_operations): Likewise.
12893         (vect_slp_bb): Likewise.
12894         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12895         (vectorizable_bswap): Likewise.
12896         (vectorizable_call): Likewise.
12897         (vectorizable_simd_clone_call): Likewise.
12898         (vectorizable_conversion): Likewise.
12899         (vectorizable_assignment): Likewise.
12900         (vectorizable_shift): Likewise.
12901         (vectorizable_operation): Likewise.
12902         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12904 2018-06-18  Martin Sebor  <msebor@redhat.com>
12906         PR tree-optimization/81384
12907         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12908         * builtins.c (expand_builtin_strnlen): New function.
12909         (expand_builtin): Call it.
12910         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12911         * builtins.def (BUILT_IN_STRNLEN): New.
12912         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12913         Warn for bounds in excess of maximum object size.
12914         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12915         single-value ranges.  Handle strnlen.
12916         (handle_builtin_strlen): Handle strnlen.
12917         (strlen_check_and_optimize_stmt): Same.
12918         * doc/extend.texi (Other Builtins): Document strnlen.
12920 2018-06-18  Maya Rashish  <coypu@sdf.org>
12922         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12923         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12924         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12926         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12927         here to ...
12928         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12930 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12932         * tree.c (escaped_string::escape): Replace cast to char * by
12933         const_cast<char *> (unescaped).
12935 2018-06-18  Nick Clifton  <nickc@redhat.com>
12937         PR 84195
12938         * tree.c (escaped_string): New class.  Converts an unescaped
12939         string into its escaped equivalent.
12940         (warn_deprecated_use): Use the new class to convert the
12941         deprecation message, if present.
12942         (test_escaped_strings): New self test.
12943         (test_c_tests): Add test_escaped_strings.
12944         * doc/extend.texi (deprecated): Add a note that the
12945         deprecation message is affected by the -fmessage-length
12946         option, and that control characters will be escaped.
12947         (#pragma GCC error): Document this pragma.
12948         (#pragma GCC warning): Likewise.
12949         * doc/invoke.texi (-fmessage-length): Document this option's
12950         effect on the #warning and #error preprocessor directives and
12951         the deprecated attribute.
12953 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
12955         * tree.c (decl_value_expr_lookup): Revert latest change.
12956         (decl_value_expr_insert): Likewise.
12958 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
12960         * gimplify.c (nonlocal_vlas): Delete.
12961         (nonlocal_vla_vars): Likewise.
12962         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12963         referenced VLAs.
12964         (gimplify_body): Do not create and destroy nonlocal_vlas.
12965         * tree-nested.c: Include diagnostic.h.
12966         (use_pointer_in_frame): Tweak.
12967         (lookup_field_for_decl): Add assertion and declare the transformation.
12968         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12969         internal error when the reference is in a wrong context.  Do not
12970         create a debug decl by default.
12971         (note_nonlocal_block_vlas): Delete.
12972         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12973         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
12974         create a debug decl by default.
12975         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12976         call is in a wrong context.
12977         (fixup_vla_decls): New function.
12978         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
12979         debug variables were created.
12980         * tree.c (decl_value_expr_lookup): Add checking assertion.
12981         (decl_value_expr_insert): Likewise.
12983 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12985         PR middle-end/82479
12986         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12987         * tree-scalar-evolution.c (interpret_expr): Likewise.
12988         (expression_expensive_p): Likewise.
12989         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12990         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12991         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12992         (ssa_defined_by_minus_one_stmt_p): New.
12994 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12996         PR middle-end/64946
12997         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12998         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12999         * dojump.c (do_jump): Likewise.
13000         * expr.c (expand_expr_real_2): Check operand type's sign.
13001         * fold-const.c (const_unop): Handle ABSU_EXPR.
13002         (fold_abs_const): Likewise.
13003         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13004         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13005         (strip_sign_op_1): Likesise.
13006         * match.pd: Add new pattern to generate ABSU_EXPR.
13007         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13008         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13009         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13010         * tree-inline.c (estimate_operator_cost): Likewise.
13011         * tree-pretty-print.c (dump_generic_node): Likewise.
13012         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13013         * tree.def (ABSU_EXPR): New.
13015 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13017         PR middle-end/86095
13018         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13019         documented as preserved for backward compatibility only.
13020         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13022         PR rtl-optimization/86108
13023         * bb-reorder.c (create_forwarder_block): Renamed to ...
13024         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13025         jump from new landing pad to the second part.
13026         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13027         Adjust callers.
13029 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13031         PR middle-end/85878
13032         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13033         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13034         Only call store_expr for halves if the mode is the same.
13036         PR middle-end/86123
13037         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13038         Fix up comment formatting.
13040 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13042         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13043         (typed_splay_tree::closure,
13044         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13045         (typed_splay_tree::typed_splay_tree,
13046         typed_splay_tree::operator =): Declared private.
13047         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13048         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13049         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13050         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13051         typed_splay_tree::splay_tree_splay,
13052         typed_splay_tree::splay_tree_foreach_helper,
13053         typed_splay_tree::splay_tree_insert,
13054         typed_splay_tree::splay_tree_remove,
13055         typed_splay_tree::splay_tree_lookup,
13056         typed_splay_tree::splay_tree_predecessor,
13057         typed_splay_tree::splay_tree_successor,
13058         typed_splay_tree::splay_tree_min,
13059         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13060         (typed_splay_tree::root, typed_splay_tree::comp,
13061         typed_splay_tree::delete_key,
13062         typed_splay_tree::delete_value): New data members.
13063         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13064         typed_splay_tree::remove.
13066 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13068         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13069         -mginv and -mno-ginv to the assembler.
13070         * config/mips/mips.opt (-mcrc): New option.
13071         (-mginv): Likewise.
13072         * doc/invoke.text (-mcrc): Document.
13073         (-mginv): Likewise.
13075 2018-06-15  Nick Clifton  <nickc@redhat.com>
13077         PR 84195
13078         * tree.c (escaped_string): New class.  Converts an unescaped
13079         string into its escaped equivalent.
13080         (warn_deprecated_use): Use the new class to convert the
13081         deprecation message, if present.
13082         (test_escaped_strings): New self test.
13083         (test_c_tests): Add test_escaped_strings.
13084         * doc/extend.texi (deprecated): Add a note that the
13085         deprecation message is affected by the -fmessage-length
13086         option, and that control characters will be escaped.
13087         (#pragma GCC error): Document this pragma.
13088         (#pragma GCC warning): Likewise.
13089         * doc/invoke.texi (-fmessage-length): Document this option's
13090         effect on the #warning and #error preprocessor directives and
13091         the deprecated attribute.
13093 2018-06-15  Richard Biener  <rguenther@suse.de>
13095         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13096         here, also noting vector size used.
13097         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13098         size used in MSG_OPTIMIZED_LOCATIONS dump.
13099         (pass_slp_vectorize::execute): Adjust.
13101 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13103         PR target/85968
13104         * config/arc/arc.c (arc_return_address_register): Fix
13105         if-condition.
13107 2018-06-15  Richard Biener  <rguenther@suse.de>
13109         PR middle-end/86159
13110         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13111         leave useless conversion stripping to force_gimple_operand_gsi.
13112         (gimplify_build2): Likewise.
13113         (gimplify_build1): Likewise.
13115 2018-06-15  Richard Biener  <rguenther@suse.de>
13117         PR middle-end/86076
13118         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13119         before adjusting their block.
13121 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13123         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13124         multilibs for *-*-rtems*.
13125         * config/riscv/t-rtems: New file.
13127 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13129         PR middle-end/86122
13130         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13131         unsigned_type_for returns NULL.
13133         PR target/85945
13134         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13135         subregs of multi-word pseudos unless the float mode has word size.
13137 2018-06-14  Richard Biener  <rguenther@suse.de>
13139         PR middle-end/86139
13140         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13141         duplicate and harmful type_hash_canon.
13142         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13144 2018-06-14  Richard Biener  <rguenther@suse.de>
13146         PR ipa/86124
13147         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13148         NULL cgraph_node.
13150 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13152         * config/rtems.h (STDINT_LONG32): Define.
13154 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13155             Prachi Godbole  <prachi.godbole@imgtec.com>
13157         * config/mips/mips-cpus.def: Define P6600.
13158         * config/mips/mips-tables.opt: Regenerate.
13159         * config/mips/mips.c (mips_ucbranch_type): New enum.
13160         (mips_rtx_cost_data): Add support for P6600.
13161         (mips_issue_rate): Likewise.
13162         (mips_multipass_dfa_lookahead): Likewise.
13163         (mips_avoid_hazard): Likewise.
13164         (mips_reorg_process_insns): Likewise.
13165         (mips_classify_branch_p6600): New function.
13166         * config/mips/mips.h (TUNE_P6600): New define.
13167         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13168         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13169         * config/mips/mips.md: Include p6600.md.
13170         (processor): Add p6600.
13171         * config/mips/p6600.md: New file.
13172         * doc/invoke.texi: Add p6600 to supported architectures.
13174 2018-06-13  Martin Sebor  <msebor@redhat.com>
13176         PR tree-optimization/86114
13177         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13178         of integer types.
13179         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13181 2018-06-13  Richard Biener  <rguenther@suse.de>
13183         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13184         Properly set vector type of the intermediate stmt.
13185         * tree-vect-stmts.c (vectorizable_operation): The destination
13186         var always has vectype_out type.
13188 2018-06-13  Jeff Law  <law@redhat.com>
13190         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13191         integer 0 for argument to print_rtl_with_bb.
13192         (rl78_reorg): Likewise.
13194 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13196         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13197         from rtx to rtx_insn *.
13198         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13199         "label".
13200         (add_sched_insns_for_speculation): Likewise for local "target",
13201         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13202         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13203         from rtx_insn ** to rtx_code_label **.
13204         (reorg_emit_nops): Likewise.
13205         (c6x_reorg): Likewise for local "call_labels".
13206         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13207         rtx to rtx_insn *.
13208         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13209         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13210         the loops over LABEL_REFS.
13211         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13212         braf_label.
13213         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13214         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13215         (split_branches): Strengthen local "olabel" from rtx to
13216         rtx_insn *, adding a safe_as_a cast.
13217         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13218         to "rtx_insn *".
13219         (add_insn_after): Likewise for first two params.
13220         (add_insn_before): Likewise.
13221         (remove_insn): Likewise for param.
13222         (emit_pattern_before_noloc): Likewise for second and third params.
13223         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13224         (emit_call_insn_before_noloc): Likewise.
13225         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13226         to "rtx_insn *".
13227         (emit_barrier_before): Likewise.
13228         (emit_label_before): Strengthen "label" param from "rtx" to
13229         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13230         "rtx_insn *".
13231         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13232         "rtx_insn *".
13233         (emit_pattern_after_noloc): Likewise.
13234         (emit_insn_after_noloc): Likewise.
13235         (emit_jump_insn_after_noloc): Likewise.
13236         (emit_call_insn_after_noloc): Likewise.
13237         (emit_debug_insn_after_noloc): Likewise.
13238         (emit_barrier_after): Likewise.
13239         (emit_label_after): Likewise for both params.
13240         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13241         "loc" param from "int" to "location_t".
13242         (emit_insn_after_setloc): Likewise.
13243         (emit_jump_insn_after_setloc): Likewise.
13244         (emit_call_insn_after_setloc): Likewise.
13245         (emit_debug_insn_after_setloc): Likewise.
13246         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13247         "loc" param from "int" to "location_t".
13248         (emit_pattern_before): Convert NULL_RTX to NULL.
13249         (emit_insn_before_setloc): Convert "loc" param from "int" to
13250         "location_t".
13251         (emit_jump_insn_before_setloc): Likewise.
13252         (emit_call_insn_before_setloc): Likewise.
13253         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13254         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13255         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13256         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13257         Convert 3rd param from "int" to "location_t".
13258         (emit_barrier_before, emit_barrier_after, next_real_insn):
13259         Strengthen param from rtx to rtx_insn *.
13260         (emit_label_before): Strengthen 1st param from "rtx" to
13261         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13262         "rtx_insn *".
13263         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13264         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13265         Strengthen 2nd param from "rtx" to "rtx_insn *".
13266         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13267         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13268         Likewise. Convert 3rd param from "int" to "location_t".
13269         (emit_label_after): Strengthen 1st param from "rtx" to
13270         "rtx_code_label *".
13271         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13272         "rtx_insn *".
13273         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13274         from "rtx" to "rtx_insn *".
13276 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13278         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13279         bodies streamed in with -Q.
13280         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13281         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13282         * gimple-streamer-out.c: Include gimple-pretty-print.h
13283         (output_bb): Dump stmts streamed.
13284         * lto-section-out.c: Include print-tree.h
13285         (lto_begin_section): Dump sections created.
13286         (lto_output_decl_index): Dump decl encoded.
13287         * lto-streamer-out.c: Include print-tree.h
13288         (create_output_block): Dump output block created.
13289         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13290         (output_function): Dump function output.
13291         (output_constructor): Dump constructor streamed.
13292         (write_global_stream): Output indexes encoded.
13293         (produce_asm_for_decls): Dump streams encoded.
13294         * lto-streamer.c (streamer_dump_file): New global var.
13295         * lto-streamer.h (streamer_dump_file): Declare.
13296         * passes.c (ipa_write_summaries): Initialize streamer dump.
13297         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13298         in.
13300 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13302         PR target/86048
13303         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13304         offsets for register save directives.  Emit a second batch of save
13305         directives, if need be, when the function accesses prior frames.
13307 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13309         * config/arc/fpu.md (fmasf4): Force operand to register.
13310         (fnmasf4): Likewise.
13312 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13314         * config/arc/arc-protos.h (arc_pad_return): Remove.
13315         * config/arc/arc.c (machine_function): Remove force_short_suffix
13316         and size_reason.
13317         (arc_print_operand): Adjust printing of '&'.
13318         (arc_verify_short): Remove conditional printing of short suffix.
13319         (arc_final_prescan_insn): Remove reference to size_reason.
13320         (pad_return): New function.
13321         (arc_reorg): Call pad_return.
13322         (arc_pad_return): Remove.
13323         (arc_init_machine_status): Remove reference to force_short_suffix.
13324         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13325         (attr length): When attribute iscompact is true force to 2
13326         regardless; in the case of maybe check if we want to force the
13327         instruction to have 4 bytes length.
13328         (nopv): Change it to generate 4 byte long nop as well.
13329         (blockage): New pattern.
13330         (simple_return): Remove call to arc_pad_return.
13331         (p_return_i): Likewise.
13333 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13335         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13337 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13339         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13340         ARC cores.
13342 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13344         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13345         for ARC700 and ARCv2.
13347 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13349         PR target/86076
13350         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13351         operands[2] instead of operands[1].
13354 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13356         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13357         case, check whether the outer register overlaps an unallocatable
13358         register, not just whether it fits the required class.
13360 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13362         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13363         are poly_ints.
13364         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13365         (memrefs_conflict_p): Likewise.
13366         (init_alias_analysis): Likewise.
13367         * cfgexpand.c (expand_debug_expr): Likewise.
13368         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13369         * cse.c (fold_rtx): Likewise.
13370         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13371         * expr.c (emit_block_move_hints): Likewise.
13372         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13373         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13374         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13375         (emit_group_store): Likewise.
13376         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13377         to read the PRE/POST_MODIFY increment.
13378         * calls.c (store_one_arg): Use strip_offset.
13379         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13380         poly_int_rtx_p.
13381         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13382         by a VEC_SELECT.
13383         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13384         (simplify_binary_operation_1): Extend CONST_INT handling to
13385         poly_int_rtx_p.
13386         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13387         than a HOST_WIDE_INT.
13388         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13389         poly_int64.
13390         (adjust_mems, add_stores): Update accodingly.
13391         (vt_canonicalize_addr): Track polynomial offsets.
13392         (emit_note_insn_var_location): Likewise.
13393         (vt_add_function_parameter): Likewise.
13394         (vt_initialize): Likewise.
13396 2018-06-12  Jeff Law  <law@redhat.com>
13398         * config.gcc (alpha*-*-freebsd*): Remove.
13399         * config/alpha/freebsd.h: Remove.
13401 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
13403         PR other/69968
13404         * spellcheck-tree.c (levenshtein_distance): Rename to...
13405         (get_edit_distance): ...this, and update for underlying renaming.
13406         * spellcheck-tree.h (levenshtein_distance): Rename to...
13407         (get_edit_distance): ...this.
13408         * spellcheck.c (levenshtein_distance): Rename to...
13409         (get_edit_distance): ...this.  Convert from Levenshtein distance
13410         to Damerau-Levenshtein distance by supporting transpositions of
13411         adjacent characters.  Rename "v1" to "v_next" and "v0" to
13412         "v_one_ago".
13413         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13414         (selftest::test_edit_distance_unit_test_oneway): ...this, and
13415         update for underlying renaming.
13416         (selftest::levenshtein_distance_unit_test): Rename to...
13417         (selftest::test_get_edit_distance_unit): ...this, and update for
13418         underlying renaming.
13419         (selftest::test_find_closest_string): Add example from PR 69968
13420         where transposition helps
13421         (selftest::test_metric_conditions): Update for renaming.
13422         (selftest::test_metric_conditions): Likewise.
13423         (selftest::spellcheck_c_tests): Likewise.
13424         * spellcheck.h (levenshtein_distance): Rename both overloads to...
13425         (get_edit_distance): ...this.
13426         (best_match::consider): Update for renaming.
13428 2018-06-12  Martin Sebor  <msebor@redhat.com>
13430         PR tree-optimization/85259
13431         * builtins.c (compute_objsize): Handle constant offsets.
13432         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13433         true iff a warning has been issued.
13434         * gimple.h (gimple_nonartificial_location): New function.
13435         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13436         gimple_nonartificial_location and handle -Wno-system-headers.
13437         (handle_builtin_stxncpy): Same.
13439 2018-06-12  Martin Sebor  <msebor@redhat.com>
13441         PR c/85931
13442         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13444 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
13446         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13447         BUILTIN_VEC_XST entries for pointer to double and long long.
13449 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
13451         PR target/85990
13452         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13453         Update comments.
13454         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13455         Likewise.
13457 2018-06-12  Martin Liska  <mliska@suse.cz>
13459         * doc/options.texi: Document IntegerRange.
13461 2018-06-12  Martin Liska  <mliska@suse.cz>
13463         * config/i386/i386.opt: Make MPX-related options as Deprecated.
13464         * opt-functions.awk: Handle Deprecated flag.
13465         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13466         and report error.
13467         (read_cmdline_option): Report warning for a deprecated option.
13468         * opts.h (struct cl_option): Add new field cl_deprecated.
13469         (CL_ERR_DEPRECATED): New.
13471 2018-06-12  Martin Liska  <mliska@suse.cz>
13473         * doc/options.texi: Document Deprecated option flag.
13475 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13477         * config/arc/arc-arch.h (arc_extras): New enum.
13478         (arc_cpu_t):Add field extra.
13479         (arc_cpu_types): Consider the extras.
13480         * config/arc/arc-cpus.def: Add extras info.
13481         * config/arc/arc-opts.h (processor_type): Consider extra field.
13482         * config/arc/arc.c (arc_override_options): Handle extra field.
13484 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13486         * config/arc/arc-arch.h: Update ARC_OPTX macro.
13487         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13488         field.
13489         * config/arc/arc.c (arc_init): Update pic warning.
13490         (irq_range): Update irq range parsing warnings.
13491         (arc_override_options): Update various warning messages.
13492         (arc_handle_aux_attribute): Likewise.
13494 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
13496         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13498 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13500         * doc/sourcebuild.texi: Document usage of line number 0 in verify
13501         compiler messages directives.
13503 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
13505         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13506         * config/mips/mips-tables.opt: Regenerate.
13507         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
13508         mips64r6.
13509         * doc/invoke.texi: Document -march=i6500.
13511 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
13513         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
13514         (i6400_gpmul): Add cpu_unit.
13515         (i6400_gpdiv): Likewise.
13516         (i6400_msa_add_d): Update reservations.
13517         (i6400_msa_int_add) Likewise.
13518         (i6400_msa_short_logic3) Likewise.
13519         (i6400_msa_short_logic2) Likewise.
13520         (i6400_msa_short_logic) Likewise.
13521         (i6400_msa_move) Likewise.
13522         (i6400_msa_cmp) Likewise.
13523         (i6400_msa_short_float2) Likewise.
13524         (i6400_msa_div_d) Likewise.
13525         (i6400_msa_long_logic1) Likewise.
13526         (i6400_msa_long_logic2) Likewise.
13527         (i6400_msa_mult) Likewise.
13528         (i6400_msa_long_float2) Likewise.
13529         (i6400_msa_long_float4) Likewise.
13530         (i6400_msa_long_float5) Likewise.
13531         (i6400_msa_long_float8) Likewise.
13532         (i6400_fpu_fadd): Include frint type.
13533         (i6400_fpu_store): New define_insn_reservation.
13534         (i6400_fpu_load): Likewise.
13535         (i6400_fpu_move): Likewise.
13536         (i6400_fpu_fcmp): Likewise.
13537         (i6400_fpu_fmadd): Likewise.
13538         (i6400_int_mult): Include imul3nc type and update reservation.
13539         (i6400_int_div): Include idiv3 type and update reservation.
13540         (i6400_int_load): Update to check type not move_type.
13541         (i6400_int_store): Likewise.
13542         (i6400_int_prefetch): Set zero latency.
13544 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
13546         * gcc.c: Document new %@{...} sequence.
13547         (LINK_COMMAND_SPEC): Use it for the -L switches.
13548         (cpp_unique_options): Use it for the -I switches.
13549         (at_file_argbuf): New global variable.
13550         (in_at_file): Likewise.
13551         (alloc_args): Create at_file_argbuf.
13552         (clear_args): Truncate at_file_argbuf.
13553         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
13554         (open_at_file): New function.
13555         (close_at_file): Likewise.
13556         (create_at_file): Delete.
13557         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
13558         <'o'>: Likewise.
13559         <'@'>: New case.
13560         (validate_switches_from_spec): Deal with %@{...} sequence.
13561         (validate_switches): Likewise.
13562         (driver::finalize): Call clear_args.
13564 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
13566         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
13568 2018-06-11  Martin Sebor  <msebor@redhat.com>
13570         * doc/invoke.texi (-Wall): List -Wc++17-compat.
13571         (Wno-class-memaccess): Add @opindex.
13572         (Wno-templates, Wno-multiple-inheritance): Same.
13573         (Wno-virtual-inheritance, Wno-namespaces): Same.
13574         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
13575         (Wno-format-overflow, Wno-format-truncation): Same.
13576         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
13577         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
13578         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
13579         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
13580         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
13581         (Wno-misspelled-isr): Same.
13583 2018-06-11  Martin Sebor  <msebor@redhat.com>
13585         * PR tree-optimization/86083
13586         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
13588 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
13590         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
13592 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
13594         PR target/85755
13595         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
13596         on the correct operand.
13597         (*movdi_internal64): Ditto.
13599 2018-06-11  Martin Liska  <mliska@suse.cz>
13601         PR tree-optimization/86089
13602         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
13604 2018-06-11  Julia Koval  <julia.koval@intel.com>
13606         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
13607         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
13608         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
13610 2018-06-11  Olivier Hainque  <hainque@adacore.com>
13612         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
13613         for Ada with strict dwarf2.
13615 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
13617         PR target/85755
13618         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
13619         addresses.
13621 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
13623         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
13625 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
13627         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
13628         TARGET_ELF.
13630 2018-06-08  Martin Liska  <mliska@suse.cz>
13632         * tree-cfg.h (debug_function): Fix argument type to match
13633         implementation.
13635 2018-06-08  Martin Liska  <mliska@suse.cz>
13637         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
13638         Remove usage of MPX-related (and removed) fields.
13639         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
13641 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
13643         * cfg.c (debug): Use TDF_NONE rather than 0.
13644         * cfghooks.c (debug): Likewise.
13645         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
13646         (struct dump_option_value_info): Convert to...
13647         (struct kv_pair): ...this template type.
13648         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
13649         rather than 0.
13650         (optinfo_verbosity_options): Likewise.
13651         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
13652         OPTGROUP_NONE.
13653         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
13654         than int for "optgroup_flags" param.
13655         (dump_generic_expr_loc): Use dump_flags_t rather than int for
13656         "dump_kind" param.
13657         (dump_dec): Likewise.
13658         (dump_finish): Use TDF_NONE rather than 0.
13659         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
13660         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
13661         than 0.  Update for change to option_ptr.
13662         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
13663         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
13664         0.  Update for changes to optinfo_verbosity_options and
13665         optgroup_options.
13666         (opt_info_switch_p): Convert optgroup_flags from int to
13667         optgroup_flags_t.
13668         (dump_basic_block): Use dump_flags_t rather than int
13669         for "dump_kind" param.
13670         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
13671         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
13672         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
13673         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
13674         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
13675         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
13676         TDF_NONE): Convert from macros to...
13677         (enum dump_flag): ...this new enum.
13678         (dump_flags_t): Update to use enum.
13679         (operator|, operator&, operator~, operator|=, operator&=):
13680         Implement for dump_flags_t.
13681         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
13682         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
13683         Convert from macros to...
13684         (enum optgroup_flag): ...this new enum.
13685         (optgroup_flags_t): New typedef.
13686         (operator|, operator|=): Implement for optgroup_flags_t.
13687         (struct dump_file_info): Convert field "alt_flags" to
13688         dump_flags_t.  Convert field "optgroup_flags" to
13689         optgroup_flags_t.
13690         (dump_basic_block): Use dump_flags_t rather than int for param.
13691         (dump_generic_expr_loc): Likewise.
13692         (dump_dec): Likewise.
13693         (dump_register): Convert param "optgroup_flags" to
13694         optgroup_flags_t.
13695         (opt_info_enable_passes): Likewise.
13696         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
13697         than 0.
13698         * gimple-pretty-print.c (debug): Likewise.
13699         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
13700         (merged_store_group::apply_stores): Likewise.
13701         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
13702         * gimple.c (verify_gimple_pp): Likewise.
13703         * graphite-poly.c (print_pbb_body): Likewise.
13704         * passes.c (pass_manager::register_one_dump_file): Convert
13705         local "optgroup_flags" to optgroup_flags_t.
13706         * print-tree.c (print_node): Use TDF_NONE rather than 0.
13707         (debug): Likewise.
13708         (debug_body): Likewise.
13709         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
13710         to optgroup_flags_t.
13711         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
13712         than 0.
13713         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
13714         (convert_mult_to_fma): Likewise.
13715         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
13716         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
13717         * tree-vect-data-refs.c (dump_lower_bound): Convert param
13718         "dump_kind" to dump_flags_t.
13720 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13722         * config/rs6000/rs6000.c (min, max): Delete.
13724 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13726         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
13727         -mabi=spe and -mabi=no-spe.
13729 2018-06-08  Martin Liska  <mliska@suse.cz>
13731         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
13732         where we expect an existing summary.
13734 2018-06-08  Martin Liska  <mliska@suse.cz>
13736         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
13737         * ipa-inline.h (estimate_edge_growth): Likewise.
13739 2018-06-08  Martin Liska  <mliska@suse.cz>
13741         * cgraph.c (function_version_hasher::hash): Use
13742         cgraph_node::get_uid ().
13743         (function_version_hasher::equal):
13744         * cgraph.h (cgraph_node::get_uid): New method.
13745         * ipa-inline.c (update_caller_keys): Use
13746         cgraph_node::get_uid ().
13747         (update_callee_keys): Likewise.
13748         * ipa-utils.c (searchc): Likewise.
13749         (ipa_reduced_postorder): Likewise.
13750         * lto-cgraph.c (input_node): Likewise.
13751         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
13752         * symbol-summary.h (symtab_insertion): Likewise.
13753         (symtab_removal): Likewise.
13754         (symtab_duplication): Likewise.
13755         * tree-pretty-print.c (dump_function_header): Likewise.
13756         * tree-sra.c (convert_callers_for_node): Likewise.
13758 2018-06-08  Martin Liska  <mliska@suse.cz>
13760         * cgraph.c (symbol_table::create_edge): Always assign a new
13761         unique number.
13762         (symbol_table::free_edge): Do not recycle numbers.
13763         * cgraph.h (cgraph_edge::get): New method.
13764         * symbol-summary.h (symtab_removal): Use it.
13765         (symtab_duplication): Likewise.
13766         (call_summary::hashable_uid): Remove.
13768 2018-06-08  Martin Liska  <mliska@suse.cz>
13770         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
13771         (initialize_growth_caches): Remove.
13772         (free_growth_caches): Likewise.
13773         (do_estimate_edge_time): Use edge_growth_cache.
13774         (do_estimate_edge_size): Likewise.
13775         (do_estimate_edge_hints): Likewise.
13776         * ipa-inline.c (reset_edge_caches): Likewise.
13777         (recursive_inlining): Likewise.
13778         (inline_small_functions): Likewise.
13779         * ipa-inline.h (initialize_growth_caches): Remove.
13780         (estimate_edge_size): Likewise.
13781         (estimate_edge_time): Likewise.
13782         (estimate_edge_hints): Likewise.
13783         (reset_edge_growth_cache): Likewise.
13784         * symbol-summary.h (call_summary::remove): New method.
13786 2018-06-08  Martin Liska  <mliska@suse.cz>
13788         * ipa-cp.c (class edge_clone_summary): New summary.
13789         (grow_edge_clone_vectors): Remove.
13790         (ipcp_edge_duplication_hook): Remove.
13791         (class edge_clone_summary_t): New call_summary class.
13792         (ipcp_edge_removal_hook): Remove.
13793         (edge_clone_summary_t::duplicate): New function.
13794         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13795         (create_specialized_node): Likewise.
13796         (ipcp_driver): Initialize edge_clone_summaries and do not
13797         register hooks.
13799 2018-06-08  Martin Liska  <mliska@suse.cz>
13801         * symbol-summary.h (get): New function.
13802         (call_summary::m_initialize_when_cloning): New class member.
13804 2018-06-08  Martin Liska  <mliska@suse.cz>
13806         * cgraph.c (cgraph_node::remove): Do not recycle uid.
13807         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13808         (symbol_table::allocate_cgraph_symbol): Do not set uid.
13809         * passes.c (uid_hash_t): Record removed_nodes by their uids.
13810         (remove_cgraph_node_from_order): Use the removed_nodes set.
13811         (do_per_function_toporder): Likwise.
13812         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13813         instead of summary_uid.
13814         (symtab_removal): Likewise.
13815         (symtab_duplication): Likewise.
13817 2018-06-08  Martin Liska  <mliska@suse.cz>
13819         * ipa-cp.c (ipcp_store_bits_results): Use
13820         ipcp_transformation_sum.
13821         (ipcp_store_vr_results): Likewise.
13822         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13823         to ...
13824         (ipcp_transformation_initialize): ... this.
13825         (ipa_set_node_agg_value_chain):
13826         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13827         (write_ipcp_transformation_info): Likewise.
13828         (read_ipcp_transformation_info): Likewise.
13829         (ipcp_update_bits): Likewise.
13830         (ipcp_update_vr): Likewise.
13831         (ipcp_transform_function): Likewise.
13832         * ipa-prop.h: Rename ipcp_transformation_summary to
13833         ipcp_transformation.
13834         (class ipcp_transformation_t): New function summary.
13835         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13836         (ipa_get_agg_replacements_for_node): Likewise.
13838 2018-06-08  Martin Liska  <mliska@suse.cz>
13840         * ipa-pure-const.c (struct funct_state_d): Do it class instead
13841         of struct.
13842         (class funct_state_summary_t): New function_summary class.
13843         (has_function_state): Remove.
13844         (get_function_state): Likewise.
13845         (set_function_state): Likewise.
13846         (add_new_function): Likewise.
13847         (funct_state_summary_t::insert): New function.
13848         (duplicate_node_data): Remove.
13849         (remove_node_data): Remove.
13850         (funct_state_summary_t::duplicate): New function.
13851         (register_hooks): Create new funct_state_summaries.
13852         (pure_const_generate_summary): Use it.
13853         (pure_const_write_summary): Likewise.
13854         (pure_const_read_summary): Likewise.
13855         (propagate_pure_const): Likewise.
13856         (propagate_nothrow): Likewise.
13857         (dump_malloc_lattice): Likewise.
13858         (propagate_malloc): Likewise.
13859         (execute): Do not register hooks, just remove summary
13860         instead.
13861         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13862         constructor.
13864 2018-06-08  Martin Liska  <mliska@suse.cz>
13866         * ipa-reference.c (remove_node_data): Remove.
13867         (duplicate_node_data): Likewise.
13868         (class ipa_ref_var_info_summary_t): New class.
13869         (class ipa_ref_opt_summary_t): Likewise.
13870         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13871         (get_reference_optimization_summary): Use
13872         ipa_ref_opt_sum_summaries.
13873         (set_reference_vars_info): Remove.
13874         (set_reference_optimization_summary): Likewise.
13875         (ipa_init): Create summaries.
13876         (init_function_info): Use function summary.
13877         (ipa_ref_opt_summary_t::duplicate): New function.
13878         (ipa_ref_opt_summary_t::remove): New function.
13879         (get_read_write_all_from_node): Fix GNU coding style.
13880         (propagate): Use function summary.
13881         (write_node_summary_p): Fix GNU coding style.
13882         (stream_out_bitmap): Likewise.
13883         (ipa_reference_read_optimization_summary): Use function summary.
13884         (ipa_reference_c_finalize): Do not release hooks.
13886 2018-06-08  Martin Liska  <mliska@suse.cz>
13888         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13889         (analyze_function_body): Extract multiple calls of get_create.
13890         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13891         * ipa-inline.c (recursive_inlining): Use ::get method.
13892         * ipa-inline.h (estimate_edge_growth): Likewise.
13894 2018-06-08  Martin Liska  <mliska@suse.cz>
13896         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13897         HSA_INVALID.
13898         (hsa_function_summary::hsa_function_summary): Use the new enum
13899         value.
13900         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13901         * hsa-gen.c (hsa_get_host_function): Likewise.
13902         (get_brig_function_name): Likewise.
13903         * ipa-hsa.c (process_hsa_functions): Likewise.
13904         (ipa_hsa_write_summary): Likewise.
13905         * symbol-summary.h (symtab_duplication): Use ::get function/
13906         (get): New function.
13908 2018-06-08  Martin Liska  <mliska@suse.cz>
13910         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13911         of get.
13912         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13913         (hsa_register_kernel): Likewise.
13914         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13915         * hsa-gen.c (hsa_get_host_function): Likewise.
13916         (get_brig_function_name): Likewise.
13917         (generate_hsa): Likewise.
13918         (pass_gen_hsail::execute): Likewise.
13919         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13920         (devirtualization_time_bonus): Likewise.
13921         (ipcp_propagate_stage): Likewise.
13922         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13923         (edge_set_predicate): Likewise.
13924         (evaluate_conditions_for_known_args): Likewise.
13925         (evaluate_properties_for_edge): Likewise.
13926         (ipa_fn_summary::reset): Likewise.
13927         (ipa_fn_summary_t::duplicate): Likewise.
13928         (dump_ipa_call_summary): Likewise.
13929         (ipa_dump_fn_summary): Likewise.
13930         (analyze_function_body): Likewise.
13931         (compute_fn_summary): Likewise.
13932         (estimate_edge_devirt_benefit): Likewise.
13933         (estimate_edge_size_and_time): Likewise.
13934         (estimate_calls_size_and_time): Likewise.
13935         (estimate_node_size_and_time): Likewise.
13936         (inline_update_callee_summaries): Likewise.
13937         (remap_edge_change_prob): Likewise.
13938         (remap_edge_summaries): Likewise.
13939         (ipa_merge_fn_summary_after_inlining): Likewise.
13940         (ipa_update_overall_fn_summary): Likewise.
13941         (read_ipa_call_summary): Likewise.
13942         (inline_read_section): Likewise.
13943         (write_ipa_call_summary): Likewise.
13944         (ipa_fn_summary_write): Likewise.
13945         (ipa_free_fn_summary): Likewise.
13946         * ipa-hsa.c (process_hsa_functions): Likewise.
13947         (ipa_hsa_write_summary): Likewise.
13948         (ipa_hsa_read_section): Likewise.
13949         * ipa-icf.c (sem_function::merge): Likewise.
13950         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13951         (do_estimate_edge_time): Likewise.
13952         (estimate_size_after_inlining): Likewise.
13953         (estimate_growth): Likewise.
13954         (growth_likely_positive): Likewise.
13955         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13956         (inline_call): Likewise.
13957         * ipa-inline.c (caller_growth_limits): Likewise.
13958         (can_inline_edge_p): Likewise.
13959         (can_inline_edge_by_limits_p): Likewise.
13960         (compute_uninlined_call_time): Likewise.
13961         (compute_inlined_call_time): Likewise.
13962         (want_inline_small_function_p): Likewise.
13963         (edge_badness): Likewise.
13964         (update_caller_keys): Likewise.
13965         (update_callee_keys): Likewise.
13966         (recursive_inlining): Likewise.
13967         (inline_small_functions): Likewise.
13968         (inline_to_all_callers_1): Likewise.
13969         (dump_overall_stats): Likewise.
13970         (early_inline_small_functions): Likewise.
13971         (early_inliner): Likewise.
13972         * ipa-inline.h (estimate_edge_growth): Likewise.
13973         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13974         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13975         * ipa-prop.h (IPA_NODE_REF): Likewise.
13976         (IPA_EDGE_REF): Likewise.
13977         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13978         (propagate_malloc): Likewise.
13979         * ipa-split.c (execute_split_functions): Likewise.
13980         * symbol-summary.h: Rename get to get_create.
13981         (get): Likewise.
13982         (get_create): Likewise.
13983         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13985 2018-06-08  Martin Liska  <mliska@suse.cz>
13987         * symbol-summary.h (release): Move definition out of class
13988         declaration.
13989         (symtab_removal): Likewise.
13990         (symtab_duplication): Likewise.
13992 2018-06-08  Martin Liska  <mliska@suse.cz>
13994         * symbol-summary.h (function_summary): Move constructor
13995         implementation out of class declaration.
13996         (release): Likewise.
13997         (symtab_insertion): Likewise.
13998         (symtab_removal): Likewise.
13999         (symtab_duplication): Likewise.
14000         (get): Likewise.
14002 2018-06-08  Martin Liska  <mliska@suse.cz>
14004         * Makefile.in: Remove support for MPX (macros, related functions,
14005         fields in cgraph_node, ...).
14006         * builtin-types.def (BT_BND): Likewise.
14007         (BT_FN_BND_CONST_PTR): Likewise.
14008         (BT_FN_CONST_PTR_BND): Likewise.
14009         (BT_FN_VOID_PTR_BND): Likewise.
14010         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14011         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14012         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14013         (expand_builtin_mempcpy_with_bounds): Likewise.
14014         (expand_builtin_memset_with_bounds): Likewise.
14015         (expand_builtin_memset_args): Likewise.
14016         (std_expand_builtin_va_start): Likewise.
14017         (expand_builtin): Likewise.
14018         (expand_builtin_with_bounds): Likewise.
14019         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14020         (DEF_LIB_BUILTIN_CHKP): Likewise.
14021         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14022         (DEF_CHKP_BUILTIN): Likewise.
14023         (BUILT_IN_MEMCPY): Likewise.
14024         (BUILT_IN_MEMMOVE): Likewise.
14025         (BUILT_IN_MEMPCPY): Likewise.
14026         (BUILT_IN_MEMSET): Likewise.
14027         (BUILT_IN_STPCPY): Likewise.
14028         (BUILT_IN_STRCAT): Likewise.
14029         (BUILT_IN_STRCHR): Likewise.
14030         (BUILT_IN_STRCPY): Likewise.
14031         (BUILT_IN_STRLEN): Likewise.
14032         (BUILT_IN_MEMCPY_CHK): Likewise.
14033         (BUILT_IN_MEMMOVE_CHK): Likewise.
14034         (BUILT_IN_MEMPCPY_CHK): Likewise.
14035         (BUILT_IN_MEMSET_CHK): Likewise.
14036         (BUILT_IN_STPCPY_CHK): Likewise.
14037         (BUILT_IN_STRCAT_CHK): Likewise.
14038         (BUILT_IN_STRCPY_CHK): Likewise.
14039         * calls.c (store_bounds): Likewise.
14040         (emit_call_1): Likewise.
14041         (special_function_p): Likewise.
14042         (maybe_warn_nonstring_arg): Likewise.
14043         (initialize_argument_information): Likewise.
14044         (finalize_must_preallocate): Likewise.
14045         (compute_argument_addresses): Likewise.
14046         (expand_call): Likewise.
14047         * cfgexpand.c (expand_call_stmt): Likewise.
14048         (expand_return): Likewise.
14049         (expand_gimple_stmt_1): Likewise.
14050         (pass_expand::execute): Likewise.
14051         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14052         (cgraph_node::remove): Likewise.
14053         (cgraph_node::dump): Likewise.
14054         (cgraph_node::verify_node): Likewise.
14055         * cgraph.h (chkp_function_instrumented_p): Likewise.
14056         (symtab_node::get_alias_target): Likewise.
14057         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14058         (cgraph_local_p): Likewise.
14059         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14060         (cgraph_edge::rebuild_references): Likewise.
14061         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14062         (walk_polymorphic_call_targets): Likewise.
14063         (cgraph_node::expand_thunk): Likewise.
14064         (symbol_table::output_weakrefs): Likewise.
14065         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14066         (ix86_handle_option): Likewise.
14067         * config/i386/constraints.md: Likewise.
14068         * config/i386/i386-builtin-types.def (BND): Likewise.
14069         (VOID): Likewise.
14070         (PVOID): Likewise.
14071         (ULONG): Likewise.
14072         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14073         (BDESC_FIRST): Likewise.
14074         (BDESC): Likewise.
14075         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14076         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14077         * config/i386/i386.c (enum reg_class): Likewise.
14078         (ix86_target_string): Likewise.
14079         (ix86_option_override_internal): Likewise.
14080         (ix86_conditional_register_usage): Likewise.
14081         (ix86_valid_target_attribute_inner_p): Likewise.
14082         (ix86_set_indirect_branch_type): Likewise.
14083         (ix86_set_current_function): Likewise.
14084         (ix86_function_arg_regno_p): Likewise.
14085         (init_cumulative_args): Likewise.
14086         (ix86_function_arg_advance): Likewise.
14087         (ix86_function_arg): Likewise.
14088         (ix86_pass_by_reference): Likewise.
14089         (ix86_function_value_regno_p): Likewise.
14090         (ix86_function_value_1): Likewise.
14091         (ix86_function_value_bounds): Likewise.
14092         (ix86_return_in_memory): Likewise.
14093         (ix86_setup_incoming_vararg_bounds): Likewise.
14094         (ix86_va_start): Likewise.
14095         (indirect_thunk_need_prefix): Likewise.
14096         (print_reg): Likewise.
14097         (ix86_print_operand): Likewise.
14098         (ix86_expand_call): Likewise.
14099         (ix86_output_function_return): Likewise.
14100         (reg_encoded_number): Likewise.
14101         (BDESC_VERIFYS): Likewise.
14102         (ix86_init_mpx_builtins): Likewise.
14103         (ix86_init_builtins): Likewise.
14104         (ix86_emit_cmove): Likewise.
14105         (ix86_emit_move_max): Likewise.
14106         (ix86_expand_builtin): Likewise.
14107         (ix86_builtin_mpx_function): Likewise.
14108         (ix86_get_arg_address_for_bt): Likewise.
14109         (ix86_load_bounds): Likewise.
14110         (ix86_store_bounds): Likewise.
14111         (ix86_load_returned_bounds): Likewise.
14112         (ix86_store_returned_bounds): Likewise.
14113         (ix86_class_likely_spilled_p): Likewise.
14114         (ix86_hard_regno_mode_ok): Likewise.
14115         (x86_order_regs_for_local_alloc): Likewise.
14116         (ix86_mitigate_rop): Likewise.
14117         (ix86_bnd_prefixed_insn_p): Likewise.
14118         (ix86_mpx_bound_mode): Likewise.
14119         (ix86_make_bounds_constant): Likewise.
14120         (ix86_initialize_bounds): Likewise.
14121         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14122         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14123         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14124         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14125         (TARGET_CHKP_BOUND_MODE): Likewise.
14126         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14127         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14128         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14129         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14130         * config/i386/i386.h (TARGET_MPX): Likewise.
14131         (TARGET_MPX_P): Likewise.
14132         (VALID_BND_REG_MODE): Likewise.
14133         (FIRST_BND_REG): Likewise.
14134         (LAST_BND_REG): Likewise.
14135         (enum reg_class): Likewise.
14136         (BND_REG_P): Likewise.
14137         (BND_REGNO_P): Likewise.
14138         (BNDmode): Likewise.
14139         (ADJUST_INSN_LENGTH): Likewise.
14140         * config/i386/i386.md: Likewise.
14141         * config/i386/i386.opt: Likewise.
14142         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14143         (defined): Likewise.
14144         (LINK_MPX): Likewise.
14145         (MPX_SPEC): Likewise.
14146         (LIBMPX_SPEC): Likewise.
14147         (LIBMPXWRAPPERS_SPEC): Likewise.
14148         (CHKP_SPEC): Likewise.
14149         * config/i386/predicates.md: Likewise.
14150         * dbxout.c (dbxout_type): Likewise.
14151         * doc/extend.texi: Likewise.
14152         * doc/invoke.texi: Likewise.
14153         * doc/md.texi: Likewise.
14154         * doc/tm.texi: Likewise.
14155         * doc/tm.texi.in: Likewise.
14156         * dwarf2out.c (is_base_type): Likewise.
14157         (gen_formal_types_die): Likewise.
14158         (gen_subprogram_die): Likewise.
14159         (gen_type_die_with_usage): Likewise.
14160         (gen_decl_die): Likewise.
14161         (dwarf2out_late_global_decl): Likewise.
14162         * expr.c (expand_assignment): Likewise.
14163         (emit_storent_insn): Likewise.
14164         (store_expr_with_bounds): Likewise.
14165         (store_expr): Likewise.
14166         (expand_expr_real_1): Likewise.
14167         * expr.h (store_expr_with_bounds): Likewise.
14168         * function.c (use_register_for_decl): Likewise.
14169         (struct bounds_parm_data): Likewise.
14170         (assign_parms_augmented_arg_list): Likewise.
14171         (assign_parm_find_entry_rtl): Likewise.
14172         (assign_parm_is_stack_parm): Likewise.
14173         (assign_parm_load_bounds): Likewise.
14174         (assign_bounds): Likewise.
14175         (assign_parms): Likewise.
14176         (expand_function_start): Likewise.
14177         * gcc.c (CHKP_SPEC): Likewise.
14178         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14179         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14180         (wrestrict_dom_walker::check_call): Likewise.
14181         * gimple.c (gimple_build_call_from_tree): Likewise.
14182         * gimple.h (enum gf_mask): Likewise.
14183         (gimple_call_with_bounds_p): Likewise.
14184         (gimple_call_set_with_bounds): Likewise.
14185         * gimplify.c (gimplify_init_constructor): Likewise.
14186         * ipa-cp.c (initialize_node_lattices): Likewise.
14187         (propagate_constants_across_call): Likewise.
14188         (find_more_scalar_values_for_callers_subset): Likewise.
14189         * ipa-hsa.c (process_hsa_functions): Likewise.
14190         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14191         * ipa-icf.c (sem_function::merge): Likewise.
14192         * ipa-inline.c (early_inliner): Likewise.
14193         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14194         (warn_function_cold): Likewise.
14195         (propagate_pure_const): Likewise.
14196         * ipa-ref.h (enum GTY): Likewise.
14197         * ipa-split.c (find_retbnd): Likewise.
14198         (consider_split): Likewise.
14199         (split_function): Likewise.
14200         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14201         * ipa.c (walk_polymorphic_call_targets): Likewise.
14202         (symbol_table::remove_unreachable_nodes): Likewise.
14203         (process_references): Likewise.
14204         (cgraph_build_static_cdtor_1): Likewise.
14205         * lto-cgraph.c (lto_output_node): Likewise.
14206         (output_refs): Likewise.
14207         (compute_ltrans_boundary): Likewise.
14208         (input_overwrite_node): Likewise.
14209         (input_node): Likewise.
14210         (input_cgraph_1): Likewise.
14211         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14212         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14213         (class pass_chkp_instrumentation_passes): Likewise.
14214         (make_pass_chkp_instrumentation_passes): Likewise.
14215         * passes.def: Likewise.
14216         * rtl.h (struct GTY): Likewise.
14217         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14218         * stor-layout.c (layout_type): Likewise.
14219         * symtab.c: Likewise.
14220         * target.def: Likewise.
14221         * targhooks.c (default_chkp_bound_type): Likewise.
14222         (default_chkp_bound_mode): Likewise.
14223         (default_builtin_chkp_function): Likewise.
14224         (default_chkp_function_value_bounds): Likewise.
14225         (default_chkp_make_bounds_constant): Likewise.
14226         (default_chkp_initialize_bounds): Likewise.
14227         * targhooks.h (default_chkp_bound_type): Likewise.
14228         (default_chkp_bound_mode): Likewise.
14229         (default_builtin_chkp_function): Likewise.
14230         (default_chkp_function_value_bounds): Likewise.
14231         (default_chkp_make_bounds_constant): Likewise.
14232         (default_chkp_initialize_bounds): Likewise.
14233         * toplev.c (compile_file): Likewise.
14234         (process_options): Likewise.
14235         * tree-core.h (DEF_BUILTIN): Likewise.
14236         (DEF_BUILTIN_CHKP): Likewise.
14237         * tree-inline.c (declare_return_variable): Likewise.
14238         (remap_gimple_stmt): Likewise.
14239         (copy_bb): Likewise.
14240         (initialize_inlined_parameters): Likewise.
14241         (expand_call_inline): Likewise.
14242         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14243         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14244         (make_pass_ipa_chkp_produce_thunks): Likewise.
14245         (make_pass_chkp): Likewise.
14246         (make_pass_chkp_opt): Likewise.
14247         (make_pass_chkp_instrumentation_passes): Likewise.
14248         * tree-pretty-print.c (dump_generic_node): Likewise.
14249         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14250         * tree-ssa-dce.c (propagate_necessity): Likewise.
14251         (eliminate_unnecessary_stmts): Likewise.
14252         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14253         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14254         * tree-ssa-sccvn.h: Likewise.
14255         * tree-ssa-strlen.c (get_string_length): Likewise.
14256         (valid_builtin_call): Likewise.
14257         (adjust_last_stmt): Likewise.
14258         (handle_builtin_strchr): Likewise.
14259         (handle_builtin_strcpy): Likewise.
14260         (handle_builtin_stxncpy): Likewise.
14261         (handle_builtin_memcpy): Likewise.
14262         (handle_builtin_strcat): Likewise.
14263         (strlen_check_and_optimize_stmt): Likewise.
14264         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14265         * tree-streamer-in.c: Likewise.
14266         * tree-streamer.c (record_common_node): Likewise.
14267         * tree.c (tree_code_size): Likewise.
14268         (wide_int_to_tree_1): Likewise.
14269         (type_contains_placeholder_1): Likewise.
14270         (build_common_tree_nodes): Likewise.
14271         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14272         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14273         (POINTER_BOUNDS_P): Likewise.
14274         (BOUNDED_TYPE_P): Likewise.
14275         (BOUNDED_P): Likewise.
14276         (CALL_WITH_BOUNDS_P): Likewise.
14277         (pointer_bounds_type_node): Likewise.
14278         * value-prof.c (gimple_ic): Likewise.
14279         * var-tracking.c (vt_add_function_parameters): Likewise.
14280         * varasm.c (make_decl_rtl): Likewise.
14281         (assemble_start_function): Likewise.
14282         (output_constant): Likewise.
14283         (maybe_assemble_visibility): Likewise.
14284         * varpool.c (ctor_for_folding): Likewise.
14285         * chkp-builtins.def: Remove.
14286         * ipa-chkp.c: Remove.
14287         * ipa-chkp.h: Remove.
14288         * rtl-chkp.c: Remove.
14289         * rtl-chkp.h: Remove.
14290         * tree-chkp-opt.c: Remove.
14291         * tree-chkp.c: Remove.
14292         * tree-chkp.h: Remove.
14294 2018-06-07  Carl Love  <cel@us.ibm.com>
14296         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14297         vextract_fp_from_shortl): Add BE support.
14299 2018-06-07  Paul Koning  <ni1d@arrl.net>
14301         * compare-elim.c (try_merge_compare): Don't merge compare if
14302         address contains a side effect.
14303         (try_eliminate_compare): Likewise.
14305 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14307         * config.gcc: Support "tremont".
14308         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14309         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14310         PROCESSOR_TREMONT.
14311         * config/i386/i386.c (m_TREMONT): Define.
14312         (processor_target_table): Add "tremont".
14313         (PTA_TREMONT): Define.
14314         (ix86_lea_outperforms): Add TARGET_TREMONT.
14315         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14316         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14317         and M_INTEL_GOLDMONT_PLUS.
14318         (fold_builtin_cpu): Add "tremont".
14319         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14320         (ix86_option_override_internal): Add "tremont".
14321         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14322         (processor_type): Add PROCESSOR_TREMONT.
14323         * config/i386/x86-tune.def: Add m_TREMONT.
14324         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14326 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14328         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14329         symbol defined for msp430i* devices to be lower case.
14331 2018-06-07  Richard Biener  <rguenther@suse.de>
14333         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14334         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14335         Properly wrap signed arithmetic if overflow wraps.
14337 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14339         PR tree-optimization/69615
14340         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14341         of a cast from a same precision integral SSA_NAME in a bb dominated
14342         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14343         cast to utype if rhs2 has already a compatible type.
14345 2018-06-07  Richard Biener  <rguenther@suse.de>
14347         PR tree-optimization/85935
14348         * graphite-scop-detection.c (find_params_in_bb): Analyze
14349         condition operands with respect to the correct loop.  Assert
14350         the analysis doesn't fail.
14352 2018-06-04  Carl Love  <cel@us.ibm.com>
14354         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14355         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14356         as it is slightly cheaper.
14357         (first_match_or_eos_index_<mode>):
14358         Calculate index using natural element order.
14359         (first_match_index_<mode>):
14360         Calculate index using natural element order.
14361         (first_match_or_eos_index_<mode>):
14362         Calculate index using natural order.
14363         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14364         for BE and LE modes.
14365         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14366         P9V_BUILTIN_VCLZLSBB_V16QI.
14367         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14368         specific.
14370 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14372         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14373         indentation and line wrap for many prototypes.  Add missing
14374         @smallexample directives around block of prototypes for vec_xl and
14375         vec_xst.
14377 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14379         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14380         track if we pass or return IEEE 128-bit floating point.
14381         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14382         C++ mangling that is compatible with GCC 8.1.
14383         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14384         (init_cumulative_args): Note if we pass or return IEEE 128-bit
14385         floating point types.
14386         (rs6000_function_arg_advance_1): Likewise.
14387         (rs6000_mangle_type): Optionally generate mangled names that match
14388         what GCC 8.1 generated for IEEE 128-bit floating point types.
14389         (rs6000_globalize_decl_name): If we have an external function that
14390         passes or returns IEEE 128-bit types, generate a weak reference
14391         from the mangled name used in GCC 8.1 to the current mangled
14392         name.
14393         (rs6000_init_builtins): Make __ibm128 use the long double type if
14394         long double is IBM extended double.  Make __float128 use the long
14395         double type if long double is IEEE 128-bit.
14397         PR target/85657
14398         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14399         macro for __ibm128 built-in functions.
14400         (PACK_IF): Add __ibm128 pack/unpack functions.
14401         (UNPACK_IF): Likewise.
14402         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14403         enable long double built-in functions if long double is IEEE
14404         128-bit floating point.
14405         (rs6000_invalid_builtin): Update long double built-in function
14406         error message.
14407         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14408         functions, adjust the built-in function to use the long double
14409         built-in function if __ibm128 and long double are the same type.
14410         * doc/extend.texi (PowerPC builtins): Update documention for
14411         __builtin_{,un}pack_longdouble.  Add documentation for
14412         __builtin_{,un}pack_ibm128.
14414 2018-06-06  Jim Wilson  <jimw@sifive.com>
14416         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14417         (struct machine_function): New field interrupt_mode.
14418         (riscv_handle_type_attribute): New function.  Add forward declaration.
14419         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14420         (riscv_expand_epilogue): Check interrupt_mode field.
14421         (riscv_set_current_function): Check interrupt attribute args and
14422         set interrupt_mode field.
14423         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14424         (riscv_sret, riscv_uret): New.
14425         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14426         new arguments to interrupt attribute.
14428 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
14430         PR target/63177
14431         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14432         Don't handle -mcpu=power8 if -mpower9-vector is also used.
14434 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14436         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14437         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14438         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14439         several redundant entries.
14441 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
14443         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14444         type from "rtx" to "rtx_insn *".
14445         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14446         for local "call_insn", removing cast.
14447         (ix86_expand_call): Likewise, introducing a "call_insn" local.
14449 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
14451         PR tree-optimization/86066
14452         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14453         for BIT_INSERT_EXPR stores.
14455 2018-06-06  Richard Biener  <rguenther@suse.de>
14457         PR tree-optimization/86062
14458         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14459         component refs ontop
14460         of to be offsetted base.
14462 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14464         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14465         to be static and remove check on interrupt attribute name.
14467 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14469         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14470         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14472 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
14474         PR target/79924
14475         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14476         second argument.
14477         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14478         Remove second argument, change how error is called.
14479         (aarch64_layout_arg): Remove second argument from
14480         aarch64_err_no_fpadvsimd call.
14481         (aarch64_init_cumulative_args): Ditto.
14482         (aarch64_gimplify_va_arg_expr): Ditto.
14483         * config/aarch64/aarch64.md (mov<mode>): Ditto.
14485 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
14487         * config/i386/i386.md (simple_return_indirect_internal): New expander.
14488         (*simple_return_indirect_internal<mode>): Rename from
14489         simple_return_indirect_internal.  Use W mode iterator.
14490         (rstorssp): New expander.
14491         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
14492         (clrssbsy): New expander.
14493         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
14495 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14497         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14498         __typeof__.
14499         (cmse_check_pointed_object): Likewise.
14501 2018-06-05  Martin Liska  <mliska@suse.cz>
14503         PR gcov-profile/47618
14504         * doc/invoke.texi: Document how -fprofile-dir format
14505         is extended.
14507 2018-06-05  Richard Biener  <rguenther@suse.de>
14509         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
14510         removal pretend DOM info isn't available so we do not update
14511         it and only remove edges, not dominated blocks.  Actually free
14512         DOM info in case we removed something.  Remove unreachable blocks.
14513         (mfb_keep_latches): Work with either DOM info or marked backedges.
14514         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
14515         first.  Mark backedges if DOM info isn't available.
14516         (Re-)compute DOM info after cleanup_control_flow_pre.
14518 2018-06-05  Richard Biener  <rguenther@suse.de>
14520         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
14521         (locus_discrim_hasher::hash): Adjust.
14522         (locus_discrim_hasher::equal): Likewise.
14523         (next_discriminator_for_locus): Work on line directly.
14524         (same_line_p): Pass in expanded locus1 as well.
14525         (assign_discriminators): Avoid redundant location expansions.
14527 2018-06-05  Richard Biener  <rguenther@suse.de>
14529         PR tree-optimization/86046
14530         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
14531         if required after clearing TREE_ADDRESSABLE.
14533 2018-06-05  Richard Biener  <rguenther@suse.de>
14535         PR tree-optimization/86047
14536         * tree-ssa-loop.c (for_each_index): Glob handling of all
14537         decls and constants and really handle all of them.
14539 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14541         PR target/81497
14542         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
14543         qualifier_void_pointer and qualifier_const_void_pointer.
14544         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
14545         (arm_init_builtins): Handle the above.
14546         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
14547         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
14548         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
14549         void intrinsics.
14551 2018-06-05  Martin Liska  <mliska@suse.cz>
14553         * auto-profile.c (read_autofdo_file): Do not use
14554         gcov_ctr_summary struct.
14555         (afdo_callsite_hot_enough_for_early_inline): Likewise.
14556         * coverage.c (struct counts_entry): Likewise.
14557         (read_counts_file): Read just single summary entry.
14558         (get_coverage_counts): Use gcov_summary struct.
14559         * coverage.h (get_coverage_counts): Likewise.
14560         * gcov-dump.c (dump_working_sets): Likewise.
14561         (tag_summary): Dump just single summary.
14562         * gcov-io.c (gcov_write_summary): Write just histogram
14563         summary.
14564         (gcov_read_summary): Read just single summary.
14565         (compute_working_sets): Use gcov_summary struct.
14566         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
14567         of GCOV_COUNTERS_SUMMABLE.
14568         (GCOV_COUNTERS_SUMMABLE): Remove.
14569         (GCOV_FIRST_VALUE_COUNTER): Replace with
14570         GCOV_COUNTER_V_INTERVAL.
14571         (struct gcov_ctr_summary): Remove.
14572         (struct gcov_summary): Directly use fields of former
14573         gcov_ctr_summary.
14574         (compute_working_sets): Use gcov_summary struct.
14575         * gcov.c (read_count_file): Do not use ctrs fields.
14576         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
14577         struct.
14578         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
14579         struct.
14580         * profile.c: Likewise.
14581         * profile.h: Likewise.
14583 2018-06-05  Martin Liska  <mliska@suse.cz>
14585         PR gcov-profile/84846
14586         * gcov.c (output_lines): Print working directory only
14587         in intermediate format.
14589 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
14591         * config/s390/s390-builtin-types.def: Add void function type.
14592         * config/s390/s390-builtins.def: Use the function type for the
14593         tbeginc builtin.
14595 2018-06-04  Jim Wilson  <jimw@sifive.com>
14597         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
14598         to int.
14599         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
14600         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
14601         handle EH_RETURN_DATA_REGNO registers properly.
14602         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
14603         (riscv_expand_epilogue): Update comment.  Change argument name and
14604         type.  Update code to use new name and type.  Pass new args to
14605         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
14606         EXCEPTION_RETURN.
14607         * config/riscv/riscv.md (NORMAL_RETURN): New.
14608         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
14609         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
14610         (eh_return): Call gen_eh_return_internal and emit barrier.
14611         (eh_return_internal): Call riscv_expand_epilogue.
14613 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
14615         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
14616         bit_insertion field and declare can_be_merged_into method.
14617         (merged_store_group::can_be_merged_into): New method.
14618         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
14619         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
14620         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
14622 2018-06-04  Richard Biener  <rguenther@suse.de>
14624         PR tree-optimization/85955
14625         * builtins.c (fold_builtin_sincos): Convert pointers to
14626         destination to appropriate type before dereferencing.
14628 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
14630         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
14632 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
14634         * expr.c (expand_expr_real_1): Force the operand into memory if
14635         its TYPE_MODE is BLKmode and if there is no integer mode for
14636         the number of bits being extracted.
14638 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14640         PR target/85832
14641         PR target/86036
14642         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14643         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
14645 2018-06-04  Richard Biener  <rguenther@suse.de>
14647         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
14648         (cleanup_tree_cfg_noloop): ... single caller.  Do
14649         start_recording_case_labels later.
14651 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
14653         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
14654         to _IMMINTRIN_H_INCLUDED.
14655         * config/i386/pconfigintrin.h: Ditto.
14656         * config/i386/waitpkgintrin.h: Ditto.
14657         * config/i386/immintrin.h: Add includes for sgxintrin.h,
14658         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
14659         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
14660         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
14661         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
14662         waitpkgintrin.h and cldemoteintrin.h.
14664 2018-06-04  Richard Biener  <rguenther@suse.de>
14666         PR tree-optimization/86038
14667         * tracer.c (find_best_successor): Check probability for
14668         being initialized, bail out if not.
14670 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
14672         PR target/86003
14673         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
14674         of bits to ignore when comparing architectures.
14676 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14678         PR tree-optimization/69615
14679         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
14680         maximum or minimum of the type, try to merge it also as if
14681         range1 is + [-, x - 1] or + [x + 1, -].
14683         PR c++/86025
14684         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
14686 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
14688         PR tree-optimization/86034
14689         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
14690         the unsigned bitfield type in a bit insertion sequence if it does not
14691         have a larger precision than the bitfield size.
14692         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
14694 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
14696         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
14698 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14700         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
14701         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
14702         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
14703         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
14705 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14707         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
14708         Disable -fdelete-null-pointer-checks for ELF toolchain.
14710 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14711             Kito Cheng  <kito.cheng@gmail.com>
14713         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
14714         (nds32le-*-*, nds32be-*-*): Integrate checking process.
14715         (nds32*-*-*): Add glibc and uclibc conditions.
14716         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
14717         (TARGET_EXCEPT_UNWIND_INFO): Define.
14718         * config/nds32/elf.h: New file.
14719         * config/nds32/linux.h: New file.
14720         * config/nds32/nds32-elf.opt: New file.
14721         * config/nds32/nds32-linux.opt: New file.
14722         * config/nds32/nds32-fp-as-gp.c
14723         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
14724         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
14725         TARGET_LINUX_ABI.
14726         (nds32_asm_file_end): Ditto.
14727         (nds32_print_operand): Ditto.
14728         (nds32_insert_attributes): Ditto.
14729         (nds32_init_libfuncs): New function.
14730         (TARGET_HAVE_TLS): Define.
14731         (TARGET_INIT_LIBFUNCS): Define.
14732         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
14733         spec content.
14734         (TARGET_ELF): Apply different mcmodel setting.
14735         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
14736         been migrated into elf.h and linux.h files.
14737         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
14738         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
14739         (mcmodel): The content has been migrated into nds32-elf.opt and
14740         nds32-linux.opt files.
14741         * config/nds32/t-elf: New file.
14742         * config/nds32/t-linux: New file.
14744 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14745             Shiva Chen  <shiva0217@gmail.com>
14747         * config/nds32/constants.md (unspec_volatile_element): Add
14748         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
14749         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
14750         optimization.
14751         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
14752         (make_pass_nds32_fp_as_gp): Declare.
14753         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
14754         optmization pass.
14755         (nds32_asm_function_end_prologue): Remove unused asm output.
14756         (nds32_asm_function_begin_epilogue): Remove unused asm output.
14757         (nds32_asm_file_start): Output necessary fp_as_gp information.
14758         (nds32_option_override): Adjust register usage.
14759         (nds32_expand_prologue): Consider fp_as_gp situation.
14760         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
14761         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
14762         (epilogue): Ditto.
14763         (return): Ditto.
14764         (simple_return): Ditto.
14765         (omit_fp_begin): Output special directive for fp_as_gp.
14766         (omit_fp_end): Output special directive for fp_as_gp.
14767         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
14768         mforbid-fp-as-gp): New options.
14770 2018-06-01  Mark Wielaard  <mark@klomp.org>
14772         * dwarf2out.c (dwarf2out_finish): Remove generation of
14773         DW_AT_loclists_base.
14775 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
14777         * gimple-ssa-store-merging.c: Include gimple-fold.h.
14778         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
14779         (struct merged_store_group): Add bit_insertion field.
14780         (dump_char_array): Use standard hexadecimal format.
14781         (merged_store_group::merged_store_group): Set bit_insertion to false.
14782         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
14783         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
14784         also print the mask in the dump file.
14785         (pass_store_merging::gate): Minor tweak.
14786         (imm_store_chain_info::coalesce_immediate): Fix wrong association
14787         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
14788         stores with INTEGER_CST stores.
14789         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14790         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14791         and use it throughout.  Generate bit insertion sequences if need be.
14792         (pass_store_merging::process_store): Remove redundant condition.
14793         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14795 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
14797         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14798         the 128-bit floating point types.  Fix function comment.
14800 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14802         * config/aarch64/aarch64-simd.md
14803         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14804         mnemonics.
14805         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14806         mnemonics.
14808 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14810         PR tree-optimization/85989
14811         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14812         variable.
14813         (backprop::intersect_uses): Check it when deciding whether this
14814         is a backedge reference.
14815         (backprop::process_block): Add each phi to m_visited_phis
14816         after visiting it, then clear it at the end.
14818 2018-06-01  Richard Biener  <rguenther@suse.de>
14820         * tree-vectorizer.h (vect_dr_stmt): New function.
14821         (vect_get_load_cost): Adjust.
14822         (vect_get_store_cost): Likewise.
14823         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14824         Use vect_dr_stmt instead of DR_SMTT.
14825         (vect_record_base_alignments): Likewise.
14826         (vect_calculate_target_alignment): Likewise.
14827         (vect_compute_data_ref_alignment): Likewise and make static.
14828         (vect_update_misalignment_for_peel): Likewise.
14829         (vect_verify_datarefs_alignment): Likewise.
14830         (vector_alignment_reachable_p): Likewise.
14831         (vect_get_data_access_cost): Likewise.  Pass down
14832         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14833         (vect_get_peeling_costs_all_drs): Likewise.
14834         (vect_peeling_hash_get_lowest_cost): Likewise.
14835         (vect_enhance_data_refs_alignment): Likewise.
14836         (vect_find_same_alignment_drs): Likewise.
14837         (vect_analyze_data_refs_alignment): Likewise.
14838         (vect_analyze_group_access_1): Likewise.
14839         (vect_analyze_group_access): Likewise.
14840         (vect_analyze_data_ref_access): Likewise.
14841         (vect_analyze_data_ref_accesses): Likewise.
14842         (vect_vfa_segment_size): Likewise.
14843         (vect_small_gap_p): Likewise.
14844         (vectorizable_with_step_bound_p): Likewise.
14845         (vect_prune_runtime_alias_test_list): Likewise.
14846         (vect_analyze_data_refs): Likewise.
14847         (vect_supportable_dr_alignment): Likewise.
14848         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14849         (vect_gen_prolog_loop_niters): Likewise.
14850         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14851         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14852         modify DR_STMT.
14853         (vect_recog_mask_conversion_pattern): Likewise.
14854         (vect_try_gather_scatter_pattern): Likewise.
14855         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14856         to vect_get_store_cost.
14857         (vect_get_store_cost): Get stmt_info instead of DR.
14858         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14859         (vect_get_load_cost): Get stmt_info instead of DR.
14861 2018-06-01  Richard Biener  <rguenther@suse.de>
14863         PR middle-end/86017
14864         * gimple-fold.c (var_decl_component_p): Also allow offsetted
14865         vars wrapped in MEM_REFs.
14867 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14869         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14870         Fix subreg tests so that we only return a choice between
14871         GENERAL_REGS and FP_REGS if the original classes included both.
14873 2018-06-01  Richard Biener  <rguenther@suse.de>
14875         PR ipa/85960
14876         * tree-ssa-structalias.c (get_function_part_constraint):
14877         Handle NULL fi->decl.
14878         (find_func_aliases_for_call): Properly handle indirect
14879         fi from direct call.
14880         (find_func_clobbers): Likewise.
14881         (ipa_pta_execute): Likewise.
14882         (create_variable_info_for): For functions that are ifunc_resolver
14883         resolve to a varinfo that contains the result of the resolver call.
14884         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14885         aliases.
14887 2018-05-31  Michael Collison  <michael.collison@arm.com>
14889         * config/aarch64/aarch64.md:
14890         (*fix_to_zero_extenddfdi2): New pattern.
14891         * gcc.target/aarch64/fix_extend1.c: New testcase.
14893 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
14895         PR middle-end/78809
14896         PR middle-end/83026
14897         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14898         and BUILT_IN_STRNCMP_EQ.
14899         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14900         BUILT_IN_STRNCMP_EQ.
14901         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14902         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14903         (gimple_fold_builtin): Likewise.
14904         * tree-ssa-strlen.c (compute_string_length): New function.
14905         (determine_min_obsize): New function.
14906         (handle_builtin_string_cmp): New function to handle calls to
14907         string compare functions.
14908         (strlen_optimize_stmt): Add handling to builtin string compare
14909         calls.
14910         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14911         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14912         * tree.c (build_common_builtin_nodes): Add new defines of
14913         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14915 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14917         PR target/85984
14918         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14919         functions with naked attribute.
14921 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14923         * config/i386/sse.md (avx_vec_concat<mode>):
14924         Substitute concat_tg_mode mode attribute with xtg_mode.
14925         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14926         (concat_tg_mode): Remove mode attribute.
14928 2018-05-31  Martin Sebor  <msebor@redhat.com>
14930         PR c/82063
14931         * calls.c (alloc_max_size): Correct a logic error/typo.
14932         Treat excessive arguments as infinite.  Warn for invalid arguments.
14933         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14935 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14937         PR target/85829
14938         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14939         and movx for Haswell.
14941 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14942             Cesar Philippidis  <cesar@codesourcery.com>
14944         PR middle-end/85879
14945         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14946         when emitting error on private/firstprivate reductions.
14947         * omp-low.c (lower_omp_target): Avoid reference-type processing
14948         on pointers for firstprivate clause.
14950 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14952         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14953         (st1x2): Likewise.
14954         (st1x3): Likewise.
14955         * config/aarch64/aarch64-simd.md
14956         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14957         (aarch64_ld1_x3_<mode>): Likewise
14958         (aarch64_st1x2<VALLDIF:mode>): Likewise
14959         (aarch64_st1_x2_<mode>): Likewise
14960         (aarch64_st1x3<VALLDIF:mode>): Likewise
14961         (aarch64_st1_x3_<mode>): Likewise
14962         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14963         (vld1_s8_x3): Likewise.
14964         (vld1_u16_x3): Likewise.
14965         (vld1_s16_x3): Likewise.
14966         (vld1_u32_x3): Likewise.
14967         (vld1_s32_x3): Likewise.
14968         (vld1_u64_x3): Likewise.
14969         (vld1_s64_x3): Likewise.
14970         (vld1_f16_x3): Likewise.
14971         (vld1_f32_x3): Likewise.
14972         (vld1_f64_x3): Likewise.
14973         (vld1_p8_x3): Likewise.
14974         (vld1_p16_x3): Likewise.
14975         (vld1_p64_x3): Likewise.
14976         (vld1q_u8_x3): Likewise.
14977         (vld1q_s8_x3): Likewise.
14978         (vld1q_u16_x3): Likewise.
14979         (vld1q_s16_x3): Likewise.
14980         (vld1q_u32_x3): Likewise.
14981         (vld1q_s32_x3): Likewise.
14982         (vld1q_u64_x3): Likewise.
14983         (vld1q_s64_x3): Likewise.
14984         (vld1q_f16_x3): Likewise.
14985         (vld1q_f32_x3): Likewise.
14986         (vld1q_f64_x3): Likewise.
14987         (vld1q_p8_x3): Likewise.
14988         (vld1q_p16_x3): Likewise.
14989         (vld1q_p64_x3): Likewise.
14990         (vst1_s64_x2): Likewise.
14991         (vst1_u64_x2): Likewise.
14992         (vst1_f64_x2): Likewise.
14993         (vst1_s8_x2): Likewise.
14994         (vst1_p8_x2): Likewise.
14995         (vst1_s16_x2): Likewise.
14996         (vst1_p16_x2): Likewise.
14997         (vst1_s32_x2): Likewise.
14998         (vst1_u8_x2): Likewise.
14999         (vst1_u16_x2): Likewise.
15000         (vst1_u32_x2): Likewise.
15001         (vst1_f16_x2): Likewise.
15002         (vst1_f32_x2): Likewise.
15003         (vst1_p64_x2): Likewise.
15004         (vst1q_s8_x2): Likewise.
15005         (vst1q_p8_x2): Likewise.
15006         (vst1q_s16_x2): Likewise.
15007         (vst1q_p16_x2): Likewise.
15008         (vst1q_s32_x2): Likewise.
15009         (vst1q_s64_x2): Likewise.
15010         (vst1q_u8_x2): Likewise.
15011         (vst1q_u16_x2): Likewise.
15012         (vst1q_u32_x2): Likewise.
15013         (vst1q_u64_x2): Likewise.
15014         (vst1q_f16_x2): Likewise.
15015         (vst1q_f32_x2): Likewise.
15016         (vst1q_f64_x2): Likewise.
15017         (vst1q_p64_x2): Likewise.
15018         (vst1_s64_x3): Likewise.
15019         (vst1_u64_x3): Likewise.
15020         (vst1_f64_x3): Likewise.
15021         (vst1_s8_x3): Likewise.
15022         (vst1_p8_x3): Likewise.
15023         (vst1_s16_x3): Likewise.
15024         (vst1_p16_x3): Likewise.
15025         (vst1_s32_x3): Likewise.
15026         (vst1_u8_x3): Likewise.
15027         (vst1_u16_x3): Likewise.
15028         (vst1_u32_x3): Likewise.
15029         (vst1_f16_x3): Likewise.
15030         (vst1_f32_x3): Likewise.
15031         (vst1_p64_x3): Likewise.
15032         (vst1q_s8_x3): Likewise.
15033         (vst1q_p8_x3): Likewise.
15034         (vst1q_s16_x3): Likewise.
15035         (vst1q_p16_x3): Likewise.
15036         (vst1q_s32_x3): Likewise.
15037         (vst1q_s64_x3): Likewise.
15038         (vst1q_u8_x3): Likewise.
15039         (vst1q_u16_x3): Likewise.
15040         (vst1q_u32_x3): Likewise.
15041         (vst1q_u64_x3): Likewise.
15042         (vst1q_f16_x3): Likewise.
15043         (vst1q_f32_x3): Likewise.
15044         (vst1q_f64_x3): Likewise.
15045         (vst1q_p64_x3): Likewise.
15047 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15049         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15050         user_label_prefix to name.
15052         * tree-core.h: Update comment about the format of NAME string
15053         passed to handler in attribute_spec.
15055         * config/msp430/msp430.md: Remove erroneous subreg expression from
15056         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15057         zero_extend{q,h}isi2.
15059 2018-05-30  Borislav Petkov  <bp@suse.de>
15061         * doc/extend.texi: Document some architecture specific
15062         constraints and sort entries.
15064 2018-05-30  Martin Sebor  <msebor@redhat.com>
15066         PR middle-end/85369
15067         * builtins.c (expand_builtin_stpcpy_1): New function.
15068         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15069         only if the former succeeds.
15071 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15073         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15074         in saphira.
15076 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15078         * doc/invoke.texi (-flinker-output): Document
15080 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15082         * passes.c (ipa_write_summaries): Only modify statements if body
15083         is in memory.
15084         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15085         incrementally linking.
15086         (ipa_passes): Likewise.
15087         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15088         pass down resolution info.
15089         * common.opt (flag_incremental_link): Update info.
15090         * gcc.c (plugin specs): Turn flinker-output=* to
15091         -plugin-opt=-linker-output-known
15092         * toplev.c (compile_file): Also cut compilation when doing incremental
15093         link.
15094         * flag-types. (enum lto_partition_model): Add
15095         LTO_LINKER_OUTPUT_NOLTOREL.
15096         (invoke.texi): Add -flinker-output docs.
15097         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15098         link same way as WPA; do not stream in dead initializers.
15100         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15101         darf2out_register_external_decl): Support incremental link.
15103 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15105         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15107 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15109         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15110         it down to simple_object_copy_lto_debug_sections.
15111         (run_gcc): Determine incremental LTO link time and configure
15112         lto1 into non-wpa mode, disable renaming of debug sections.
15114 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15116         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15117         descriptions of various incorrectly documented functions.
15119 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15121         Revert:
15122         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15123         address check not strict.
15125 2018-05-30  Richard Biener  <rguenther@suse.de>
15127         PR tree-optimization/85964
15128         * tracer.c (better_p): Drop initialized count check, we only
15129         call the function with initialized counts now.
15130         (find_best_successor): Do find a best edge if one
15131         has uninitialized count.
15132         (find_best_predecessor): Likewise.  Do BB frequency check only
15133         if count is initialized.
15135 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15137         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15138         (aarch64_ldrstr_offset_compare): New.
15139         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15140         load/store orderings.
15141         (aarch64_gen_adjusted_ldpstp): Likewise.
15143 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15145         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15146         Check for subset of GENERAL_REGS and FP_REGS.
15147         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15148         r=w alternative.
15150 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15152         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15153         and wi::to_poly_offset.  Add the current offset and then check
15154         whether the sum fits, rather than using an unchecked addition of
15155         a checked term.  Check for a shwi rather than a uhwi.
15156         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15157         (store_constructor): Use poly_int_tree_p.
15158         (expand_expr_real_1): Likewise.
15159         * function.c (assign_temp): Likewise.
15160         * fold-const.c (const_binop): Use poly_int_tree_p and
15161         wi::to_poly_offset.
15162         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15163         division.
15164         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15165         to_poly_offset for MEM offsets.
15166         * ipa-icf.c (sem_variable::equals): Likewise.
15167         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15168         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15169         wi::to_poly_offset for BIT_FIELD_REF offsets.
15170         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15171         wi::to_poly_offset.
15172         * var-tracking.c (emit_note_insn_var_location): Use
15173         tree_to_poly_uint64.
15175 2018-05-29  Jim Wilson  <jimw@sifive.com>
15177         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15179 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15181         PR target/85950
15182         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15183         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15184         sequence.
15185         (sse4_1_round<mode>2): Use nonimmediate_operand
15186         for operand 1 predicate.
15188 2018-05-29  Martin Sebor  <msebor@redhat.com>
15189             Richard Biener  <rguenther@suse.de>
15191         PR testsuite/85888
15192         * calls.c (get_size_range): Call determine_value_range instead
15193         of get_value_range..
15194         * tree-vrp.h (determine_value_range): Declared new function.
15195         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15197 2018-05-29  Richard Biener  <rguenther@suse.de>
15199         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15200         sure to use non-pattern stmts for get_earlier_stmt arguments.
15201         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15202         called on pattern stmts.
15203         (get_later_stmt): Likewise.
15205 2018-05-29  Martin Liska  <mliska@suse.cz>
15207         PR gcov-profile/85759
15208         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15209         env variables.
15211 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15213         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15214         VEC_UNPACK_*_EXPR.
15215         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15216         VEC_PACK_*_EXPR.
15218         PR target/85918
15219         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15220         VEC_PACK_FLOAT_EXPR): New tree codes.
15221         * tree-pretty-print.c (op_code_prio): Handle
15222         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15223         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15224         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15225         * tree-inline.c (estimate_operator_cost): Likewise.
15226         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15227         * fold-const.c (const_binop): Likewise.
15228         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15229         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15230         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15231         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15232         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15233         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15234         * expr.c (expand_expr_real_2): Likewise.
15235         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15236         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15237         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15238         optabs.
15239         * optabs.c (expand_widen_pattern_expr): For
15240         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15241         sign from result type rather than operand's type.
15242         (expand_binop_directly): For vec_packu_float_optab and
15243         vec_packs_float_optab allow result type to be different from operand's
15244         type.
15245         * optabs-tree.c (optab_for_tree_code): Handle
15246         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15247         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15248         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15249         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15250         VEC_PACK_FLOAT_EXPR.
15251         * tree-vect-stmts.c (supportable_widening_operation): Handle
15252         FIX_TRUNC_EXPR.
15253         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15254         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15255         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15256         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15257         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15258         mode attributes.
15259         (vec_pack<floatprefix>_float_<mode>): New expander.
15260         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15261         attributes.
15262         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15263         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15264         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15265         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15266         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15267         Document.
15268         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15269         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15270         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15271         VEC_PACK_FLOAT_EXPR): Document.
15273 2018-05-29  Richard Biener  <rguenther@suse.de>
15275         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15276         member.
15277         (stmt_vec_info_vec): Make pointer.
15278         (init_stmt_vec_info_vec): Remove.
15279         (free_stmt_vec_info_vec): Likewise.
15280         (set_stmt_vec_info_vec): New function.
15281         (free_stmt_vec_infos): Likewise.
15282         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15283         (set_vinfo_for_stmt): Likewise.
15284         (get_earlier_stmt): Likewise.
15285         (get_later_stmt): Likewise.
15286         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15287         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15288         (vec_info::~vec_info): Free stmt_vec_infos.
15289         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15290         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15291         (pass_slp_vectorize::execute): Likewise.
15292         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15293         (free_stmt_vec_info_vec): Likewise.
15294         (set_stmt_vec_info_vec): New function.
15295         (free_stmt_vec_infos): Likewise.
15296         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15297         the global stmt_vec_info_vec.
15298         * tree-parloops.c (gather_scalar_reductions): Use
15299         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15300         vector.
15302 2018-05-29  Richard Biener  <rguenther@suse.de>
15304         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15306 2018-05-29  Martin Liska  <mliska@suse.cz>
15307             David Malcolm  <dmalcolm@redhat.com>
15309         * vec.c (test_reverse): New.
15310         (vec_c_tests): Add new test.
15311         * vec.h (vl_ptr>::reverse): New function.
15313 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15315         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15317         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15318         and later.
15320 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15322         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15324 2018-05-28  Richard Biener  <rguenther@suse.de>
15326         PR tree-optimization/85933
15327         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15328         look at stmts marked as vectorizable.
15330 2018-05-28  Richard Biener  <rguenther@suse.de>
15332         PR tree-optimization/85934
15333         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15334         vector boolean check before scalar optimization.
15336 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15338         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15339         for armv5te.
15341 2018-05-28  Mark Wielaard  <mark@klomp.org>
15343         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15344         if it is an expression containing a minus sign.
15346 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15348         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15350 2018-05-27  Paul Koning  <ni1d@arrl.net>
15352         * config/pdp11/pdp11.md (truncsihi2): Remove.
15354 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15355             Chung-Ju Wu  <jasonwucj@gmail.com>
15357         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15358         implementation.
15359         (unaligned_store_dw): Ditto.
15360         * config/nds32/nds32-memory-manipulation.c
15361         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15362         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15363         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15364         (emit_setmem_word_loop): Rename to ...
15365         (emit_setmem_doubleword_loop): ... this.
15366         (nds32_gen_dup_4_byte_to_word_value): New function.
15367         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15368         (nds32_expand_setmem_loop): Refine implementation.
15369         (nds32_expand_setmem_loop_v3m): Ditto.
15370         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15371         pattern.
15373 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15375         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15377 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15379         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15380         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15381         machine->attr_no_prologue_p.
15382         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15383         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15384         (nds32_expand_epilogue): Consider attr_naked_p.
15385         (nds32_expand_epilogue_v3pop): Likewise.
15386         (nds32_can_use_return_insn): Likewise.
15387         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15388         attr_no_prologue_p fields.
15389         * config/nds32/nds32.opt (mret-in-naked-func): New option.
15391 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
15393         PR target/85918
15394         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15395         attributes.
15396         * config/i386/sse.md
15397         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15398         Rename to ...
15399         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15400         ... this.
15401         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15402         Rename to ...
15403         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15404         ... this.
15405         (*<floatsuffix>floatv2div2sf2): Rename to ...
15406         (*float<floatunssuffix>v2div2sf2): ... this.
15407         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15408         (float<floatunssuffix>v2div2sf2_mask): ... this.
15409         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15410         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15411         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15412         to ...
15413         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15414         ... this.
15415         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15416         Rename to ...
15417         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15418         ... this.
15419         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15420         Rename to ...
15421         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15422         ... this.
15423         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15424         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15425         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15426         gen_ufix_truncv8dfv8si2.
15427         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15428         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15429         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15430         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15431         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15432         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15433         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15434         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15436 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15438         PR target/85900
15439         PR target/85345
15440         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15442 2018-05-25  Jim Wilson  <jimw@sifive.com>
15444         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15445         * config/riscv/riscv.c (struct machine_function): Add
15446         interrupt_handler_p and attribute_checked_p fields.
15447         (riscv_attribute_table): Add interrupt.
15448         (riscv_interrupt_type_p): New.
15449         (riscv_save_reg_p): Save extra regs for interrupt handler.
15450         (riscv_use_save_libcall): Return false  for interrupt handler.
15451         (riscv_first_stack_step): Add forward declaration.
15452         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
15453         for interrupt handler with large frame.  Use it for saved reg list.
15454         (riscv_expand_prologue): Move flag_stack_usage_info support to
15455         eliminate duplication.
15456         (riscv_expand_epilogue): Generate mret for interrupt handler.
15457         (riscv_epilogue_uses): New.
15458         (riscv_can_use_return_insn): Return false for interrupt handler.
15459         (riscv_function_ok_for_sibcall): Likewise.
15460         (riscv_set_current_function): Add interrupt handler support.
15461         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15462         * config/riscv/riscv.md (UNSPECV_MRET): New.
15463         (GP_REGNUM): New.
15464         (riscv_frflags, riscv_fsflags): Use tab after opcode.
15465         (riscv_mret): New.
15466         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15468 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
15470         PR tree-optimization/85712
15471         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15472         this candidate has already been replaced in-situ by a copy.
15474 2018-05-25  Jason Merrill  <jason@redhat.com>
15476         PR c++/80485 - inline function non-zero address.
15477         * symtab.c (nonzero_address): Check DECL_COMDAT.
15479 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
15481         PR target/83628
15482         * config/alpha/alpha.md (ashlsi3): New insn pattern.
15483         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
15484         extension of SImode operation.  Use const123_operand predicate.
15485         (*saddsi_1): Remove.
15486         (*saddl_se_1): Ditto.
15487         (*ssubsi_1): Ditto.
15488         (*ssubl_se_1): Ditto.
15489         * config/alpha/predicates.md (const123_operand): New predicate.
15490         * config/alpha/constraints.md (P): Use IN_RANGE.
15492 2018-05-25  Richard Biener  <rguenther@suse.de>
15494         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15495         defaulted to true.
15496         (ref_maybe_used_by_stmt_p): Likewise.
15497         (stmt_may_clobber_ref_p): Likewise.
15498         (stmt_may_clobber_ref_p_1): Likewise.
15499         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15500         and pass it along.
15501         (ref_maybe_used_by_stmt_p): Likewise.
15502         (stmt_may_clobber_ref_p): Likewise.
15503         (stmt_may_clobber_ref_p_1): Likewise.
15504         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15505         the alias oracle to disambiguate DRs with stmts DR analysis
15506         couldn't handle.
15507         (vect_analyze_data_refs): Do not give up on not analyzable
15508         DRs for BB vectorization.  Remove code truncating the dataref
15509         vector.
15511 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
15513         PR target/85832
15514         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15515         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
15516         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
15518 2018-05-25  Richard Biener  <rguenther@suse.de>
15520         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
15521         function, combining stmt data ref gathering and fatal analysis
15522         parts.
15523         (vect_analyze_data_refs): Remove now redudnant code and simplify.
15524         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
15525         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
15526         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
15527         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
15529 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15531         PR tree-optimization/85720
15532         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
15533         SCC if all partitions are builtins.
15534         (version_loop_by_alias_check): New parameter.  Generate cancelable
15535         runtime alias check if all partitions are builtins.
15536         (distribute_loop): Update call to above function.
15538 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15540         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
15541         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
15542         (parm_default_def_partition_arg): Ditto.
15543         (set_parm_default_def_partition): Ditto.
15544         (get_parm_default_def_partitions): Ditto and make it static.
15545         (get_undefined_value_partitions): Ditto and make it static.
15546         (remove_ssa_form): Refactor call to init_var_map here.
15547         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
15548         computation for loop region.
15549         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
15550         (register_default_def): Delete.
15551         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15552         (parm_default_def_partition_arg): Ditto.
15553         (set_parm_default_def_partition): Ditto.
15554         (get_parm_default_def_partitions): Ditto and make it static.
15555         (get_undefined_value_partitions): Ditto and make it static.
15556         (coalesce_with_default, coalesce_with_default): Update comment.
15557         (create_coalesce_list_for_region): New func factored out from
15558         create_outofssa_var_map.
15559         (populate_coalesce_list_for_outofssa): New func factored out from
15560         create_outofssa_var_map and coalesce_ssa_name.
15561         (create_outofssa_var_map): Delete.
15562         (coalesce_ssa_name): Refactor to support live range computation.
15563         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
15564         (get_parm_default_def_partitions): Delete.
15565         (get_undefined_value_partitions): Ditto.
15566         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
15567         computation for loop region.
15568         (new_tree_live_info, loe_visit_block): Ditto.
15569         (live_worklist, set_var_live_on_entry): Ditto.
15570         (calculate_live_on_exit, verify_live_on_entry): Ditto.
15571         * tree-ssa-live.h (struct _var_map): New fields.
15572         (init_var_map): Change decl.
15573         (region_contains_p): New.
15575 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15577         * tree-ssa-live.h (live_merge_and_clear): Delete.
15579 2018-05-25  Richard Biener  <rguenther@suse.de>
15581         PR c++/85912
15582         * tree-dump.c (dequeue_and_dump): Remove access to removed
15583         operand 2 of a SWITCH_EXPR.
15585 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15587         * doc/sourcebuild.texi (vect_double_cond_arith): Include
15588         multiplication and division.
15589         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
15590         (cond_udiv@var{m}, cond_umod@var{m}): Document.
15591         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
15592         (cond_udiv_optab, cond_umod_optab): New optabs.
15593         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
15594         (IFN_COND_RDIV): New internal functions.
15595         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
15596         TRUNC_MOD_EXPR and RDIV_EXPR.
15597         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
15598         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
15599         New unspecs.
15600         (SVE_INT_BINARY): Include mult.
15601         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
15602         (optab, sve_int_op): Handle mult.
15603         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
15604         UNSPEC_COND_DIV.
15605         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
15606         for SVE_INT_BINARY_SD.
15608 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15610         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
15611         (optab, sve_int_op): Handle div and udiv.
15612         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
15613         for SVE_INT_BINARY_SD.
15614         (*<optab><mode>3): New insn for the same.
15616 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15618         * tree-vect-patterns.c: Include predict.h.
15619         (vect_recog_divmod_pattern): Restrict check for division support
15620         to when optimizing for size.
15622 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15624         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
15625         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
15626         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
15627         (gimple_match_op::set_op): Likewise.
15628         (gimple_resimplify4): Declare.
15629         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
15630         (expr::gen_transform): Likewise.
15631         (decision_tree::gen): Generate a simplification routine for 4 operands.
15632         * gimple-match-head.c (gimple_simplify): Add an overload for
15633         4 operands.  In the top-level function, handle up to 4 call
15634         arguments and call gimple_resimplify4.
15635         (gimple_resimplify4): New function.
15636         (build_call_internal): Pass a fourth operand.
15637         (maybe_push_to_seq): Likewise.
15638         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
15639         Fold VEC_COND_EXPRs of an operation and a default value into
15640         an IFN_COND_* function if possible.
15641         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
15642         New unspecs.
15643         (SVE_COND_FP_BINARY): Include them.
15644         (optab, sve_fp_op): Handle them.
15645         (SVE_INT_BINARY_REV): New code iterator.
15646         (SVE_COND_FP_BINARY_REV): New int iterator.
15647         (commutative): New int attribute.
15648         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
15649         Declare.
15650         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
15651         function.
15652         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
15653         (*cond_<optab><mode>): New patterns for reversed operands.
15655 2018-05-25  Richard Biener  <rguenther@suse.de>
15657         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
15658         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
15659         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
15660         (STMT_VINFO_GROUPED_ACCESS): Adjust.
15661         * tree-vect-data-refs.c (everywhere): Adjust users.
15662         * tree-vect-loop.c (everywhere): Likewise.
15663         * tree-vect-slp.c (everywhere): Likewise.
15664         * tree-vect-stmts.c (everywhere): Likewise.
15665         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15667 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15669         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
15670         Rename to...
15671         (gcc_cv_as_section_exclude): ... this.
15672         Try Solaris as #exclude syntax.
15673         * configure: Regenerate.
15674         * config.in: Regenerate.
15675         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
15676         SECTION_EXCLUDE.
15677         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
15678         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
15680         * varasm.c (default_elf_asm_named_section): Don't check if
15681         HAVE_GAS_SECTION_EXCLUDE is defined.
15683 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15685         * doc/md.texi: Update the documentation of the cond_* optabs
15686         to mention the new final operand.  Fix GET_MODE_NUNITS call.
15687         Describe the scalar case too.
15688         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
15689         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
15690         instead of 2.
15691         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
15692         (get_conditional_internal_fn): Update comment.
15693         * tree-vect-loop.c (vectorizable_reduction): Pass the original
15694         accumulator value as a final argument to conditional functions.
15695         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
15696         a define_expand and add an "else" operand.  Assert for now that
15697         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
15698         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
15699         (*cond_<optab><mode>): New patterns.
15700         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
15701         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15702         (UNSPEC_COND_EOR): Delete.
15703         (optab): Remove associated mappings.
15704         (SVE_INT_BINARY): New code iterator.
15705         (sve_int_op): Remove int attribute and add "minus" to the code
15706         attribute.
15707         (SVE_COND_INT_OP): Delete.
15708         (SVE_COND_FP_OP): Rename to...
15709         (SVE_COND_FP_BINARY): ...this.
15711 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15713         * optabs.c (can_reuse_operands_p): New function.
15714         (maybe_legitimize_operands): Try to reuse the results for
15715         earlier operands.
15717 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15719         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
15720         Add {q} suffix to insn mnemonic.
15722 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15724         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
15725         (msp430_warn_func_return): New.
15727 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
15729         * fold-const.c (tree_nonzero_bits): New function.
15730         * fold-const.h (tree_nonzero_bits): Likewise.
15731         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
15732         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
15734 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15736         PR target/85900
15737         PR target/85345
15738         * varasm.c (assemble_alias): Check ifunc_resolver only on
15739         FUNCTION_DECL.
15741 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15743         PR target/85903
15744         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
15745         when memory input operand is handled.
15747 2018-05-24  Luis Machado  <luis.machado@linaro.org>
15749         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
15750         global.
15751         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
15753 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15755         * match.pd: Delay FMA folds until after vectorization.
15757 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15759         PR target/83009
15760         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15761         address check not strict.
15763 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15765         * gimple-match.h (gimple_match_op): New class.
15766         (mprts_hook): Replace parameters with a gimple_match_op *.
15767         (maybe_build_generic_op): Likewise.
15768         (gimple_simplified_result_is_gimple_val): Replace parameters with
15769         a const gimple_match_op *.
15770         (gimple_simplify): Replace code_helper * and tree * parameters with
15771         a gimple_match_op * parameter.
15772         (gimple_resimplify1): Replace code_helper *, tree and tree *
15773         parameters with a gimple_match_op * parameter.
15774         (gimple_resimplify2): Likewise.
15775         (gimple_resimplify3): Likewise.
15776         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
15777         parameters with a gimple_match_op * parameter.
15778         * gimple-match-head.c (gimple_simplify): Change prototypes of
15779         auto-generated functions to take a gimple_match_op * instead of
15780         separate code_helper * and tree * parameters.  Make the same
15781         change in the top-level overload and update calls to the
15782         gimple_resimplify routines.  Update calls to the auto-generated
15783         functions and to maybe_push_res_to_seq in the publicly-facing
15784         operation-specific gimple_simplify overloads.
15785         (gimple_match_op::MAX_NUM_OPS): Define.
15786         (gimple_resimplify1): Replace rcode and ops with a single res_op
15787         parameter.  Update call to gimple_simplify.
15788         (gimple_resimplify2): Likewise.
15789         (gimple_resimplify3): Likewise.
15790         (mprts_hook): Replace parameters with a gimple_match_op *.
15791         (maybe_build_generic_op): Likewise.
15792         (build_call_internal): Replace type, nargs and ops with
15793         a gimple_match_op *.
15794         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15795         with a single gimple_match_op *.  Update calls to mprts_hook,
15796         build_call_internal and gimple_simplified_result_is_gimple_val.
15797         Factor out code that is common to the tree_code and combined_fn cases.
15798         * genmatch.c (expr::gen_transform): Replace tem_code and
15799         tem_ops with a gimple_match_op called tem_op.  Update calls
15800         to the gimple_resimplify functions and maybe_push_res_to_seq.
15801         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15802         res_ops.  Update call to the gimple_resimplify functions.
15803         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15804         (decision_tree::gen): Make the functions take a gimple_match_op *
15805         called res_op instead of separate res_code and res_ops parameters.
15806         Update call accordingly.
15807         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15808         and ops with a single res_op parameter.  Update calls to
15809         maybe_build_generic_op and maybe_push_res_to_seq.
15810         (fold_stmt_1): Update calls to gimple_simplify and
15811         replace_stmt_with_simplification.
15812         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15813         and gimple_simplified_result_is_gimple_val.
15814         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15815         gimple_simplify.
15816         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15817         with a gimple_match_op *.
15818         (vn_nary_build_or_lookup): Likewise.  Update call to
15819         vn_nary_build_or_lookup_1.
15820         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15821         gimple_match_op *.  Update calls to the gimple_resimplify routines
15822         and to gimple_simplified_result_is_gimple_val.
15823         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15824         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15825         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15826         (visit_nary_op): Likewise.
15827         (visit_reference_op_load): Likewise.
15829 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15831         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15832         modifier for printing the step amount.
15834 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
15836         PR target/78849
15837         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15838         types.
15840 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
15842         * doc/sourcebuild.texi (Endianness): New subsubsection.
15844 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15846         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15847         <prefetch_dynamic_strides>: New const bool field.
15848         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15849         prefetch_dynamic_strides.
15850         (exynosm1_prefetch_tune): Likewise.
15851         (thunderxt88_prefetch_tune): Likewise.
15852         (thunderx_prefetch_tune): Likewise.
15853         (thunderx2t99_prefetch_tune): Likewise.
15854         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15855         false.
15856         (aarch64_override_options_internal): Update to set
15857         PARAM_PREFETCH_DYNAMIC_STRIDES.
15858         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15859         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15860         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15861         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15862         prefetch-dynamic-strides setting.
15864 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15866         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15867         <minimum_stride>: New const int field.
15868         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15869         minimum_stride field defaulting to -1.
15870         (exynosm1_prefetch_tune): Likewise.
15871         (thunderxt88_prefetch_tune): Likewise.
15872         (thunderx_prefetch_tune): Likewise.
15873         (thunderx2t99_prefetch_tune): Likewise.
15874         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15875         <default_opt_level>: Set to 3.
15876         (aarch64_override_options_internal): Update to set
15877         PARAM_PREFETCH_MINIMUM_STRIDE.
15878         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15879         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15880         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15881         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15882         stride is constant and is below the minimum stride threshold.
15884 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15886         * config/arm/arm-cpus.in (mode26): Delete.
15887         (armv4): Delete mode26 reference.
15888         * config/arm/arm.c (arm_configure_build_target): Delete use of
15889         isa_bit_mode26.
15891 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15893         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15894         New insn pattern.
15895         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15896         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15897         for non-SSE modes.
15898         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15899         (floatunsdidf2): Ditto.
15901 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15903         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15904         (fixuns_trunc<mode>si2_avx512f): Ditto.
15905         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15906         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15907         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15909 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15911         PR rtl-optimization/79985
15912         * df-scan.c (df_insn_refs_collect): Remove special case for
15913         global registers and asm statements.
15915 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15917         * extend.texi (Global Register Variables): Rewrite the bullet list.
15918         Note that the register is available for allocation. Note that access
15919         via inline asm must use constraints. Add note about async-signal
15920         handlers. Remove paragraph about automagic register selection.
15922 2018-05-23  Richard Biener  <rguenther@suse.de>
15924         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15925         of fixed offset from memset VN.
15927 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15929         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15930         first_interp field.
15931         (alloc_cand_and_find_basis): Initialize first_interp field.
15932         (slsr_process_mul): Modify first_interp field.
15933         (slsr_process_add): Likewise.
15934         (slsr_process_cast): Modify first_interp field for each new
15935         interpretation.
15936         (slsr_process_copy): Likewise.
15937         (dump_candidate): Dump first_interp field.
15938         (replace_mult_candidate): Process all interpretations, not just
15939         subsequent ones.
15940         (replace_rhs_if_not_dup): Likewise.
15941         (replace_one_candidate): Likewise.
15943 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
15945         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15946         Add new boolean.
15947         (aarch64_needs_frame_chain): New function.
15948         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15950 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
15952         PR target/84882
15953         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15954         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15955         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15956         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15957         as true for strict-align.
15958         (aarch64_can_inline_p): Perform checks even when callee has no
15959         attributes to check for strict alignment.
15960         * doc/extend.texi (AArch64 Function Attributes): Document
15961         no-strict-align.
15962         * doc/invoke.texi: (AArch64 Options): Likewise.
15964 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
15966         PR tree-optimization/85853
15967         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15968         the handling of the root of the node to...
15969         (vect_slp_analyze_node_operations_1): ...this new function,
15970         and run the whole thing with the child nodes' def types
15971         set according to their SLP node's def type.
15973 2018-05-23  Richard Biener  <rguenther@suse.de>
15975         PR middle-end/85874
15976         * tree-data-ref.c (create_runtime_alias_checks): Defer
15977         and ignore overflow warnings.
15979 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
15981         PR tree-optimization/85822
15982         * tree-vrp.c (is_masked_range_test): Fix handling of negative
15983         constants.
15985 2018-05-23  Richard Biener  <rguenther@suse.de>
15987         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15988         memset constants via native_interpret_expr.
15990 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
15992         PR target/85345
15993         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15994         attribute.
15995         (cgraph_node::create_alias): Likewise.
15996         (cgraph_node::get_availability): Check ifunc_resolver instead
15997         of looking up ifunc attribute.
15998         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15999         * varasm.c (do_assemble_alias): Likewise.
16000         (assemble_alias): Likewise.
16001         (default_binds_local_p_3): Likewise.
16002         * cgraph.h (cgraph_node): Add ifunc_resolver.
16003         (cgraph_node::only_called_directly_or_aliased_p): Return false
16004         for IFUNC resolver.
16005         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16006         attribute.
16007         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16008         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16009         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16010         instead of looking up ifunc attribute.
16012 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16014         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16016 2018-05-22  Martin Sebor  <msebor@redhat.com>
16018         PR middle-end/85359
16019         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16020         only when expasion succeeds.
16021         (expand_builtin_strcmp): Same.
16022         (expand_builtin_strncmp): Same.
16024 2018-05-22  Martin Sebor  <msebor@redhat.com>
16026         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16028 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16029             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16031         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16032         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16033         code.
16034         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16035         Add check for legitimate_address.
16036         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16037         (aarch64_swap_ldrstr_operands): New.
16038         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16039         Define prototype.
16041 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16042             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16044         * config/aarch64/aarch64.md: New patterns to generate stp
16045         and ldp.
16046         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16047         single words and double words.
16048         (load_pair_sw, load_pair_dw): Likewise.
16049         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16050         Delete.
16051         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16052         Delete.
16053         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16054         for different mode ldpstp and add peephole for merged zero stores.
16055         Likewise for loads.
16056         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16057         Add size check.
16058         (aarch64_gen_store_pair): Rename calls to match new patterns.
16059         (aarch64_gen_load_pair): Rename calls to match new patterns.
16060         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16061         (load_pair<DREG:mode><DREG2:mode>): ... This.
16062         (store_pair<mode>): Rename to...
16063         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16064         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16065         New mode iterators.
16066         (V_INT_EQUIV): Handle SImode.
16067         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16068         New predicate.
16070 2018-05-22  Martin Sebor  <msebor@redhat.com>
16072         PR c/85623
16073         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16074         or ajust the presumed bound on an operation to avoid unnecessary
16075         warnings.
16077 2018-05-22  Martin Sebor  <msebor@redhat.com>
16079         PR tree-optimization/85826
16080         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16081         assuming that a DECL necesarily has a constant size.
16083 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16085         PR middle-end/85862
16086         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16088 2018-05-22  Richard Biener  <rguenther@suse.de>
16090         PR tree-optimization/85834
16091         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16092         non-constant and non-zero memset arguments.
16094 2018-05-22  Martin Liska  <mliska@suse.cz>
16096         PR ipa/85607
16097         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16099 2018-05-22  Richard Biener  <rguenther@suse.de>
16101         PR tree-optimization/85863
16102         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16103         comparisons when vectype is specified.
16104         (vectorizable_condition): Do not specify vectype for
16105         vect_is_simple_cond when SLP vectorizing.
16107 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16109         PR target/85657
16110         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16111         define __ibm128 as long double.
16112         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16113         as a distinct type when IEEE 128-bit support is enabled.
16114         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16115         128-bit types to use the correct functions.
16116         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16117         convert between 128-bit floating point types that have different
16118         modes but the same representation, instead of using gen_lowpart to
16119         makean alias.
16120         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16121         KFmode.
16122         (IFKF_reg): New attributes to give the register constraints for
16123         IFmode and KFmode.
16124         (extend<mode>tf2_internal): New insns to mark an explicit
16125         conversion between 128-bit floating point types that have a
16126         different mode but share the same representation.
16128 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16130         PR tree-optimization/85814
16131         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16132         a null return from get_strinfo when unsharing the next
16133         strinfo in the chain.
16135 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16137         PR gcc/84923
16138         * varasm.c (weak_finish): Clean up weak_decls.
16140 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16142         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16143         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16144         UNSPEC_UADALP values.
16145         * config/aarch64/iterators.md (ABAL): New int iterator.
16146         (ABDL2): Likewise.
16147         (ADALP): Likewise.
16148         (sur): Add mappings for the above.
16149         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16150         New define_insn.
16151         (aarch64_<sur>abal<mode>_4): Likewise.
16152         (aarch64_<sur>adalp<mode>_3): Likewise.
16153         (<sur>sadv16qi): New define_expand.
16155 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16157         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16158         (*movdf_internal): Ditto.
16159         (*rcpsf2_sse): Ditto.
16160         (*rsqrtsf2_sse): Ditto.
16161         (*sqrt<mode>2_sse): Ditto.
16163 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16165         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16166         eor3q<mode>4.
16167         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16168         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16169         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16170         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16171         vbcaxq_s64): New.
16172         * config/aarch64/arm_neon.h: Likewise.
16173         * config/aarch64/iterators.md (VQ_I): New.
16175 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16177         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16178         arc*-*-linux*.
16179         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16180         MULTILIB_DIRNAMES
16182 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16184         * config/nds32/constraints.md (S): New constraint.
16185         * config/nds32/nds32.md (call_internal): Use constraint S.
16186         (call_value_internal): Likewise.
16187         (sibcall_internal): Likewise.
16188         (sibcall_value_internal): Likewise.
16190 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16191             Chung-Ju Wu  <jasonwucj@gmail.com>
16193         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16194         into consideration.
16196 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16197             Chung-Ju Wu  <jasonwucj@gmail.com>
16199         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16200         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16201         (nds32_rtx_costs_impl): Simplify.
16202         (nds32_address_cost_impl): Simplify.
16203         (nds32_init_rtx_costs): New function.
16204         (nds32_rtx_costs_speed_prefer): Likewise.
16205         (nds32_rtx_costs_size_prefer): Likewise.
16206         (nds32_address_cost_speed_prefer): Likewise.
16207         (nds32_address_cost_speed_fwprop): Likewise.
16208         (nds32_address_cost_size_prefer): Likewise.
16209         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16210         * config/nds32/nds32.c (nds32_option_override): Use
16211         nds32_init_rtx_costs function.
16213 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16215         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16216         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16217         (TARGET_PIPELINE_N8): Likewise.
16218         (TARGET_PIPELINE_N10): Likewise.
16219         (TARGET_PIPELINE_N13): Likewise.
16220         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16222 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16224         * config/nds32/nds32-fpu.md: Update copyright year.
16226 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16228         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16230 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16232         * config/nds32/nds32.c
16233         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16234         * config/nds32/nds32.opt (minline-asm-r15): New option.
16236 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16238         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16239         MASK_HW_ABS.
16240         * config/nds32/nds32.md (abssi2): New pattern.
16242 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16244         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16245         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16246         Merge insn pattern from sse_cvtsi2ss<round_name> and
16247         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16248         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16249         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16250         using SWI48 mode iterator.
16251         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16252         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16253         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16254         pattern from sse_cvttss2si<round_saeonly_name>
16255         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16256         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16257         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16258         using SWI48 mode iterator.
16259         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16260         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16261         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16262         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16263         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16264         using SWI48 mode iterator.
16265         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16266         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16267         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16268         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16269         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16270         SWI48 mode iterator.
16271         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16272         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16273         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16274         pattern from sse_cvttsd2si<round_saeonly_name>
16275         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16277 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16279         * config/nds32/nds32-md-auxiliary.c
16280         (nds32_valid_smw_lwm_base_p): Refine.
16281         (nds32_output_smw_single_word): Refine.
16282         (nds32_output_smw_double_word): New.
16283         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16285 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16287         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16288         (nds32_output_stack_pop): Refine.
16289         (nds32_expand_unaligned_load): Refine.
16290         (nds32_expand_unaligned_store): Refine.
16292 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16293             Chung-Ju Wu  <jasonwucj@gmail.com>
16295         * config/nds32/constants.md: Add TP_REGNUM constant.
16296         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16297         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16298         UNSPEC_ADD32.
16299         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16300         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16301         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16302         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16303         and PIC code generation.
16304         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16305         code generation.
16306         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16307         optimization.
16308         * config/nds32/nds32.md: Support TLS and PIC.
16309         * config/nds32/nds32.c: Support TLS and PIC.
16310         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16311         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16312         predicate.
16314 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16316         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16317         mode with E_ prefix.
16319 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16320             Chung-Ju Wu  <jasonwucj@gmail.com>
16322         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16323         * config/nds32/nds32-md-auxiliary.c
16324         (symbolic_reference_mentioned_p): New.
16325         (nds32_legitimize_ict_address): New.
16326         (nds32_expand_ict_move): New.
16327         (nds32_indirect_call_referenced_p): New.
16328         (nds32_symbol_binds_local_p): Delete.
16329         (nds32_long_call_p): Modify.
16330         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16331         * config/nds32/nds32-protos.h
16332         (symbolic_reference_mentioned_p): Declare.
16333         (nds32_legitimize_ict_address): Declare.
16334         (nds32_expand_ict_move): Declare.
16335         (nds32_indirect_call_referenced_p): Declare.
16336         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16337         (nds32_relax_group): Use nds32_ict_const_p as condition.
16338         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16339         (nds32_asm_file_start): Output ict_model directive in asm code.
16340         (nds32_legitimate_address_p): Consider indirect call.
16341         (nds32_print_operand): Consider indirect call.
16342         (nds32_print_operand_address): Consider indirect call.
16343         (nds32_insert_attributes): Handle "indirect_call" attribute.
16344         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16345         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16346         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16347         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16348         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16349         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16350         (TARGET_ICT_MODEL_SMALL): Define.
16351         (TARGET_ICT_MODEL_LARGE): Define.
16352         * config/nds32/nds32.md (movsi): Consider ict model.
16353         (call, call_value): Consider ict model.
16354         (sibcall, sibcall_value): Consider ict model.
16355         * config/nds32/nds32.opt (mict-model): New option.
16356         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16357         model.
16359 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16360             Monk Chiang  <sh.chiang04@gmail.com>
16361             Jim Wilson <jimw@sifive.com>
16363         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16364         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16365         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16366         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16367         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16368         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16369         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16370         compute save_libcall_adjustment properly.
16371         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16372         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16373         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16374         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16375         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16376         (ABI_SPEC): Handle mabi=ilp32e.
16377         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16378         (RVE): Add RVE mask.
16379         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16380         <-march>: Add rv32e as an example.
16382 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
16384         PR c++/82899
16385         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16386         (intra_create_variable_infos): Handle C++ constructors.
16388 2018-05-18  Martin Liska  <mliska@suse.cz>
16390         * passes.def: Remove a redundant pass.
16392 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
16394         PR bootstrap/85838
16395         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16397 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16399         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16400         (ARMv4): Update.
16401         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16402         (ARMv6m): Update.
16403         (armv2, armv2a, armv3, armv3m): Delete architectures.
16404         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16405         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16406         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16407         Delete cpus.
16408         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16409         (*mulsidi3adddi): Likewise.
16410         (mulsidi3): Likewise.
16411         (*mulsidi3_nov6): Likewise.
16412         (umulsidi3): Likewise.
16413         (umulsidi3_nov6): Likewise.
16414         (umaddsidi4): Likewise.
16415         (*umulsidi3adddi): Likewise.
16416         (smulsi3_highpart): Likewise.
16417         (*smulsi3_highpart_nov6): Likewise.
16418         (umulsi3_highpart): Likewise.
16419         (*umulsi3_highpart_nov6): Likewise.
16420         * config/arm/arm.h (arm_arch3m): Delete.
16421         * config/arm/arm.c (arm_arch3m): Delete.
16422         (arm_option_override_internal): Update armv3-related comment.
16423         (arm_configure_build_target): Delete use of isa_bit_mode32.
16424         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16425         (arm_rtx_costs_internal): Delete check of arm_arch3m.
16426         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16427         (mulsa3): Likewise.
16428         (mulusa3): Likewise.
16429         * config/arm/arm-protos.h (arm_arch3m): Delete.
16430         * config/arm/arm-tables.opt: Regenerate.
16431         * config/arm/arm-tune.md: Likewise.
16432         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16433         deleted architectures.
16435 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16437         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16438         (armv5t, armv5te): New features.
16439         (ARMv5, ARMv5e): Delete fgroups.
16440         (ARMv5t, ARMv5te): Adjust for above changes.
16441         (ARMv6m): Likewise.
16442         (armv5, armv5e): Delete arches.
16443         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16444         arm_arch5.
16445         (*call_reg_arm): Likewise.
16446         (*call_value_reg_armv5): Likewise.
16447         (*call_value_reg_arm): Likewise.
16448         (*call_symbol): Likewise.
16449         (*call_value_symbol): Likewise.
16450         (*sibcall_insn): Likewise.
16451         (*sibcall_value_insn): Likewise.
16452         (clzsi2): Likewise.
16453         (prefetch): Likewise.
16454         (define_split and define_peephole2 dependent on arm_arch5):
16455         Likewise.
16456         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16457         arm_arch5e.
16458         (TARGET_ARM_QBIT): Likewise.
16459         (TARGET_DSP_MULTIPLY): Likewise.
16460         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16461         (arm_arch5, arm_arch5e): Delete.
16462         (arm_arch5t, arm_arch5te): Declare.
16463         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16464         (arm_arch5t): Declare.
16465         (arm_option_reconfigure_globals): Update for the above.
16466         (arm_options_perform_arch_sanity_checks): Update comment, replace
16467         use of arm_arch5 with arm_arch5t.
16468         (use_return_insn): Likewise.
16469         (arm_emit_call_insn): Likewise.
16470         (output_return_instruction): Likewise.
16471         (arm_final_prescan_insn): Likewise.
16472         (arm_coproc_builtin_available): Likewise.
16473         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16474         arm_arch5e with arm_arch5t and arm_arch5te.
16475         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16476         (arm_arch5t, arm_arch5te): Declare.
16477         * config/arm/arm-tables.opt: Regenerate.
16478         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16479         * config/arm/t-multilib: Likewise.
16480         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16481         instead of arm_arch5.
16482         (*call_reg_thumb1): Likewise.
16483         (*call_value_reg_thumb1_v5): Likewise.
16484         (*call_value_reg_thumb1): Likewise.
16485         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16486         unreachable path.
16487         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16489 2018-05-18  Martin Liska  <mliska@suse.cz>
16491         PR gcov-profile/84846
16492         * doc/gcov.texi: Document -t option of gcov tool.
16494 2018-05-18  Martin Liska  <mliska@suse.cz>
16496         PR gcov-profile/84846
16497         * gcov.c (print_usage): Add new -t option.
16498         (process_args): Handle the option.
16499         (generate_results): Use stdout as output when requested by
16500         the option.
16502 2018-05-18  Martin Liska  <mliska@suse.cz>
16504         PR gcov-profile/84846
16505         * coverage.c (coverage_init): Write PWD to .gcno file.
16506         * doc/gcov.texi: Document how working directory is printed.
16507         * gcov-dump.c (dump_gcov_file): Print PWD.
16508         * gcov.c (output_intermediate_file): Likewise.
16509         (read_graph_file): Read PWD string.
16510         (output_lines): Print PWD.
16512 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16514         PR middle-end/85817
16515         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
16516         for retval and return false if all args to phi are zero.
16518 2018-05-18  Richard Biener  <rguenther@suse.de>
16520         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
16521         method.
16522         (evrp_dom_walker::before_dom_children): Call it.
16524 2018-05-18  Richard Biener  <rguenther@suse.de>
16526         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
16527         results when processing array refs with variable index.
16529 2018-05-18  Toon Moene  <toon@moene.org>
16531         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
16532         directly after that of -floop-interchange. Indicate that both
16533         options are enabled by default when specifying -O3.
16535 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16537         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
16538         iterator.  Delete separate integer-mode vec_set<mode> expander.
16539         (aarch64_simd_vec_setv2di): Delete.
16540         (vec_setv2di): Delete.
16541         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
16542         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
16543         the "w, r" alternative.
16545 2018-05-18  Martin Liska  <mliska@suse.cz>
16547         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
16548         * tree-pass.h (make_pass_lower_switch_O0): New function.
16549         * tree-switch-conversion.c (node_has_low_bound): Remove.
16550         (node_has_high_bound): Likewise.
16551         (node_is_bounded): Likewise.
16552         (class pass_lower_switch): Make it a template type and create
16553         two instances.
16554         (pass_lower_switch::execute): Add template argument.
16555         (make_pass_lower_switch): New function.
16556         (make_pass_lower_switch_O0): New function.
16557         (do_jump_if_equal): Remove.
16558         (emit_case_nodes): Simplify to just handle all 3 cases and leave
16559         all the hard work to tree optimization passes.
16561 2018-05-18  Martin Liska  <mliska@suse.cz>
16563         * dbgcnt.c (limit_low): Renamed from limit.
16564         (limit_high): New variable.
16565         (dbg_cnt_is_enabled): Check for upper limit.
16566         (dbg_cnt): Adjust dumping.
16567         (dbg_cnt_set_limit_by_index): Add new argument for high
16568         value.
16569         (dbg_cnt_set_limit_by_name): Likewise.
16570         (dbg_cnt_process_single_pair): Parse new format.
16571         (dbg_cnt_process_opt): Use strtok.
16572         (dbg_cnt_list_all_counters): Remove 'value' and add
16573         'limit_high'.
16574         * doc/invoke.texi: Document changes.
16576 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
16578         * doc/sourcebuild.texi (scalar_all_fma): Document.
16579         * tree.def (FMA_EXPR): Delete.
16580         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
16581         * internal-fn.c (ternary_direct): New macro.
16582         (expand_ternary_optab_fn): Likewise.
16583         (direct_ternary_optab_supported_p): Likewise.
16584         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
16585         * builtins.c (fold_builtin_fma): Delete.
16586         (fold_builtin_3): Don't call it.
16587         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
16588         * expr.c (expand_expr_real_2): Likewise.
16589         * fold-const.c (operand_equal_p): Likewise.
16590         (fold_ternary_loc): Likewise.
16591         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
16592         * gimple.c (DEFTREECODE): Likewise.
16593         * gimplify.c (gimplify_expr): Likewise.
16594         * optabs-tree.c (optab_for_tree_code): Likewise.
16595         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16596         * tree-eh.c (operation_could_trap_p): Likewise.
16597         (stmt_could_throw_1_p): Likewise.
16598         * tree-inline.c (estimate_operator_cost): Likewise.
16599         * tree-pretty-print.c (dump_generic_node): Likewise.
16600         (op_code_prio): Likewise.
16601         * tree-ssa-loop-im.c (stmt_cost): Likewise.
16602         * tree-ssa-operands.c (get_expr_operands): Likewise.
16603         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
16604         * fold-const-call.h (fold_fma): Delete.
16605         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
16606         CFN_FNMA and CFN_FNMS.
16607         (fold_fma): Delete.
16608         * genmatch.c (combined_fn): New enum.
16609         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
16610         (commutative_op): New function.
16611         (commutate): Use it.  Handle more than 2 operands.
16612         (dt_operand::gen_gimple_expr): Use commutative_op.
16613         (parser::parse_expr): Allow :c to be used with non-binary
16614         operators if the commutative operand is known.
16615         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
16616         CFN_FMS, CFN_FNMA and CFN_FNMS.
16617         (backprop::process_assign_use): Remove FMA_EXPR handling.
16618         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
16619         (gen_hsa_fma): New function.
16620         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
16621         IFN_FNMA and IFN_FNMS.
16622         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
16623         * gimple-fold.h (follow_all_ssa_edges): Declare.
16624         * gimple-fold.c (follow_all_ssa_edges): New function.
16625         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
16626         gimple_build interface and use follow_all_ssa_edges to fold the result.
16627         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
16628         instead of checking for optabs directly.
16629         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
16630         rather than FMA_EXPRs.
16631         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
16632         call to IFN_FMA instead of an FMA_EXPR.
16634 2018-05-17  Jim Wilson  <jimw@sifive.com>
16636         * expr.c (do_tablejump): When converting index to Pmode, if we have a
16637         sign extended promoted subreg, and the range does not have the sign bit
16638         set, then do a sign extend.
16640         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
16641         test, check for sign extended subreg and/or constant operands, and
16642         do a sign extend in that case.
16644 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
16646         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
16647         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
16648         Add untyped.
16649         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
16650         Change logics_shift_reg to logics_shift_imm.
16651         (thunderx2t99_fp_loadpair_basic): Delete.
16652         (thunderx2t99_fp_storepair_basic): Delete.
16653         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
16654         (thunderx2t99_asimd_polynomial): Delete.
16655         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
16656         and neon_fp_mul_d_scalar_q.
16657         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
16658         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
16659         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
16660         (thunderx2t99_asimd_lut): Add missing tbl types.
16661         (thunderx2t99_asimd_ext): Delete.
16662         (thunderx2t99_asimd_load1_1_mult): Delete.
16663         (thunderx2t99_asimd_load1_2_mult): Delete.
16664         (thunderx2t99_asimd_load1_ldp): New.
16665         (thunderx2t99_asimd_load1): New.
16666         (thunderx2t99_asimd_load2): Add missing *load2* types.
16667         (thunderx2t99_asimd_load3): New.
16668         (thunderx2t99_asimd_load4): New.
16669         (thunderx2t99_asimd_store1_1_mult): Delete.
16670         (thunderx2t99_asimd_store1_2_mult): Delete.
16671         (thunderx2t99_asimd_store2_mult): Delete.
16672         (thunderx2t99_asimd_store2_onelane): Delete.
16673         (thunderx2t99_asimd_store_stp): New.
16674         (thunderx2t99_asimd_store1): New.
16675         (thunderx2t99_asimd_store2): New.
16676         (thunderx2t99_asimd_store3): New.
16677         (thunderx2t99_asimd_store4): New.
16679 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
16681         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
16682         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
16684 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
16685             Segher Boessenkool  <segher@kernel.crashing.org>
16687         PR target/85698
16688         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
16689         operand.
16691 2018-05-17  Richard Biener  <rguenther@suse.de>
16693         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
16694         for pruning loop and prune defs feeding only already visited PHIs.
16696 2018-05-17  Richard Biener  <rguenther@suse.de>
16698         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
16700 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
16701             Richard Biener  <rguenther@suse.de>
16703         PR tree-optimization/85793
16704         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
16705         for VMAT_ELEMENTWISE.
16707 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16709         * internal-fn.h (lookup_internal_fn): Declare
16710         * internal-fn.c (lookup_internal_fn): New function.
16711         * gimple.c (gimple_build_call_from_tree): Handle calls to
16712         internal functions.
16713         * gimple-pretty-print.c (dump_gimple_call): Print "." before
16714         internal function names.
16715         * tree-pretty-print.c (dump_generic_node): Likewise.
16716         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
16718 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16720         * gimple-fold.h (gimple_build): Make the function forms take
16721         combined_fn rather than built_in_function.
16722         (gimple_simplify): Likewise.
16723         * gimple-match-head.c (gimple_simplify): Likewise.
16724         * gimple-fold.c (gimple_build): Likewise.
16725         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
16726         rather than gimple_build_call_internal.
16727         (get_initial_defs_for_reduction): Likewise.
16728         (vect_create_epilog_for_reduction): Likewise.
16729         (vectorizable_live_operation): Likewise.
16731 2018-05-17  Martin Liska  <mliska@suse.cz>
16733         * gimple-ssa-sprintf.c (format_directive): Do not use
16734         space in between 'G_' and '('.
16736 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
16738         PR target/85323
16739         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
16740         even if the mask is not all ones.
16742         PR target/85323
16743         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
16744         vector.
16745         (ix86_gimple_fold_builtin): Likewise.
16747         PR target/85323
16748         * config/i386/i386.c: Include tree-vector-builder.h.
16749         (ix86_vector_shift_count): New function.
16750         (ix86_fold_builtin): Fold shift builtins by scalar count.
16751         (ix86_gimple_fold_builtin): Likewise.
16753         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
16754         _mm512_setzero): New intrinsics.
16756 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
16757             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16759         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
16760         code generation for cases where splatting a value is not useful.
16761         * simplify-rtx.c (simplify_ternary_operation): Simplify
16762         vec_merge across a vec_duplicate and a paradoxical subreg forming
16763         a vector mode to a vec_concat.
16765 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
16767         * config.gcc: Support "goldmont-plus".
16768         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16769         "goldmont-plus".
16770         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16771         PROCESSOR_GOLDMONT_PLUS.
16772         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
16773         (processor_target_table): Add "goldmont-plus".
16774         (PTA_GOLDMONT_PLUS): Define.
16775         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
16776         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
16777         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
16778         (fold_builtin_cpu): Add "goldmont-plus".
16779         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
16780         (ix86_option_override_internal): Add "goldmont-plus".
16781         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
16782         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
16783         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
16784         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
16786 2018-05-17  Richard Biener  <rguenther@suse.de>
16788         PR tree-optimization/85757
16789         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16790         remove defs that only feed that PHI from further processing.
16792 2018-05-16  Jim Wilson  <jimw@sifive.com>
16794         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16795         asterisk to name.
16796         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16798 2018-05-16  Mark Wielaard  <mark@klomp.org>
16800         * dwarf2out.c (count_index_strings): New function.
16801         (output_indirect_strings): Call count_index_strings and generate
16802         header for dwarf_version >= 5.
16804 2018-05-16  Mark Wielaard  <mark@klomp.org>
16806         * dwarf2out.c (dwarf_FORM): New function.
16807         (set_indirect_string): Use dwarf_FORM.
16808         (reset_indirect_string): Likewise.
16809         (size_of_die): Likewise.
16810         (value_format): Likewise.
16811         (output_die): Likewise.
16812         (add_skeleton_AT_string): Likewise.
16813         (output_macinfo_op): Likewise.
16814         (index_string): Likewise.
16815         (output_index_string_offset): Likewise.
16816         (output_index_string): Likewise.
16817         (count_index_strings): Likewise.
16819 2018-05-16  Carl Love  <cel@us.ibm.com>
16821         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16822         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16824 2018-05-16  Martin Jambor  <mjambor@suse.cz>
16826         * ipa-prop.c (ipa_free_all_edge_args): Remove.
16827         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16829 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
16831         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16832         (fnma<mode>4): Likewise.
16833         (fms<mode>4): Likewise.
16834         (fnms<mode>4): Likewise.
16835         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16836         (aarch64_fnma<mode>4): Likewise.
16837         (aarch64_fms<mode>4): Likewise.
16838         (aarch64_fnms<mode>4): Likewise.
16839         (aarch64_fnmadd<mode>4): Likewise.
16841 2018-05-16  Jason Merrill  <jason@redhat.com>
16843         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
16845 2018-05-16  Richard Biener  <rguenther@suse.de>
16847         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16848         (dump_stmt_cost): Declare.
16849         (add_stmt_cost): Dump cost we add.
16850         (add_stmt_costs): New function.
16851         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16852         No longer exported.
16853         (vect_analyze_stmt): Adjust prototype.
16854         (vectorizable_condition): Likewise.
16855         (vectorizable_live_operation): Likewise.
16856         (vectorizable_reduction): Likewise.
16857         (vectorizable_induction): Likewise.
16858         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16859         cost vector to pass to vectorizable_ and record afterwards.
16860         (vect_model_reduction_cost): Take cost vector argument and adjust.
16861         (vect_model_induction_cost): Likewise.
16862         (vectorizable_reduction): Likewise.
16863         (vectorizable_induction): Likewise.
16864         (vectorizable_live_operation): Likewise.
16865         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16866         SLP_TREE_NUMBER_OF_VEC_STMTS.
16867         (vect_analyze_slp_cost_1): Remove.
16868         (vect_analyze_slp_cost): Likewise.
16869         (vect_slp_analyze_node_operations): Take visited args and
16870         a target cost vector.  Avoid processing already visited stmt sets.
16871         (vect_slp_analyze_operations): Use a local cost vector to gather
16872         costs and register those of non-discarded instances.
16873         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16874         (vect_schedule_slp_instance): Remove copying of
16875         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
16876         zero.
16877         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16878         adding cost.  Record cost entry location.
16879         (vect_prologue_cost_for_slp_op): Function to compute cost of
16880         a constant or invariant generated for SLP vect in the prologue,
16881         split out from vect_analyze_slp_cost_1.
16882         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
16883         (vect_model_promotion_demotion_cost): Likewise.
16884         (vect_model_store_cost): Likewise, make static.
16885         (vect_model_load_cost): Likewise.
16886         (vectorizable_bswap): Add cost vector arg and adjust.
16887         (vectorizable_call): Likewise.
16888         (vectorizable_simd_clone_call): Likewise.
16889         (vectorizable_conversion): Likewise.
16890         (vectorizable_assignment): Likewise.
16891         (vectorizable_shift): Likewise.
16892         (vectorizable_operation): Likewise.
16893         (vectorizable_store): Likewise.
16894         (vectorizable_load): Likewise.
16895         (vectorizable_condition): Likewise.
16896         (vectorizable_comparison): Likewise.
16897         (can_vectorize_live_stmts): Likewise.
16898         (vect_analyze_stmt): Likewise.
16899         (vect_transform_stmt): Adjust calls to vectorizable_*.
16900         * tree-vectorizer.c: Include gimple-pretty-print.h.
16901         (dump_stmt_cost): New function.
16903 2018-05-16  Richard Biener  <rguenther@suse.de>
16905         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16906         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16907         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16908         (check_name): New callback.
16909         (dse_classify_store): Track cycles via a visited bitmap of PHI
16910         defs and simplify handling of in-loop and across loop dead stores
16911         and properly fail for loop-variant refs.  Handle byte-tracking with
16912         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16913         limiting the walk.
16915 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16917         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16918         (vect_get_mask_type_for_stmt): Likewise.
16919         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16920         split out from...
16921         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16922         to determine the statement's vector type and the vector type that
16923         should be used for calculating nunits.  Deal with cases in which
16924         the type has to be deferred.
16925         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16926         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16927         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16928         (vect_determine_vf_for_stmt): New functions, split out from...
16929         (vect_determine_vectorization_factor): ...here.
16930         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16931         (vect_get_mask_type_for_stmt): New functions, split out from
16932         vect_determine_vectorization_factor.
16934 2018-05-16  Richard Biener  <rguenther@suse.de>
16936         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16937         verify the [VEC_]COND_EXPR embedded comparison.
16939 2018-05-15  Martin Sebor  <msebor@redhat.com>
16941         PR tree-optimization/85753
16942         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16943         RECORD_TYPE in addition to ARRAY_TYPE.
16945 2018-05-15  Martin Sebor  <msebor@redhat.com>
16947         PR middle-end/85643
16948         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16950 2018-05-15  Richard Biener  <rguenther@suse.de>
16952         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16953         add by_clobber_p one.  Change algorithm to collect all defs
16954         representing uses we need to walk and try reducing them to
16955         a single one before failing.
16956         (dse_dom_walker::dse_optimize_stmt): Adjust.
16958 2018-05-13  Mark Wielaard  <mark@klomp.org>
16960         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16961         (size_of_loc_descr): Likewise.
16962         (output_loc_operands): Likewise.
16963         (output_loc_operands_raw): Likewise.
16964         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16965         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16966         (hash_loc_operands): Likewise.
16967         (compare_loc_operands): Likewise.
16969 2018-05-14  Mark Wielaard  <mark@klomp.org>
16971         * dwarf2out.c (count_index_addrs): New function.
16972         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16974 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16976         PR tree-optimization/83648
16977         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16978         return value as malloc candidate.
16980 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16982         PR ipa/85734
16983         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16984         param as true in call to suggest_attribute.
16986 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
16988         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16989         -mreadonly-in-sdata.
16991 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16993         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16994         New pattern.
16995         (aarch64_crypto_aesd_fused): Likewise.
16997 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
16999         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17000         (movsi_aarch64): Likewise.
17001         (load_pairsi): Likewise.
17002         (load_pairdi): Likewise.
17003         (store_pairsi): Likewise.
17004         (store_pairdi): Likewise.
17005         (load_pairsf): Likewise.
17006         (load_pairdf): Likewise.
17007         (store_pairsf): Likewise.
17008         (store_pairdf): Likewise.
17009         (zero_extend): Likewise.
17010         (trunc): Swap alternatives.
17011         (fcvt_target): Add '?' to prefer w over r.
17013 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17015         PR target/85756
17016         * config/i386/i386.md: Disallow non-commutative arithmetics in
17017         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17018         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17019         in the peephole2 before it.
17021 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17023         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17024         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17025         (ix86_handle_option): Handle -mcldemote.
17026         * config.gcc: New header.
17027         * config/i386/cldemoteintrin.h: New file.
17028         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17029         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17030         -mcldemote.
17031         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17032         OPTION_MASK_ISA_CLDEMOTE.
17033         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17034         (ix86_valid_target_attribute_inner_p): Ditto.
17035         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17036         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17037         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17038         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17039         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17040         (cldemote): New.
17041         * config/i386/i386.opt: Add -mcldemote.
17042         * config/i386/x86intrin.h: New header.
17043         * doc/invoke.texi: Add -mcldemote.
17045 2018-05-14  Richard Biener  <rguenther@suse.de>
17047         * doc/match-and-simplify.texi: Adjust :s documentation.
17049 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17051         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17052         intended memcpy size.
17053         (REORDER_45): Likewise.
17055 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17057         * sort.cc: New file.
17058         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17059         * vec.c (qsort_chk): Use gcc_qsort.
17060         * Makefile.in (OBJS-libcommon): Add sort.o.
17061         (build/sort.o): New target.  Use it...
17062         (BUILD_RTL): ... here, and...
17063         (build/gencfn-macros): ... here, and...
17064         (build/genmatch): ... here.
17066 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17067             Chung-Ju Wu  <jasonwucj@gmail.com>
17069         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17070         * config/nds32/nds32-graywolf.md: New file.
17071         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17072         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17073         pipeline.
17074         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17075         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17076         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17077         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17078         * config/nds32/pipelines.md: Include n15 settings.
17080 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17081             Chung-Ju Wu  <jasonwucj@gmail.com>
17083         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17084         * config/nds32/nds32-n13.md: New file.
17085         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17086         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17087         pipeline.
17088         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17089         * config/nds32/nds32.md (pipeline_model): Add n13.
17090         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17091         * config/nds32/pipelines.md: Include n13 settings.
17093 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17094             Chung-Ju Wu  <jasonwucj@gmail.com>
17096         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17097         * config/nds32/nds32-n10.md: New file.
17098         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17099         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17100         pipeline.
17101         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17102         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17103         * config/nds32/nds32.md (pipeline_model): Add n10.
17104         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17105         * config/nds32/pipelines.md: Include n10 settings.
17107 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17108             Kito Cheng  <kito.cheng@gmail.com>
17109             Chung-Ju Wu  <jasonwucj@gmail.com>
17111         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17112         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17113         Add enum values for DSP extension instructions.
17114         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17115         New constraints.
17116         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17117         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17118         New code iterators.
17119         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17120         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17121         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17122         * config/nds32/nds32-intrinsic.md: Likewise.
17123         * config/nds32/nds32_intrinsic.h: Likewise.
17124         * config/nds32/nds32-md-auxiliary.c: Likewise.
17125         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17126         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17127         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17128         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17129         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17130         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17131         TYPE_DMAC in switch statement.
17132         * config/nds32/nds32.c: New checking and implementation for DSP
17133         extension instructions.
17134         * config/nds32/nds32.h: Likewise.
17135         * config/nds32/nds32.md: Likewise.
17136         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17137         * config/nds32/predicates.md: Implement new predicates for DSP
17138         extension.
17140 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17142         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17143         Reformat alternatives and attributes so it is easier to identify
17144         which constraints/attributes go with which instruction.
17145         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17146         (mov<mode>_softfloat32, FMOVE64): Likewise.
17147         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17148         (mov<mode>_softfloat64, FMOVE64): Likewise.
17150 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17152         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17153         subsection.
17154         (Basic PowerPC Built-in Functions): The new name of the
17155         subsection previously known as "PowerPC Built-in Functions".
17156         (Basic PowerPC Built-in Functions Available on all Configurations):
17157         New subsubsection.
17158         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17159         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17160         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17161         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17163 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17165         PR ipa/85655
17166         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17167         single const.
17169 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17171         PR target/85733
17172         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17174 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17176         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17177         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17178         (ix86_handle_option): Handle -mwaitpkg.
17179         * config.gcc: New header.
17180         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17181         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17182         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17183         function type.
17184         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17185         OPTION_MASK_ISA_WAITPKG.
17186         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17187         (ix86_option_override_internal): Add PTA_WAITPKG.
17188         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17189         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17190         IX86_BUILTIN_TPAUSE.
17191         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17192         __builtin_ia32_umwait and __builtin_ia32_tpause.
17193         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17194         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17195         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17196         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17197         UNSPECV_TPAUSE): New.
17198         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17199         * config/i386/i386.opt: Add -mwaitpkg.
17200         * config/i386/waitpkgintrin.h: New file.
17201         * config/i386/x86intrin.h: New header.
17202         * doc/invoke.texi: Add -mwaitpkg.
17204 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17206         PR target/85606
17207         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17208         equivalent.
17209         (cortex-m0): Use armv6s-m isa.
17210         (cortex-m0plus): Likewise.
17211         (cortex-m1): Likewise.
17212         (cortex-m0.small-multiply): Likewise.
17213         (cortex-m0plus.small-multiply): Likewise.
17214         (cortex-m1.small-multiply): Likewise.
17216 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17217             Jakub Jelinek  <jakub@redhat.com>
17219         PR tree-optimization/85692
17220         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17221         source permute as well.
17223 2018-05-11  Martin Liska  <mliska@suse.cz>
17225         PR sanitizer/85556
17226         * doc/extend.texi: Document LLVM style format for no_sanitize
17227         attribute.
17229 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17231         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17232         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17233         (mode_supports_vsx_dform_quad): Likewise.
17234         (mode_supports_vmx_dform): Move these functions to be next to the
17235         other mode_supports functions.
17236         (mode_supports_dq_form): Likewise.
17237         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17238         mode_supports_dq_form.
17239         (reg_offset_addressing_ok_p): Likewise.
17240         (offsettable_ok_by_alignment): Likewise.
17241         (rs6000_legitimate_offset_address_p): Likewise.
17242         (legitimate_lo_sum_address_p): Likewise.
17243         (rs6000_legitimize_address): Likewise.
17244         (rs6000_legitimize_reload_address): Likewise.
17245         (rs6000_secondary_reload_inner): Likewise.
17246         (rs6000_preferred_reload_class): Likewise.
17247         (rs6000_output_move_128bit): Likewise.
17249 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17251         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17252         Generate SImode target register for null target.
17253         <case IX86_BUILTIN_XGETBV>: Ditto.
17254         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17255         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17257 2018-05-10  Carl Love  <cel@us.ibm.com>
17259         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17260         dcbtt and dcbtstt if operands[2] is 0.
17262 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17264         PR target/85693
17265         * config/i386/sse.md (usadv64qi): New expander.
17267 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17269         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17270         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17271         -maltivec=be support.
17272         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17273         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17274         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17275         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17276         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17277         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17278         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17279         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17280         altivec_vsumsws): Adjust.
17281         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17282         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17283         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17284         support.
17285         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17286         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17287         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17288         (altivec_lve<VI_char>x): Delete expand.
17289         (*altivec_lve<VI_char>x_internal): Rename to...
17290         (altivec_lve<VI_char>x): ... this.
17291         (altivec_lvxl_<mode>): Delete expand.
17292         (*altivec_lvxl_<mode>_internal): Rename to ...
17293         (altivec_lvxl_<mode>): ... this.
17294         (altivec_stvxl_<mode>): Delete expand.
17295         (*altivec_stvxl_<mode>_internal): Rename to ...
17296         (altivec_stvxl_<mode>): ... this.
17297         (altivec_stve<VI_char>x): Delete expand.
17298         (*altivec_stve<VI_char>x_internal): Rename to ...
17299         (altivec_stve<VI_char>x): ... this.
17300         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17301         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17302         reduc_plus_scal_<mode>): Adjust.
17303         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17304         comment.
17305         (rs6000_cpu_cpp_builtins): Adjust.
17306         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17307         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17308         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17309         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17310         -maltivec=be support.
17311         (rs6000_split_vec_extract_var): Adjust.
17312         (rs6000_split_v4si_init): Adjust.
17313         (swap_selector_for_mode): Delete.
17314         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17315         altivec_expand_stvex_be): Delete.
17316         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17317         -maltivec=be support.
17318         (rs6000_gimple_fold_builtin): Ditto.
17319         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17320         Adjust.
17321         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17322         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17323         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17324         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17325         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17326         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17327         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17328         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17329         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17330         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17331         anonymous split): Adjust.
17332         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17333         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17335 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17337         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17338         when --with-gxx-include-dir is also specified.
17339         * configure: Regenerate.
17341 2018-05-09  Jim Wilson  <jimw@sifive.com>
17343         PR target/84797
17344         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17345         * config/riscv/t-withmultilib: New.
17346         * config/riscv/withmultilib.h: New.
17347         * doc/install.texi: Document RISC-V --with-multilib-list support.
17349 2018-05-09  Richard Biener  <rguenther@suse.de>
17351         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17352         vector.
17353         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17354         actual scalar cost using the cost vector and the add_stmt_cost
17355         machinery.
17357 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17359         PR rtl-optimization/85645
17360         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17361         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17363 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17365         PR rtl-optimization/85645
17366         * shrink-wrap.c (spread_components): Return a boolean saying if
17367         anything was changed.
17368         (try_shrink_wrapping_separate): Iterate spread_components until
17369         nothing changes anymore.
17371 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17373         PR rtl-optimization/85645
17374         * regrename.c (build_def_use): Also kill the chains that include the
17375         destination of a REG_CFA_REGISTER note.
17377 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17379         PR rtl-optimization/85645
17380         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17381         insn that has a REG_CFA_REGISTER note.
17383 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
17385         * cfgexpand.c (expand_clobber): New function.
17386         (expand_gimple_stmt_1): Use it.
17387         * tree-vect-stmts.c (vect_clobber_variable): New function,
17388         split out from...
17389         (vectorizable_simd_clone_call): ...here.
17390         (vectorizable_store): Emit a clobber either side of an
17391         IFN_STORE_LANES sequence.
17392         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17394 2018-05-09  Tom de Vries  <tom@codesourcery.com>
17396         PR target/85626
17397         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17398         (define_insn "trap_if_false"): Add exit after trap.
17400 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
17402         PR rtl-optimization/85638
17403         * bb-reorder.c: Include common/common-target.h.
17404         (create_forwarder_block): New function extracted from...
17405         (fix_up_crossing_landing_pad): ...here.  Rename into...
17406         (dw2_fix_up_crossing_landing_pad): ...this.
17407         (sjlj_fix_up_crossing_landing_pad): New function.
17408         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17409         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17410         from both partitions and exit the loop after one iteration.
17412 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17414         Revert:
17415         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17416         subsection.
17417         (Basic PowerPC Built-in Functions): The new name of the
17418         subsection previously known as "PowerPC Built-in Functions".
17419         (Basic PowerPC Built-in Functions Available on all Configurations):
17420         New subsubsection.
17421         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17422         subsubsection.
17423         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17424         subsubsection.
17425         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17426         subsubsection.
17427         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17428         subsubsection.
17430 2018-05-08  Jim Wilson  <jimw@sifive.com>
17432         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17433         (LD_EMUL_SUFFIX): New.
17434         (LINK_SPEC): Use it.
17436 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17438         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17439         subsection.
17440         (Basic PowerPC Built-in Functions): The new name of the
17441         subsection previously known as "PowerPC Built-in Functions".
17442         (Basic PowerPC Built-in Functions Available on all Configurations):
17443         New subsubsection.
17444         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17445         subsubsection.
17446         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17447         subsubsection.
17448         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17449         subsubsection.
17450         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17451         subsubsection.
17453 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17455         PR target/85683
17456         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17457         after cmpelim optimization.
17459 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
17461         * config.gcc: Support "goldmont".
17462         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17463         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17464         PROCESSOR_GOLDMONT.
17465         * config/i386/i386.c (m_GOLDMONT): Define.
17466         (processor_target_table): Add "goldmont".
17467         (PTA_GOLDMONT): Define.
17468         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17469         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17470         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17471         (fold_builtin_cpu): Add "goldmont".
17472         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17473         (ix86_option_override_internal): Add "goldmont".
17474         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17475         (processor_type): Add PROCESSOR_GOLDMONT.
17476         * config/i386/i386.md: Add CPU "glm".
17477         * config/i386/glm.md: New file.
17478         * config/i386/x86-tune.def: Add m_GOLDMONT.
17479         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17481 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17483         PR target/85572
17484         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17485         E_V4DImode.
17486         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17487         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
17488         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
17490         PR target/85317
17491         * config/i386/i386.c (ix86_fold_builtin): Handle
17492         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17494         PR target/85480
17495         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17496         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17498 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
17500         PR target/85658
17501         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17502         (check_arch): Likewise.
17503         (check_fpu): Return the result rather than printing it.
17504         (end arch): Fix operator precedence.
17505         (end cpu): Likewise.
17506         (END): Print the result from check_fpu.
17508 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17509             Alan Hayward  <alan.hayward@arm.com>
17510             David Sherwood  <david.sherwood@arm.com>
17512         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
17513         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
17514         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
17515         (*fcmuo<mode>_and): New patterns.
17517 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17519         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
17520         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
17521         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
17522         (cmp_op, sve_imm_con): New code attributes.
17523         (SVE_COND_INT_CMP, imm_con): Delete.
17524         (cmp_op): Remove above unspecs from int attribute.
17525         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
17526         to...
17527         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
17528         comparison-specific unspecs.
17529         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
17530         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
17531         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
17532         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17533         (*vec_fcm<cmp_op><mode>): Rename to...
17534         (*fcm<cmp_op><mode>): ...this and adjust likewise.
17535         (*vec_fcmuo<mode>): Rename to...
17536         (*fcmuo<mode>): ...this and adjust likewise.
17537         (*pred_fcm<cmp_op><mode>): New pattern.
17538         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
17539         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
17540         functions.
17541         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
17542         and UNORDERED.
17543         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
17544         (aarch64_emit_sve_predicated_cond): New function.
17545         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
17546         (aarch64_emit_unspec_cond_or): Replace with...
17547         (aarch64_emit_sve_or_conds): ...this new function.  Use
17548         aarch64_emit_sve_ptrue_op for the individual comparisons and
17549         aarch64_emit_binop to OR them together.
17550         (aarch64_emit_inverted_unspec_cond): Replace with...
17551         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
17552         aarch64_emit_sve_ptrue_op for the comparison and
17553         aarch64_emit_unop to invert the result.
17554         (aarch64_expand_sve_vec_cmp_float): Update after the above
17555         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
17557 2018-05-07  Nathan Sidwell  <nathan@acm.org>
17559         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
17560         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
17561         (Backwards Compatibility): Likewise.
17563 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17565         PR bootstrap/85681
17566         Revert:
17567         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17569         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17570         <prefetch_dynamic_strides>: New const bool field.
17571         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17572         prefetch_dynamic_strides.
17573         (exynosm1_prefetch_tune): Likewise.
17574         (thunderxt88_prefetch_tune): Likewise.
17575         (thunderx_prefetch_tune): Likewise.
17576         (thunderx2t99_prefetch_tune): Likewise.
17577         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17578         to false.
17579         (aarch64_override_options_internal): Update to set
17580         PARAM_PREFETCH_DYNAMIC_STRIDES.
17581         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17582         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17583         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17584         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17585         prefetch-dynamic-strides setting.
17587         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17589         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17590         <minimum_stride>: New const int field.
17591         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17592         minimum_stride field.
17593         (exynosm1_prefetch_tune): Likewise.
17594         (thunderxt88_prefetch_tune): Likewise.
17595         (thunderx_prefetch_tune): Likewise.
17596         (thunderx2t99_prefetch_tune): Likewise.
17597         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17598         (aarch64_override_options_internal): Update to set
17599         PARAM_PREFETCH_MINIMUM_STRIDE.
17600         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17601         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17602         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17603         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17604         stride is constant and is below the minimum stride threshold.
17606 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17608         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
17609         to 512.
17611 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17613         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17614         <prefetch_dynamic_strides>: New const bool field.
17615         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17616         prefetch_dynamic_strides.
17617         (exynosm1_prefetch_tune): Likewise.
17618         (thunderxt88_prefetch_tune): Likewise.
17619         (thunderx_prefetch_tune): Likewise.
17620         (thunderx2t99_prefetch_tune): Likewise.
17621         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17622         to false.
17623         (aarch64_override_options_internal): Update to set
17624         PARAM_PREFETCH_DYNAMIC_STRIDES.
17625         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17626         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17627         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17628         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17629         prefetch-dynamic-strides setting.
17631 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17633         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17634         <minimum_stride>: New const int field.
17635         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17636         minimum_stride field.
17637         (exynosm1_prefetch_tune): Likewise.
17638         (thunderxt88_prefetch_tune): Likewise.
17639         (thunderx_prefetch_tune): Likewise.
17640         (thunderx2t99_prefetch_tune): Likewise.
17641         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17642         (aarch64_override_options_internal): Update to set
17643         PARAM_PREFETCH_MINIMUM_STRIDE.
17644         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17645         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17646         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17647         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17648         stride is constant and is below the minimum stride threshold.
17650 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
17652         PR c++/85659
17653         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
17654         the type is addressable.  Don't force op into register if it has
17655         BLKmode.
17657 2018-05-05  Roland McGrath  <mcgrathr@google.com>
17659         PR other/77609
17660         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
17661         any section for which we don't know a specific type it should have,
17662         regardless of name.  Previously this was done only for the exact
17663         names ".init_array", ".fini_array", and ".preinit_array".
17664         (default_elf_asm_named_section): Add comment about
17665         relationship with default_section_type_flags and SECTION_NOTYPE.
17666         (get_section): Don't consider it a type conflict if one side has
17667         SECTION_NOTYPE and the other doesn't, as long as neither has the
17668         SECTION_BSS et al used in the default_section_type_flags logic.
17670 2018-05-05  Tom de Vries  <tom@codesourcery.com>
17672         PR target/85653
17673         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
17674         (workaround_barsyncs): New function.
17675         (nvptx_reorg): Use workaround_barsyncs.
17676         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17677         (define_expand "nvptx_membar_cta"): New define_expand.
17678         (define_insn "*nvptx_membar_cta"): New insn.
17680 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
17682         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
17683         To improve optimization opportunities.
17684         * builtin-types.def: The new needed builtin types for the above.
17686 2018-05-04  Richard Biener  <rguenther@suse.de>
17688         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
17689         * gimple-ssa-store-merging.c
17690         (imm_store_chain_info::output_merged_store): Remove redundant create,
17691         release split_store vector contents on failure.
17692         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
17693         scalar stmt vector on cache hit.
17695 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
17697         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
17698         Xilinx FP support.
17699         * config.gcc (powerpc-xilinx-eabi*): Remove.
17700         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
17701         support.
17702         (fusion_addis_mem_combo_load): Ditto.
17703         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
17704         FP support.
17705         (rs6000_cpu_cpp_builtins): Ditto.
17706         * config/rs6000/rs6000-linux.c
17707         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
17708         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
17709         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
17710         support.
17711         (rs6000_setup_reg_addr_masks): Ditto.
17712         (rs6000_init_hard_regno_mode_ok): Ditto.
17713         (rs6000_option_override_internal): Ditto.
17714         (legitimate_lo_sum_address_p): Ditto.
17715         (rs6000_legitimize_address): Ditto.
17716         (rs6000_legitimize_reload_address): Ditto.
17717         (rs6000_legitimate_address_p): Ditto.
17718         (abi_v4_pass_in_fpr): Ditto.
17719         (setup_incoming_varargs): Ditto.
17720         (rs6000_gimplify_va_arg): Ditto.
17721         (rs6000_split_multireg_move): Ditto.
17722         (rs6000_savres_strategy): Ditto.
17723         (rs6000_emit_prologue_components): Ditto.
17724         (rs6000_emit_epilogue_components): Ditto.
17725         (rs6000_emit_prologue): Ditto.
17726         (rs6000_emit_epilogue): Ditto.
17727         (rs6000_elf_file_end): Ditto.
17728         (rs6000_function_value): Ditto.
17729         (rs6000_libcall_value): Ditto.
17730         * config/rs6000/rs6000.h: Ditto.
17731         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
17732         (TARGET_MINMAX): ... this.  New.
17733         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
17734         * config/rs6000/rs6000.md: Remove Xilinx FP support.
17735         (*movsi_internal1_single): Delete.
17736         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
17737         mfpu=, mxilinx-fpu): Delete.
17738         * config/rs6000/singlefp.h: Delete.
17739         * config/rs6000/sysv4.h: Remove Xilinx FP support.
17740         * config/rs6000/t-rs6000: Ditto.
17741         * config/rs6000/t-xilinx: Delete.
17742         * config/rs6000/titan.md: Adjust for fp_type removal.
17743         * config/rs6000/vsx.md: Remove Xilinx FP support.
17744         (VStype_simple): Delete.
17745         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
17746         * config/rs6000/xfpu.h: Delete.
17747         * config/rs6000/xfpu.md: Delete.
17748         * config/rs6000/xilinx.h: Delete.
17749         * config/rs6000/xilinx.opt: Delete.
17750         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
17751         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
17753 2018-05-04  Tom de Vries  <tom@codesourcery.com>
17755         PR libgomp/85639
17756         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
17757         if ignore == 0.
17759 2018-05-04  Richard Biener  <rguenther@suse.de>
17761         PR middle-end/85627
17762         * tree-complex.c (update_complex_assignment): We are always in SSA form.
17763         (expand_complex_div_wide): Likewise.
17764         (expand_complex_operations_1): Likewise.
17765         (expand_complex_libcall): Preserve EH info of the original stmt.
17766         (tree_lower_complex): Handle removed blocks.
17767         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
17768         on complex multiplication and division libcall builtins.
17770 2018-05-04  Richard Biener  <rguenther@suse.de>
17772         PR middle-end/85574
17773         * fold-const.c (negate_expr_p): Restrict negation of operand
17774         zero of a division to when we know that can happen without
17775         overflow.
17776         (fold_negate_expr_1): Likewise.
17778 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
17780         PR libstdc++/85466
17781         * real.h (real_nextafter): Declare.
17782         * real.c (real_nextafter): New function.
17783         * fold-const-call.c (fold_const_nextafter): New function.
17784         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
17785         CASE_CFN_NEXTTOWARD.
17786         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
17787         even when arg1_mode is different from arg0_mode.
17789 2018-05-03  Nathan Sidwell  <nathan@acm.org>
17791         * doc/extend.texi (Deprecated Features): Remove
17792         -ffriend-injection.
17793         (Backwards Compatibility): Likewise.
17794         * doc/invoke.texi (C++ Language Options): Likewise.
17795         (C++ Dialect Options): Likewise.
17797 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
17799         PR target/85530
17800         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17801         _mm512_mask_mullox_epi64): New intrinsics.
17803 2018-05-03  Tom de Vries  <tom@codesourcery.com>
17805         PR testsuite/85106
17806         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17807         dump files): Add offload-tree.
17809 2018-05-03  Richard Biener  <rguenther@suse.de>
17811         PR tree-optimization/85615
17812         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17813         to loops not nested in BBs loop father to avoid creating multi-entry
17814         loops.
17816 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17818         PR tree-optimization/70291
17819         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17820         arguments.  Change return type to tree.  Emit libcall as a new
17821         statement rather than replacing existing one when inplace_p is true.
17822         (expand_complex_multiplication_components): New function.
17823         (expand_complex_multiplication): Expand floating-point complex
17824         multiplication using the above.
17825         (expand_complex_division): Rename inner_type parameter to type.
17826         Update expand_complex_libcall call-site.
17827         (expand_complex_operations_1): Update expand_complex_multiplication
17828         and expand_complex_division call-sites.
17830 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17832         PR target/85582
17833         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17834         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17835         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17836         the highest significant bit of the shift count mask is clear.  In
17837         check whether and[sq]i3 is needed verify that all significant bits
17838         of the shift count other than the highest are set.
17840 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17842         PR libgomp/82428
17843         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17844         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17845         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17846         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17847         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17848         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17849         __builtin_goacc_parlevel_size.
17851 2018-05-02  Richard Biener  <rguenther@suse.de>
17853         PR tree-optimization/85597
17854         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17855         do not use split vect_get_vec_defs call but call vect_get_slp_defs
17856         directly.
17858 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17860         PR testsuite/85106
17861         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17862         dump files): Add ltrans-tree.
17864 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17866         PR testsuite/85106
17867         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17868         dump files): Add wpa-ipa.
17870 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
17872         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
17873         powerpc*-*-linux*paired* target.
17874         * config/rs6000/750cl.h: Delete.
17875         * config/rs6000/paired.h: Delete.
17876         * config/rs6000/paired.md: Delete.
17877         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17878         float support.
17879         * config/rs6000/rs6000-builtin.def: Remove paired float support.
17880         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17881         comment.  Remove paired float support.
17882         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17883         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17884         VECTOR_PAIRED.
17885         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17886         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17887         declarations.
17888         * config/rs6000/rs6000.c: Remove paired float support.
17889         (paired_expand_vector_init, paired_expand_vector_move,
17890         paired_emit_vector_compare, paired_emit_vector_cond_expr,
17891         (paired_expand_lv_builtin, paired_expand_stv_builtin,
17892         paired_expand_builtin, paired_expand_predicate_builtin,
17893         paired_init_builtins): Delete.
17894         * config/rs6000/rs6000.h: Remove paired float support.
17895         * config/rs6000/rs6000.md: Remove paired float support.
17896         (move_from_CR_ov_bit): Delete.
17897         * config/rs6000/rs6000.opt (mpaired): Delete.
17898         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17899         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17901 2018-05-02  Richard Biener  <rguenther@suse.de>
17903         PR middle-end/85567
17904         * gimplify.c (gimplify_save_expr): When in SSA form allow
17905         SAVE_EXPRs to compute to SSA vars.
17907 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17909         PR target/85582
17910         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17911         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17912         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17913         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17915 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17917         PR tree-optimization/85586
17918         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17919         exit early for statements in the same group if the accesses are
17920         not strided.
17922 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17924         PR lto/85451
17925         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17926         error message.
17928 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17930         PR tree-optimization/85143
17931         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17933 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17935         PR lto/85451
17936         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17937         not found" error message.
17939 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17941         PR other/83786
17942         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17943         * vec.c (test_ordered_remove_if): New function.
17944         (vec_c_tests): Call test_ordered_remove_if.
17945         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17946         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17947         * tree-vect-patterns.c (vect_pattern_recog_1): Use
17948         VEC_ORDERED_REMOVE_IF.
17950 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17952         PR tree-optimization/82665
17953         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17954         pointer subtraction where arguments come from a memchr call.
17956 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
17958         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17959         --push-state --as-needed and --pop-state instead of --as-needed and
17960         --no-as-needed if ld supports it.
17961         * configure: Regenerated.
17963         PR web/85578
17964         * doc/install.texi2html: Replace _002d with - and _002a with * in
17965         generated html files using sed.
17967 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17969         PR c++/85523
17970         * gcc-rich-location.c (blank_line_before_p): New function.
17971         (use_new_line): New function.
17972         (gcc_rich_location::add_fixit_insert_formatted): New function.
17973         * gcc-rich-location.h
17974         (gcc_rich_location::add_fixit_insert_formatted): New function.
17976 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17978         * selftest.c (assert_streq): Rename "expected" and "actual" to
17979         "val1" and "val2".  Extend NULL-handling to cover both inputs
17980         symmetrically, while still requiring both to be non-NULL for a pass.
17981         * selftest.h (assert_streq): Rename "expected" and "actual" to
17982         "val1" and "val2".
17983         (ASSERT_EQ): Likewise.
17984         (ASSERT_EQ_AT): Likewise.
17985         (ASSERT_KNOWN_EQ): Likewise.
17986         (ASSERT_KNOWN_EQ_AT): Likewise.
17987         (ASSERT_NE): Likewise.
17988         (ASSERT_MAYBE_NE): Likewise.
17989         (ASSERT_MAYBE_NE_AT): Likewise.
17990         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
17991         the assertion to pass.
17992         (ASSERT_STREQ_AT): Likewise.
17994 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17996         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17997         interaction with -pie.
17999 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18001         * selftest.h: Fix alphabetization of per-source-file selftest
18002         declarations.
18004 2018-04-30  Jason Merrill  <jason@redhat.com>
18006         PR c++/61982 - dead stores to destroyed objects.
18007         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18008         of clobber.
18010 2018-04-30  Jason Merrill  <jason@redhat.com>
18012         * tree.c (build_clobber): New.
18013         * tree.h: Declare it.
18014         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18016 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18018         * diagnostic-show-locus.c (layout::layout): Update for
18019         location_get_source_line returning a char_span.
18020         (struct char_span): Move to input.h.
18021         (struct correction): Update for fields in char_span becoming
18022         private.
18023         (struct source_line): Update for location_get_source_line
18024         returning a char_span.
18025         (layout::print_line): Likewise.
18026         * edit-context.c (edited_file::print_content): Likewise.
18027         (edited_file::print_diff_hunk): Likewise.
18028         (edited_file::print_run_of_changed_lines): Likewise.
18029         (edited_file::get_num_lines): Likewise.
18030         (edited_line::edited_line): Likewise.
18031         * final.c (asm_show_source): Likewise.
18032         * input.c (location_get_source_line): Convert return type
18033         from const char * to char_span, losing the final "line_len"
18034         param.
18035         (dump_location_info): Update for the above.
18036         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18037         when handling the literal within the line.
18038         (test_reading_source_line): Update for location_get_source_line
18039         returning a char_span.
18040         * input.h (class char_span): Move here from
18041         diagnostic-show-locus.c, converting from a struct to a class.
18042         Make data members private.
18043         (char_span::operator bool): New.
18044         (char_span::length): New.
18045         (char_span::get_buffer): New.
18046         (char_span::operator[]): New.
18047         (char_span::subspan): Make const.
18048         (char_span::xstrdup): New.
18049         (location_get_source_line): Convert return type from const char *
18050         to char_span, losing the final "line_size" param.
18052 2018-04-30  Jan Hubicka  <jh@suse.cz>
18054         * lto-wrapper.c (ltrans_priorities): New static var.
18055         (cmp_priority): New.
18056         (run_gcc): Read priorities and if doing parallel build order
18057         the Makefile by them.
18059 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18061         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18063 2018-04-30  Richard Biener  <rguenther@suse.de>
18065         * tree-cfg.c (verify_address): Remove base argument, add
18066         flag whether to check TREE_ADDRESSABLE and do that.
18067         (verify_expr): Remove.
18068         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18069         (verify_gimple_assign_single): Likewise.
18070         (verify_gimple_switch): Likewise.
18071         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18072         piece from verify_expr.
18073         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18075 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18077         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18079 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18081         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18082         (small_data_pattern): Likewise.
18083         (arc_rewrite_small_data): Likewise.
18084         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18085         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18086         (get_symbol_alignment): New function.
18087         (legitimate_small_data_address_p): Likewise.
18088         (legitimate_scaled_address): Update, call
18089         legitimate_small_data_address_p.
18090         (output_sdata): New static variable.
18091         (arc_print_operand): Update how we handle small data operands.
18092         (arc_print_operand_address): Likewise.
18093         (arc_legitimate_address_p): Update, use
18094         legitimate_small_data_address_p.
18095         (arc_rewrite_small_data_p): Remove.
18096         (arc_rewrite_small_data_1): Likewise.
18097         (arc_rewrite_small_data): Likewise.
18098         (small_data_pattern): Likewise.
18099         (compact_sda_memory_operand): Update to use
18100         legitimate_small_data_address_p and get_symbol_alignment.
18101         (prepare_move_operands): Don't rewite sdata pattern.
18102         (prepare_extend_operands): Remove.
18103         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18104         pattern.
18105         (zero_extendqisi2): Likewise.
18106         (zero_extendhisi2): Likewise.
18107         (extendqihi2): Likewise.
18108         (extendqisi2): Likewise.
18109         (extendhisi2): Likewise.
18110         (addsi3): Likewise.
18111         (subsi3): Likewise.
18112         (andsi3): Likewise.
18113         * config/arc/constraints.md (Usd): Change it to memory constraint.
18115 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18117         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18118         as source of std instructions.
18119         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18120         allow 6-bit constants as source for store instructions.
18121         (movdi_insn): Update instruction pattern to allow 6-bit constants
18122         as source for store instructions.
18124 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18126         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18128 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18129             Sandra Loosemore <sandra@codesourcery.com>
18131         * dumpfile.c (dump_open): Allow '-' for stdout.
18132         * doc/invoke.texi (Developer Options): Document dump filename
18133         determination early.  Document stdin/stdout selection.
18135 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18137         Microblaze Target: PIC data text relative
18139         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18140         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18141         Add declaration.
18142         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18143         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18144         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18145         New addressing mode for data-text relative position indepenedent code.
18146         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18147         'ADDRESS_SYMBOLIC_TXT_REL'.
18148         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18149         (microblaze_legitimate_pic_operand): Exclude function calls from
18150         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18151         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18152         addresses cases.
18153         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18154         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18155         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18156         for 'address + offset'.
18157         (microblaze_expand_prologue): Add new function prologue call for
18158         'r20' assignation.
18159         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18160         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18161         table in case of TARGET_PIC_DATA_TEXT_REL.
18162         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18163         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18164         Add new macros 'UNSPEC_TEXT',
18165         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18166         + exclude function calls from 'UNSPEC_PLT' in case of data text
18167         relative mode.
18168         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18169         new target hook for generating address diff vector tables in case of
18170         flag_pic.
18171         * doc/tm.texi : Regenerate.
18172         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18173         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18174         of addr diff vector generation.
18175         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18176         target hook definition.
18177         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18178         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18179         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18180         option for data text relative.
18182 2018-04-30  Richard Biener  <rguenther@suse.de>
18184         * tree-chrec.h (evolution_function_is_constant_p): Remove
18185         redundant check.
18186         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18188 2018-04-30  Richard Biener  <rguenther@suse.de>
18190         PR bootstrap/85571
18191         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18193 2018-04-30  Richard Biener  <rguenther@suse.de>
18195         PR tree-optimization/28364
18196         PR tree-optimization/85275
18197         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18198         copying first exit test.
18200 2018-04-28  Mark Wielaard  <mark@klomp.org>
18202         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18203         dwarf_version >= 5.
18204         (dwarf_AT): Handle DW_AT_addr_base.
18205         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18207 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18209         PR target/84431
18210         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18211         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18212         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18213         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18215 2018-04-28  Richard Biener  <rguenther@suse.de>
18217         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18218         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18219         to reflect use.  Only add interesting stmts.
18221 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18223         PR ipa/85549
18224         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18225         the jump function allows for passing through aggregate values.
18227 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18229         * input.h (in_system_header_at): Convert from macro to inline
18230         function.
18231         (from_macro_expansion_at): Likewise.
18232         (from_macro_definition_at): Likewise.
18234 2018-04-27  Jeff Law  <law@redhat.com>
18236         * config.gcc: Mark tile* targets as deprecated/obsolete.
18238 2018-04-27  Richard Biener  <rguenther@suse.de>
18240         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18241         fix for ILP32.
18243 2018-04-27  Richard Biener  <rguenther@suse.de>
18245         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18247 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18249         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18250         with Yd constraint. Set "preferred_for_speed" attribute from
18251         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18252         with Yd constraint.
18253         (*movdi_internal): Ditto.
18254         (movti_interunit splitters): Remove
18255         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18256         (movdi_interunit splitters): Ditto.
18257         * config/i386/constraints.md (Ye): Remove.
18258         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18260 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18262         PR target/85512
18263         * config/aarch64/constraints.md (Usg): Limit to 31.
18264         (Usj): Limit to 63.
18266 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18268         PR tree-optimization/85529
18269         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18270         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18271         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18272         zero extension or masking of the MSB bit.
18273         (optimize_range_tests): Add FIRST_BB argument, pass it through
18274         to optimize_range_tests_var_bound.
18275         (maybe_optimize_range_tests, reassociate_bb): Adjust
18276         optimize_range_tests callers.
18278 2018-04-26  Richard Biener  <rguenther@suse.de>
18279             Jakub Jelinek  <jakub@redhat.com>
18281         * cgraph.h (symbol_table): Just declare debug method here.
18282         * symtab.c (symbol_table::debug): Define.
18284 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18286         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18288 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18290         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18291         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18292         (*movdi_internal): Substitute Yi and Yj constraint with x
18293         and Ym and Yn constraint with y constraint.  Update "isa"
18294         attribute and set "preferred_for_speed" attribute from
18295         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18296         (*movsi_internal): Ditto.
18297         (*movdf_internal): Ditto.
18298         (*movsf_internal): Ditto.
18299         (*zero_extendsidi2): Ditto.
18300         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18301         (sse2_loadld): Ditto.
18302         (*vec_extract<ssevecmodelower>_0): Ditto.
18303         (*vec_extractv4si_0_zext_sse4): Ditto.
18304         (vec_concatv2di): Ditto.
18305         (*vec_dup<mode>): Ditto.
18306         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18307         * config/i386/constraints.md (Yi): Remove.
18308         (Yj): Remove.
18309         (Ym): Remove.
18310         (Yn): Remove.
18312 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18314         * dumpfile.c (dump_open): New.
18315         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18316         (dump_finish): Detect stdio/stderr by value not name.
18318 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18320         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18322 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18324         PR target/84952
18325         * config/nvptx/nvptx.c (verify_neutering_jumps)
18326         (verify_neutering_labels): New function
18327         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18329 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18331         PR target/84025
18332         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18333         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18334         need neutering.
18336 2018-04-26  Richard Biener <rguenther@suse.de>
18337             Tom de Vries  <tom@codesourcery.com>
18339         PR lto/85422
18340         * lto-streamer-out.c (output_function): Fixup loops if required to match
18341         discovery done in the reader.
18343 2018-04-26  Richard Biener  <rguenther@suse.de>
18345         PR tree-optimization/85116
18346         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18347         have a loop exit from the single latch predecessor.  Remove
18348         case of header with just condition.
18349         (ch_base::copy_headers): Exclude infinite loops from any
18350         processing.
18351         (pass_ch::execute): Record exits.
18353 2018-04-26  Richard Biener  <rguenther@suse.de>
18355         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18356         prologue cost vector and pass it to vect_get_load_cost.
18357         (vect_get_peeling_costs_all_drs): Likewise.
18358         (vect_peeling_hash_get_lowest_cost): Likewise.
18359         (vect_enhance_data_refs_alignment): Likewise.
18361 2018-04-26  Richard Biener  <rguenther@suse.de>
18363         PR middle-end/85450
18364         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18365         checking of integer<->pointer conversions.
18366         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18367         sign-/zero-extending pointer types.
18368         (expand_omp_for_static_chunk): Likewise.
18370 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18371             Jean Lee  <xiaoyur347@gmail.com>
18373         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18374         (TARGET_ASAN_SHADOW_OFFSET): Define.
18375         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18376         true for -fsanitize=address.
18378 2018-04-25  Mark Wielaard  <mark@klomp.org>
18380         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18381         shorter ones.
18383 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18385         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18386         than "alu", remove explicit "memory" and "imm_disp" attributes.
18387         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18389         PR middle-end/85414
18390         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18391         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18392         gen_lowpart_no_emit.
18394 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
18396         PR target/85473
18397         * config/i386/i386.c (ix86_expand_builtin): Change memory
18398         operand to XI, extend p0 to Pmode.
18399         * config/i386/i386.md: Change unspec volatile and operand
18400         1 mode to XI, change operand 0 mode to P.
18402 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18404         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18405         GET_MODE_MASK before any checking.
18406         (nds32_can_use_bset_p): Likewise.
18407         (nds32_can_use_btgl_p): Likewise.
18409 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18411         * config/nds32/nds32-doubleword.md: New define_split pattern for
18412         illegal register number.
18414 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18416         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18418 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18420         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18422 2018-04-25  Richard Biener  <rguenther@suse.de>
18424         * lto-streamer.h (LTO_major_version): Bump to 8.
18426 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18428         * BASE-VER: Set to 9.0.0.
18430 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
18432         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18433         in __abskf2 and __powikf2.
18435 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18437         PR target/85512
18438         * config/aarch64/constraints.md (Usg, Usj): New constraints.
18439         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18440         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18441         Use the above on operand 2.  Reindent.
18442         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18444 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18446         PR target/85485
18447         * common/config/i386/i386-common.c (ix86_handle_option): Don't
18448         handle OPT_mcet.
18449         * config/i386/i386.opt (mcet): Removed.
18450         * doc/install.texi: Remove -mcet documentation.
18451         * doc/invoke.texi: Likewise.
18453 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18455         PR target/85485
18456         * doc/install.texi: Remove -mcet from bootstrap-cet.
18458 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18460         PR target/85511
18461         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18462         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18463         if TARGET_64BIT.
18465         PR target/85503
18466         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18467         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18468         containing a CONST_VECTOR.
18470 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
18472         * doc/install.texi: Update newlib dependency for nvptx.
18474 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18476         PR target/85508
18477         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18478         instead of INTVAL when shifting x left.
18480 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
18482         PR tree-optimization/85478
18483         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18484         vect_grouped_store_supported for single element vectors.
18486 2018-04-24  Richard Biener  <rguenther@suse.de>
18488         PR target/85491
18489         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18490         load cost increase to the case of non-constant step.
18492 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18494         PR target/84828
18495         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18496         destination if any_malformed_asm.
18498 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
18500         PR middle-end/85496
18501         * expr.c (store_field): In the bitfield case, if the value comes from
18502         a function call and is returned in registers by means of a PARALLEL,
18503         do not change the mode of the temporary unless BLKmode and VOIDmode.
18505 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
18507         PR rtl-optimization/85423
18508         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
18509         dependencies to debug insns when the previous insn is non-debug.
18511 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18513         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
18514         enums into a single definition.
18515         (fls): Fix predicates and printing.
18516         (seti): Likewise.
18518 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18520         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
18521         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
18522         and short u6 immediate.
18523         (check_if_valid_sleep_operand): Remove.
18524         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
18526 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18528         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
18529         flag_always_save_lp condition.
18530         * config/nds32/nds32.opt (malways-save-lp): New option.
18532 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18534         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
18535         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
18536         * config/nds32/nds32.h
18537         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
18538         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
18540 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18542         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
18543         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
18545 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18546             Chung-Ju Wu  <jasonwucj@gmail.com>
18548         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
18549         Declare.
18550         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
18551         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
18553 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18555         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
18557 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18559         * config/nds32/nds32-protos.h (nds32_data_alignment,
18560         nds32_local_alignment): Declare.
18561         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
18562         nds32_local_alignment): New functions.
18563         (TARGET_CONSTANT_ALIGNMENT): Define.
18564         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
18566 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18568         * config/nds32/nds32.c
18569         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18570         (TARGET_MODES_TIEABLE_P): Likewise.
18572 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18574         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
18575         level Ofast and Og.
18577 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
18578             Chung-Ju Wu  <jasonwucj@gmail.com>
18580         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18581         for unaligned access.
18582         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18583         unaligned access.
18584         * config/nds32/nds32-intrinsic.md: Likewise.
18585         * config/nds32/nds32_intrinsic.h: Likewise.
18586         * config/nds32/nds32.h (nds32_builtins): Likewise.
18587         * config/nds32/nds32.opt (munaligned-access): New option.
18588         * config/nds32/nds32.c (nds32_asm_file_start): Display
18589         flag_unaligned_access status.
18591 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18593         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
18594         -mno-relax is present.
18595         * config/riscv/linux.h (LINK_SPEC): Ditto.
18597 2018-04-20  Martin Sebor  <msebor@redhat.com>
18599         PR c/85365
18600         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
18601         for null pointers.
18602         (gimple_fold_builtin_stxcpy_chk): Same.
18603         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
18605 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
18607         PR target/85456
18608         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
18609         __powikf2 when long double is IEEE 128-bit.
18611 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18613         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
18614         step to make sure stack always aligned.
18616 2018-04-20  Carl Love  <cel@us.ibm.com>
18618         PR target/83402
18619         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
18620         size check for arg0.
18622 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
18623             Tom de Vries  <tom@codesourcery.com>
18625         PR target/85445
18626         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
18627         Emit insns for calls too.
18628         (nvptx_find_par): Always look for worker-level predecessor insn.
18629         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
18630         calls.
18631         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
18632         (nvptx_process_pars): Propagate frames for calls.
18634 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
18636         PR target/85469
18637         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
18638         Removed.
18639         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
18640         (ix86_handle_option): Don't handle OPT_mibt.
18641         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
18642         __SHSTK__.
18643         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
18644         has_ibt and ibt.
18645         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
18646         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
18647         (ix86_target_macros): Define __CET__ with flag_cf_protection
18648         for -fcf-protection.
18649         * config/i386/i386.c (isa2_opts): Remove -mibt.
18650         * config/i386/i386.h (TARGET_IBT): Removed.
18651         (TARGET_IBT_P): Likewise.
18652         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
18653         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
18654         * config/i386/i386.opt (mcet): Update help message.
18655         (mshstk): Likewise.
18656         (mibt): Removed.
18657         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
18658         -mcet as an alias for -mshstk.
18660 2018-04-20  Richard Biener <rguenther@suse.de>
18662         PR middle-end/85475
18663         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
18664         complexity by forcing a single use of the multiply operand.
18666 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18668         ipa/85449
18669         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
18670         recursion dependency to only apply to non-clones.
18672 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18674         ipa/85447
18675         * ipa-cp.c (create_specialized_node): Check that clones of
18676         self-recursive edges exist during IPA-CP.
18678 2018-04-19  Toon Moene  <toon@moene.org>
18680         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
18681         by -O3.
18683 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18685         PR tree-optimization/85467
18686         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
18687         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
18688         VECTOR_CST element to type.
18690 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18692         PR target/85397
18693         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
18694         * config/i386/i386.md (builtin_setjmp_setup): Removed.
18695         (builtin_longjmp): Likewise.
18696         (save_stack_nonlocal): New pattern.
18697         (restore_stack_nonlocal): Likewise.
18699 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18701         PR target/85404
18702         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18703         Replace ASM_OUTPUT_LABEL with fprintf.
18705 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18707         PR target/85417
18708         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18709         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
18710         * config/i386/i386-c.c (ix86_target_macros_internal): Also
18711         define __IBT__ and __SHSTK__ for -fcf-protection.
18712         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
18713         TARGET_IBT.
18714         (ix86_trampoline_init): Likewise.
18715         (x86_output_mi_thunk): Likewise.
18716         (ix86_notrack_prefixed_insn_p): Likewise.
18717         (ix86_option_override_internal): Don't disallow -fcf-protection.
18718         * config/i386/i386.md (rdssp<mode>): Also enable for
18719         -fcf-protection.
18720         (incssp<mode>): Likewise.
18721         (nop_endbr): Likewise.
18722         * config/i386/i386.opt (mcet): Change help message to built-in
18723         functions only.
18724         (mibt): Likewise.
18725         (mshstk): Likewise.
18726         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
18727         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
18728         enable CET built-in functions.
18730 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
18732         * common/config/i386/i386-common.c
18733         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
18734         OPTION_MASK_ISA_MOVDIRI_UNSET,
18735         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
18736         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
18737         * config.gcc (movdirintrin.h): New header.
18738         * config/i386/cpuid.h (bit_MOVDIRI,
18739         bit_MOVDIR64B): New bits.
18740         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
18741         and -mmvodir64b.
18742         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
18743         (VOID, PVOID, PCVOID)): New function types.
18744         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
18745         __builtin_ia32_directstoreu_u64,
18746         __builtin_ia32_movdir64b): New builtins.
18747         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
18748         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
18749         and -mmovdiri.
18750         (ix86_valid_target_attribute_inner_p): Ditto.
18751         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
18752         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
18753         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
18754         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
18755         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
18756         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
18757         (movdiri<mode>, movdir64b_<mode>): New.
18758         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
18759         * config/i386/immintrin.h: Include movdirintrin.h.
18760         * config/i386/movdirintrin.h: New file.
18761         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
18763 2018-04-19  Richard Biener  <rguenther@suse.de>
18765         PR middle-end/85455
18766         * cfg.c (clear_bb_flags): When loop state says we have
18767         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
18769 2018-04-19  Richard Biener  <rguenther@suse.de>
18771         PR tree-optimization/84737
18772         * tree-vect-data-refs.c (vect_copy_ref_info): New function
18773         copying restrict info.
18774         (vect_setup_realignment): Use it.
18775         * tree-vectorizer.h (vect_copy_ref_info): Declare.
18776         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
18777         the first DR to all generated stores.
18778         (vectorizable_load): Likewise for loads.
18780 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18782         PR tree-optimization/85446
18783         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
18784         the integral and pointer types to have the same precision.
18786         * doc/install.texi: Document --disable-cet being the default and
18787         --enable-cet=auto.
18789 2018-04-18  Martin Liska  <mliska@suse.cz>
18791         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18792         style.
18794 2018-04-18  Martin Liska  <mliska@suse.cz>
18796         Revert
18797         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18799         PR ipa/83983
18800         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18801         arguments if they are comparable.
18803 2018-04-18  Martin Liska  <mliska@suse.cz>
18805         Revert
18806         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18808         PR lto/84805
18809         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18810         incomplete types.
18812 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
18814         PR target/85388
18815         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18816         ENDBR after calling __morestack.
18818 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
18820         PR jit/85384
18821         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18822         by using gcc_base_ver to generate a gcc_driver_version, and use
18823         it when generating GCC_DRIVER_NAME.
18824         * configure: Regenerate.
18826 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
18828         PR target/81084
18829         * config.gcc: Obsolete powerpc*-*-*spe*.
18831 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18833         PR debug/84637
18834         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18835         (stabstr_D): Change type of unum from unsigned int to
18836         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
18837         type.
18839 2018-04-17  Jim Wilson  <jimw@sifive.com>
18841         PR 84856
18842         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18843         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18844         Set arg_pointer_offset after using pretend_args_size.
18846 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18848         PR rtl-optimization/85431
18849         * dse.c (record_store): Ignore zero width stores.
18851         PR sanitizer/85230
18852         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
18853         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18854         __builtin_stack_restore rather than after it.
18855         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18856         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18857         argument instead of virtual_dynamic_stack_rtx.
18859 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18861         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18862         New prototype.
18863         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18864         Add note to error message to explain internal mapping of overloaded
18865         built-in function name to non-overloaded built-in function name.
18866         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18867         function.
18869 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18871         PR target/85424
18872         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18873         where the inputs overlap with the output.
18875 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18877         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18878         (=v, v) alternative and explicit "memory" attribute.
18879         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
18880         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18881         attributes.
18882         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18883         "sselog1" type instead of "sselog".
18884         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18885         "sselog".  Remove explicit "memory" attribute.
18886         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18887         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18888         attributes.
18889         (vec_extract_hi_v32hi): Merge all alternatives into one, use
18890         "sselog1" type instead of "sselog".  Remove explicit "memory"
18891         attribute.
18892         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18893         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18894         attribute.
18895         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18896         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18897         attributes.
18898         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18899         "sselog1" type instead of "sselog".  Remove explicit "memory"
18900         attribute.
18901         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18902         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18903         attribute.
18905         PR target/85430
18906         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18908         PR middle-end/85414
18909         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18910         on a SUBREG.
18912 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18914         PR ipa/85421
18915         * ipa-cp.c (create_specialized_node): Call
18916         expand_all_artificial_thunks if necessary.
18918 2018-04-17  Martin Liska  <mliska@suse.cz>
18920         PR lto/85405
18921         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18922         in message, remote space in between '_G' and '('.
18924 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18926         PR target/85281
18927         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18928         avx512f_vmcmp<mode>3<round_saeonly_name>,
18929         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18930         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18931         avx512f_rndscale<mode><round_saeonly_name>,
18932         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18933         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18934         Use %<iptr>2 instead of %2 for -masm=intel.
18935         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18936         avx512f_vcvttss2usi<round_saeonly_name>,
18937         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18938         -masm=intel.
18939         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18940         avx512f_vcvttsd2usi<round_saeonly_name>,
18941         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18942         Use %q1 instead of %1 for -masm=intel.
18943         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18944         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18945         of %3 for -masm=intel.
18946         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18947         -masm=intel.
18948         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18949         -masm=intel.
18950         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18951         -masm=intel.
18952         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18953         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18954         %g1.
18955         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18956         -masm=intel.
18957         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18958         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18959         %g1 and one with %0 and %1.
18960         (avx512er_vmrcp28<mode><round_saeonly_name>,
18961         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18962         %1 for -masm=intel.
18963         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18964         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18965         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18966         of %0 and %{%4%} for -masm=intel.
18967         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18968         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18969         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18970         order of %0 and %{%5%}%{z%} for -masm=intel.
18972 2018-04-17  Jan Hubicka  <jh@suse.cz>
18974         PR lto/85405
18975         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18977 2018-04-17  Martin Liska  <mliska@suse.cz>
18979         PR ipa/85329
18980         * multiple_target.c (create_dispatcher_calls): Set apostrophes
18981         for target_clone error message.  Make default implementation
18982         clone to be a local declaration.
18983         (separate_attrs): Add new argument and check for an empty
18984         string.
18985         (expand_target_clones): Handle it.
18986         (ipa_target_clone): Make redirection just for target_clones
18987         functions.
18989 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
18990             Tom de Vries  <tom@codesourcery.com>
18992         PR middle-end/84955
18993         * omp-expand.c (expand_oacc_for): Add dummy false branch for
18994         tiled basic blocks without omp continue statements.
18996 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
18998         PR target/83660
18999         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19000         vec_extract expression as having side effects to make sure it gets
19001         a cleanup point.
19003 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19005         PR target/85403
19006         * config/i386/i386.c (get_builtin_code_for_version): Check
19007         error_mark_node.
19009 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19011         PR target/84331
19012         * config.gcc: Support "skylake".
19013         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19014         PROCESSOR_SKYLAKE.
19015         * config/i386/i386.c (m_SKYLAKE): Define.
19016         (processor_target_table): Add "skylake".
19017         (ix86_option_override_internal): Add "skylake".
19018         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19019         PROCESSOR_CANNONLAKE.
19020         (get_builtin_code_for_version): Fix priority for
19021         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19022         PROCESSOR_SKYLAKE-AVX512.
19023         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19024         (processor_type): Add PROCESSOR_SKYLAKE.
19026 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19027             Jason Merrill  <jason@redhat.com>
19029         PR c++/85112
19030         * convert.c (convert_to_integer_1): Use direct recursion for
19031         enumeral types and types with a precision less than the number
19032         of bits in their mode.
19034 2018-04-16  Julia Koval  <julia.koval@intel.com>
19036         PR target/84413
19037         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19038         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19040 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19042         PR target/85293
19043         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19044         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19045         and -mno-direct-move.
19047 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19049         PR target/83402
19050         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19051         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19052         Ensure negative shifts result in {0}.
19054 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19056         PR rtl-optimization/79916
19057         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19058         regs (if any) to define how to gnerate SD moves when LRA is in
19059         progress.
19061 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19063         PR rtl-optimization/85393
19064         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19065         * except.c (expand_dw2_landing_pad_for_region): Make static.
19066         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19067         a label and unconditional jump to old_bb, rather than
19068         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19069         basic block.
19071         PR rtl-optimization/85376
19072         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19073         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19074         instead of a specific value.
19076 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19077             Bin Cheng  <bin.cheng@arm.com>
19079         PR tree-optimization/82965
19080         PR tree-optimization/83991
19081         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19082         by_profile_only parameter.
19083         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19084         information if the loop was predicted to iterate too many times.
19085         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19087 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19089         PR lto/71991
19090         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19091         always inline.
19093 2018-04-13  Martin Liska  <mliska@suse.cz>
19094             Jakub Jelinek  <jakub@redhat.com>
19096         PR middle-end/81657
19097         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19098         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19099         * builtins.c (expand_builtin_memory_copy_args): Use
19100         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19101         handle dest_addr == pc_rtx.
19103 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19105         PR target/85291
19106         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19107         asked to not generate direct moves.
19108         (fix_trunc<mode>si2_stfiwx): Similar.
19109         (fix_trunc<mode>si2_internal): Similar.
19111 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19113         PR debug/83157
19114         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19115         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19116         lookup if dest in some wider mode is known to be const0_rtx and
19117         if so, record permanent equivalence for it to be ZERO_EXTEND of
19118         the narrower mode destination.
19120 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19122         * lto-streamer-out.c (output_function): Revert 259346.
19123         * omp-expand.c (expand_oacc_for): Likewise.
19125 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19127         PR rtl-optimization/85354
19128         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19129         * sel-sched.c (sel_global_init): ... here.
19131 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19133         PR target/85238
19134         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19135         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19136         mode for PE-COFF targets.
19137         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19138         (i386_pe_asm_lto_end): Likewise.
19139         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19140         (TARGET_ASM_LTO_END): Likewise.
19141         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19142         (i386_pe_asm_lto_start): New function.
19143         (i386_pe_asm_lto_end): Likewise.
19145 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19146             Richard Biener  <rguenther@suse.de>
19148         PR middle-end/84955
19149         * lto-streamer-out.c (output_function): Fix CFG loop state before
19150         streaming out.
19151         * omp-expand.c (expand_oacc_for): Handle calls to internal
19152         functions like regular functions.
19154 2018-04-12  Richard Biener  <rguenther@suse.de>
19156         PR lto/85371
19157         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19158         for the early LTO debug to properly generate references to it
19159         during DIE emission.  Do not re-use that for the skeleton for
19160         split-dwarf.
19161         (dwarf2out_early_finish): Likewise.
19163 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19165         PR target/85328
19166         * config/i386/sse.md
19167         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19168         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19169         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19170         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19171         and output is a reg, avoid creating invalid lowpart subreg, but
19172         instead split into a 512-bit move.  Don't split if not AVX512VL,
19173         input is xmm16+ reg and output is a mem.
19174         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19175         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19176         xmm16+ reg and output is a mem.
19178 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19180         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19181         also for flag_dwarf2_cfi_asm.
19183 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19185         PR rtl-optimization/85342
19186         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19187         a bool scalar var inside of the loop instead.  Don't try to update
19188         recog_data.operand after failed apply_change_group.
19190 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19192         PR target/85296
19193         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19194         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19195         array with flexible array member as array without given dimension.
19196         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19197         argument for undefined param to true.
19199 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19201         PR target/85321
19202         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19203         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19204         from PowerPC section.
19205         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19206         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19207         help text that is too long.
19208         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19209         help text that is too long.
19210         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19211         help text that is too long.
19213 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19215         * config/alpha/alpha.md (stack_probe_internal): Rename
19216         from "probe_stack".  Update all callers.
19218 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19220         PR rtl-optimization/84566
19221         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19222         sched_macro_fuse_insns.
19224 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19226         PR target/84301
19227         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19228         (compute_block_dependences): ... from here.
19230 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19232         PR tree-optimization/85331
19233         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19234         from int to HOST_WIDE_INT.
19236 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19238         PR ipa/84149
19239         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19240         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19241         not the same as the source val.
19242         (cgraph_edge_brings_value_p): New parameter.
19243         (gather_edges_for_value): Pass destination value to
19244         cgraph_edge_brings_value_p.
19245         (perhaps_add_new_callers): Likewise.
19246         (get_info_about_necessary_edges): Likewise and exclude values brought
19247         only by self-recursive edges.
19248         (create_specialized_node): Redirect only clones of self-calling edges.
19249         (+self_recursive_pass_through_p): New function.
19250         (find_more_scalar_values_for_callers_subset): Use it.
19251         (find_aggregate_values_for_callers_subset): Likewise.
19252         (known_aggs_to_agg_replacement_list): Removed.
19253         (decide_whether_version_node): Re-calculate known constants for all
19254         remaining context clones.
19256 2018-04-11  Richard Biener  <rguenther@suse.de>
19258         PR lto/85339
19259         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19260         from early DWARF output.
19261         (dwarf2out_early_finish): Output line info unconditionally into
19262         early DWARF and add reference to it.
19264 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19266         PR target/85281
19267         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19268         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19269         other than V2DFmode using iptr mode attribute.
19270         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19272 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19274         PR rtl-optimization/84659
19275         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19277 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19279         PR debug/85302
19280         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19281         SIZEP is NULL.
19282         (output_loc_list): Pass address of a dummy size variable even in the
19283         locview handling loop.
19284         (index_location_lists): Add comment on why skip_loc_list_entry can't
19285         call size_of_locs.
19287 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19289         PR target/85261
19290         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19291         into register.
19293 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19295         PR target/85321
19296         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19297         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19298         and -mstring-compare-inline-limit.
19300 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19302         PR target/85287
19303         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19304         for stack clash protection in a register whenever we need it to be in
19305         a register.
19307 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19309         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19310         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19312 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19314         PR target/85321
19315         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19316         the help text.
19317         (mlong-double-): Ditto.
19318         * config/rs6000/sysv4.opt (msdata=): Ditto.
19319         (mtls-size=): Ditto.
19321 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19323         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19324         erroneous entries for
19325         "vector int vec_ldl (int, long int *)", and
19326         "vector unsigned int vec_ldl (int, unsigned long int *)".
19327         Add comments and entries for
19328         "vector bool char vec_ldl (int, bool char *)",
19329         "vector bool short vec_ldl (int, bool short *)",
19330         "vector bool int vec_ldl (int, bool int *)",
19331         "vector bool long long vec_ldl (int, bool long long *)",
19332         "vector pixel vec_ldl (int, pixel *)",
19333         "vector long long vec_ldl (int, long long *)",
19334         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19335         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19336         type tree bool_long_long_type_node and correct definition of
19337         bool_V2DI_type_node to make reference to this new type tree.
19338         (rs6000_mangle_type): Replace erroneous reference to
19339         bool_long_type_node with bool_long_long_type_node.
19340         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19341         comments to emphasize sign distinctions for char and int types and
19342         replace RS6000_BTI_bool_long constant with
19343         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19344         use of RS6000_BTI_pixel.
19345         (bool_long_type_node): Remove this macro definition.
19346         (bool_long_long_type_node): New macro definition
19348 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19350         PR rtl-optimization/85300
19351         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19352         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19353         simplify_unary_operation fails.
19355 2018-04-10  Martin Liska  <mliska@suse.cz>
19357         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19358         cgraph_edge and ipa_ref.
19360 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19362         PR target/85177
19363         PR target/85255
19364         * config/i386/sse.md
19365         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19366         computation of the VEC_MERGE selector from mask.
19367         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19368         Fix decoding of the VEC_MERGE selector into mask.
19370 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19372         PR tree-optimization/85286
19373         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19375 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19377         * final.c (final_1): Set insn_last_address as well as
19378         insn_current_address.
19380 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19382         PR target/85173
19383         * explow.c (emit_stack_probe): Call validize_mem on memory location
19384         before passing it to gen_probe_stack.  Create address operand and
19385         legitimize it for the probe_stack_address case.
19387 2018-04-09  Jan Hubicka  <jh@suse.cz>
19389         PR lto/85078
19390         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19391         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19392         * tree.c (free_lang_data_in_type): Fix handling of binfos;
19393         walk basetypes.
19394         (free_lang_data): Rebuild type inheritance graph.
19396 2018-04-09  Martin Sebor  <msebor@redhat.com>
19398         * invoke.texi (-finline-small-functions): Mention other optimization
19399         options.
19400         (-findirect-inlining, -fpartial-inlining): Same.
19401         (-finline-functions-called-once): Same.
19402         (-freorder-blocks-and-partition): Same.
19404 2018-04-09  Jan Hubicka  <jh@suse.cz>
19406         PR rtl/84058
19407         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19408         jumps; choose last target that matches the criteria (i.e.
19409         no partition changes for non-crossing jumps).
19410         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19411         support for redirecting crossing jumps to non-crossing.
19413 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
19415         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19416         also for naked functions.
19418 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
19420         * config/arc/arc.md (add_shift): New pattern.
19421         (add_shift2): Likewise.
19422         (sub_shift): Likewise.
19423         (sub_shift_cmp0_noout): Likewise.
19424         (compare_si_ashiftsi): Likewise.
19425         (xbfu_cmp0_noout): New combine pattern.
19426         (xbfu_cmp0"): Likewise.
19427         (movsi_set_cc_insn): Place the predicable variant first.
19428         (commutative_binary_cmp0_noout): Remove clobber.
19429         (commutative_binary_cmp0): New pattern.
19430         (noncommutative_binary_cmp0): Likewise.
19431         (noncommutative_binary_cmp0_noout): Likewise.
19432         (noncommutative_binary_comparison_result_used): Removed.
19433         (rsub_cmp0): New pattern.
19434         (rsub_cmp0_noout): Likewise.
19435         (extzvsi): Changed, keep only meaningful variants.
19436         (SQH, SEZ): New iterators.
19437         (SQH_postfix): New mode attribute.
19438         (SEZ_prefix): New code attribute.
19439         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19440         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19441         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19442         of numerical value.
19443         (noncommutative_operator): Check the availability of barrel
19444         shifter option.
19446 2018-04-09  Richard Biener  <rguenther@suse.de>
19448         PR tree-optimization/85284
19449         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19450         Only use the niter constraining form of simple_iv when the exit
19451         is always executed.
19453 2018-04-09  Tom de Vries  <tom@codesourcery.com>
19455         PR target/84041
19456         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19457         (define_expand "*memory_barrier"): New define_expand.
19458         (define_insn "memory_barrier"): New insn.
19460 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19462         PR rtl-optimization/80463
19463         PR rtl-optimization/83972
19464         PR rtl-optimization/83480
19466         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19467         correct producer for the insn.
19468         (tidy_control_flow): Fixup seqnos in case of debug insns.
19470 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19472         PR rtl-optimization/83913
19474         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19475         different sched-times when merging exprs.
19477 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19479         PR rtl-optimization/83962
19481         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19482         tidy_fallthru_edge and tidy_control_flow.
19484 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19486         PR rtl-optimization/83530
19488         * sel-sched.c (force_next_insn): New global variable.
19489         (remove_insn_for_debug): When force_next_insn is true, also leave only
19490         next insn in the ready list.
19491         (sel_sched_region): When the region wasn't scheduled, make another pass
19492         over it with force_next_insn set to 1.
19494 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19496         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19497         into tm_file.
19498         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19499         for interrupt control.
19500         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19501         functions for interrupt control.
19502         * config/nds32/nds32-intrinsic.md: Likewise.
19503         * config/nds32/nds32_intrinsic.h: Likewise.
19504         * config/nds32/nds32.h (nds32_builtins): Likewise.
19506 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19508         * config/nds32/nds32.c (nds32_init_machine_status,
19509         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
19510         strict_aligned_p field.
19511         (nds32_expand_to_rtl_hook): New function.
19512         (TARGET_EXPAND_TO_RTL_HOOK): Define.
19513         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
19515 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19516             Chung-Ju Wu  <jasonwucj@gmail.com>
19518         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
19519         * config/nds32/nds32-n7.md: New file.
19520         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
19521         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
19522         pipeline.
19523         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19524         * config/nds32/nds32.md (pipeline_model): Add n7.
19525         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
19526         * config/nds32/pipelines.md: Include n7 settings.
19528 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19529             Chung-Ju Wu  <jasonwucj@gmail.com>
19531         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
19532         * config/nds32/nds32-e8.md: New file.
19533         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
19534         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
19535         pipeline.
19536         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
19537         * config/nds32/nds32.md (pipeline_model): Add e8.
19538         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
19539         * config/nds32/pipelines.md: Include e8 settings.
19541 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19542             Chung-Ju Wu  <jasonwucj@gmail.com>
19544         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
19545         * config/nds32/nds32-n8.md: New file.
19546         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
19547         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
19548         pipeline.
19549         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
19550         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
19551         * config/nds32/nds32.md (pipeline_model): Add n8.
19552         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
19553         * config/nds32/pipelines.md: Include n8 settings.
19555 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19556             Chung-Ju Wu  <jasonwucj@gmail.com>
19558         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
19559         * config/nds32/nds32-n9-2r1w.md: New file.
19560         * config/nds32/nds32-n9-3r2w.md: New file.
19561         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
19562         nds32_register_ports): New or modify for cpu n9.
19563         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
19564         pipeline.
19565         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
19566         * config/nds32/nds32-utils.c: New file.
19567         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
19568         TARGET_MUL_SLOW): Define.
19569         * config/nds32/nds32.md (pipeline_model): New attribute.
19570         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
19571         New options that support cpu n9.
19572         * config/nds32/pipelines.md: Include n9 settings.
19573         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
19575 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19577         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
19578         information if necessary.
19579         (output_cond_branch_compare_zero): Likewise.
19580         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
19581         (nds32_target_alignment): Refine for alignment.
19582         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
19583         (FUNCTION_BOUNDARY): Modify.
19584         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
19585         align case.
19586         * config/nds32/nds32.opt (malways-align, malign-functions): New.
19588 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19590         * config/nds32/constants.md (unspec_volatile_element): Add values for
19591         TLB operation and data prefetch.
19592         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19593         functions for TLB operation and data prefetch.
19594         * config/nds32/nds32-intrinsic.md: Likewise.
19595         * config/nds32/nds32_intrinsic.h: Likewise.
19596         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
19597         (nds32_print_operand): Likewise.
19598         * config/nds32/nds32.h (nds32_builtins): Likewise.
19600 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
19601         Andrew Pinski <pinsika@gcc.gnu.org>
19603         PR middle-end/82976
19604         * match.pd: Use constant_boolean_node of correct type instead of
19605         boolean_true_node or boolean_false_node for simplifying
19606         pointer comparisons to zero.
19608 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19610         PR tree-optimization/80021
19611         * tree.c (verify_type_variant): Make error call in verify_variant_match
19612         translatable and remove final full stop.
19614 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19616         * config/nds32/constants.md (unspec_volatile_element): Add
19617         UNSPEC_VOLATILE_EH_RETURN.
19618         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
19619         nds32_output_stack_pop): Support dwarf exception handling process.
19620         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
19621         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
19622         exception handling process.
19623         (nds32_compute_stack_frame): Likewise.
19624         (nds32_return_addr_rtx): Likewise.
19625         (nds32_initial_elimination_offset): Likewise.
19626         (nds32_expand_prologue): Likewise.
19627         (nds32_expand_epilogue): Likewise.
19628         (nds32_dynamic_chain_address): New function.
19629         * config/nds32/nds32.h (machine_function): Add fields for dwarf
19630         exception handling.
19631         (DYNAMIC_CHAIN_ADDRESS): Define.
19632         (EH_RETURN_DATA_REGNO): Define.
19633         (EH_RETURN_STACKADJ_RTX): Define.
19634         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
19635         patterns for dwarf exception handling.
19637 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19639         * config/nds32/nds32.h: Clean up obsolete macros.
19641 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19643         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19644         Add enum values for particular instructions.
19645         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19646         particular intrinsic functions.
19647         * config/nds32/nds32-intrinsic.md: Likewise.
19648         * config/nds32/nds32_intrinsic.h: Likewise.
19649         * config/nds32/nds32.h (nds32_builtins): Likewise.
19650         * config/nds32/nds32.md (type): Add pbsad and pbsada.
19651         (btst, ave): New patterns for particular instructions.
19653 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19655         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19656         Add enum values for atomic load/store and memory sync.
19657         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
19658         and memory sync.
19659         * config/nds32/nds32-intrinsic.md: Likewise.
19660         * config/nds32/nds32_intrinsic.h: Likewise.
19661         * config/nds32/nds32.h (nds32_builtins): Likewise.
19663 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19665         PR tree-optimization/85257
19666         * fold-const.c (native_encode_vector): If not all elts could fit
19667         and off is -1, return 0 rather than offset.
19668         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
19669         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
19670         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
19671         adjust buffer in native_interpret_expr call.
19673 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19675         * config/nds32/constants.md (unspec_volatile_element): Add cache
19676         control enum values.
19677         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
19678         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
19679         * config/nds32/nds32.c (nds32_cctl_names): New.
19680         (nds32_print_operand): Handle cache control register names.
19681         * config/nds32/nds32.h (nds32_builtins): New enum values.
19682         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
19683         macros.
19684         * config/nds32/nds32.md (type): Add mmu.
19685         * config/nds32/pipelines.md (simple_insn): Add mmu.
19687 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19689         * config/nds32/nds32.md (type): Remove call.
19690         * config/nds32/pipelines.md (simple_insn): Likewise.
19692 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19694         * config/nds32/constants.md (unspec_volatile_element): Add
19695         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
19696         UNSPEC_VOLATILE_FMFCFG.
19697         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
19698         description for fmfcfg and fmfcsr.
19699         (bdesc_1arg): Add fmtcsr.
19700         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
19701         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
19702         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
19703         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
19704         unspec_fmfcfg): New patterns.
19705         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
19706         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
19707         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
19708         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
19709         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
19710         __nds32__fmfcfg): Define.
19712 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19714         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
19715         intrinsic register names.
19716         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
19717         intrinsic register enum values and macros.
19719 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19721         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
19722         for load/store addressing form.
19723         (nds32_print_operand_address): Likewise.
19725 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
19727         PR target/85196
19728         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
19729         based on LABEL_REF.  Remove useless assertion.
19730         (pic_address_needs_scratch): Fix formatting.
19731         (sparc_legitimize_pic_address): Minor tweaks.
19732         (sparc_delegitimize_address): Adjust assertion accordingly.
19733         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
19734         into symbolic_operand.
19735         (movsi_high_pic_label_ref): Likewise.
19736         (movsi_lo_sum_pic_label_ref): Likewise.
19737         (movdi_pic_label_ref): Likewise.
19738         (movdi_high_pic_label_ref): Likewise.
19739         (movdi_lo_sum_pic_label_ref): Likewise.
19741 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
19743         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
19744         custom LIB_SPEC setup.
19746 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
19747             Kito Cheng  <kito.cheng@gmail.com>
19749         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
19750         * config/riscv/freebsd.h: New.
19752 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19754         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
19755         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
19756         file.
19758 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19759             Kito Cheng  <kito.cheng@gmail.com>
19761         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
19762         nds32_output_call, nds32_symbol_binds_local_p): New functions.
19763         * config/nds32/nds32-protos.h (nds32_output_call,
19764         nds32_output_return): Declare.
19765         * config/nds32/nds32.md: Refine all the call and return patterns.
19767 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19769         PR debug/85252
19770         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
19771         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
19773         PR rtl-optimization/84872
19774         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
19775         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
19776         EDGE_CROSSING edge.
19778 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
19780         * expr.c (copy_blkmode_to_reg): Revert 254862.
19781         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
19783 2018-04-06  Richard Biener  <rguenther@suse.de>
19785         PR middle-end/85244
19786         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
19787         after seeing a component reference with an adjacent field.  Treat
19788         refs to arrays at struct end of external decls similar to
19789         refs to unconstrained commons.
19791 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19793         PR sanitizer/85213
19794         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19795         look through SAVE_EXPRs with non-side-effects argument.  Adjust
19796         recursive calls.
19797         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19798         save_p here.
19800 2018-04-06  Richard Biener  <rguenther@suse.de>
19802         PR middle-end/85180
19803         * alias.c (find_base_term): New wrapper around find_base_term
19804         unwinding CSELIB_VAL_PTR changes.
19805         (find_base_term): Do not restore CSELIB_VAL_PTR during the
19806         recursion.
19808 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19810         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19811         instructions.
19812         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19813         constant definitions.
19814         ("nop"): lr 0,0 -> nopr r0
19815         ("nop_lr0", "nop_lr1"): New insn definitions.
19817 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19819         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19820         NDS32_V3PUSH_AVAILABLE_P macro.
19822 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
19823             Chung-Ju Wu  <jasonwucj@gmail.com>
19825         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19826         (nds32*-*-*): Add float and fpu_config into supported_defaults.
19827         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19828         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19829         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19830         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19831         * config/nds32/constraints.md: New constraints and checking for hard
19832         float configuration.
19833         * config/nds32/iterators.md: New mode iterator and attribute for hard
19834         float configuration.
19835         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19836         patterns.
19837         * config/nds32/nds32-fpu.md: New file.
19838         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19839         deal with hard float code generation.
19840         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19841         ARCH_V3S.
19842         (abi_type, float_reg_number): New enum type.
19843         * config/nds32/nds32-predicates.c: New predicates for hard float.
19844         * config/nds32/nds32-protos.h: Declare functions for hard float.
19845         * config/nds32/nds32.c: Implementation for hard float configuration.
19846         * config/nds32/nds32.h: Definitions for hard float configuration.
19847         * config/nds32/nds32.md: Include hard float machine description and
19848         modify patterns for hard float configuration.
19849         * config/nds32/nds32.opt: New options for hard float configuration.
19850         * config/nds32/predicates.md: New predicates for hard float
19851         configuration.
19853 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19855         * common/config/nds32/nds32-common.c
19856         (nds32_option_optimization_table): Enable -mreleax-hint by default.
19858 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
19860         PR middle-end/85195
19861         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19862         CONSTRUCTOR_ELT (ctor, ...)->value.
19864 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
19866         PR target/85193
19867         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19869 2018-04-05  Tom de Vries  <tom@codesourcery.com>
19871         PR target/85204
19872         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19873         cond jump.
19875 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19876             Kito Cheng  <kito.cheng@gmail.com>
19878         * config/nds32/constraints.md (U33): Fine-tune checking condition.
19879         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19880         * config/nds32/nds32.h (nds32_16bit_address_type): Add
19881         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19883 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19884             Kito Cheng  <kito.cheng@gmail.com>
19886         * config/nds32/constraints.md (Ufe): New memory constraint.
19887         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19888         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19889         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19890         operands.
19891         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19892         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19894 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19896         * config/nds32/nds32.md: Use optimize_size in the condition for
19897         alu-shift instructions.
19899 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19901         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19903 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19905         * config/nds32/nds32.md (negsi2): Refine pattern.
19907 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19908             Chung-Ju Wu  <jasonwucj@gmail.com>
19910         * config/nds32/iterators.md (shift_rotate): New code iterator.
19911         (shift): New code attribute.
19912         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19913         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19914         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19915         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19916         bit-wise operations.
19917         (andsi3, *andsi3): Ditto.
19918         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19919         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19920         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19921         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19922         nds32_ior_operand, nds32_xor_operand): New predicates.
19924 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19926         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19927         (addsi3, subsi3): ... this.
19929 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19931         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19933 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19935         * config/nds32/nds32.md: Adjust indention.
19937 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19939         * config/nds32/nds32.md (feature): New attribute.
19941 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19943         * config/nds32/nds32.md (subtype): New attribute.
19945 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19947         PR target/85203
19948         * config/arm/arm-builtins.c (arm_expand_builtin): Change
19949         expansion to perform a bitwise AND of the argument followed by a
19950         boolean negation of the result.
19952 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
19954         PR rtl-optimization/84878
19955         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19956         the basic block.  Assert the use reference is not artificial and that
19957         it has an associated insn.
19959 2018-04-04  Michael Matz  <matz@suse.de>
19961         * builtins.c (compute_objsize): Pass correct operand
19962         to array_at_struct_end_p.
19964 2018-04-04  Richard Biener  <rguenther@suse.de>
19966         PR lto/85176
19967         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19968         from contexts for DINFO_LEVEL_TERSE and below.
19970 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19972         * config/nds32/nds32-doubleword.md (move_<mode>): Require
19973         resiter_operand condition.
19974         * config/nds32/nds32.md (*move<mode>): Ditto.
19976 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19977             Monk Chiang  <sh.chiang04@gmail.com>
19979         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19981 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19983         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19985 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19986             Kito Cheng  <kito.cheng@gmail.com>
19988         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19989         nds32_cond_code_str, output_cond_branch,
19990         output_cond_branch_compare_zero, nds32_expand_cbranch,
19991         nds32_expand_cstore, nds32_expand_movcc,
19992         nds32_output_cbranchsi4_equality_zero,
19993         nds32_output_cbranchsi4_equality_reg,
19994         nds32_output_cbranchsi4_equality_reg_or_const_int,
19995         nds32_output_cbranchsi4_greater_less_zero: New functions.
19996         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19997         nds32_expand_cstore, nds32_expand_movcc,
19998         nds32_output_cbranchsi4_equality_zero,
19999         nds32_output_cbranchsi4_equality_reg,
20000         nds32_output_cbranchsi4_equality_reg_or_const_int,
20001         nds32_output_cbranchsi4_greater_less_zero): Declare.
20002         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20003         nds32_rimm11s_operand): New predicates.
20004         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20005         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20006         patterns.
20008 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20010         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20011         * config/nds32/nds32.md: Ditto.
20012         * config/nds32/pipelines.md: Ditto.
20014 2018-04-04  Richard Biener  <rguenther@suse.de>
20016         PR tree-optimization/85168
20017         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20018         propagating abnormals.
20020 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20022         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20024 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20025             Kito Cheng  <kito.cheng@gmail.com>
20027         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20028         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20029         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20030         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20031         * config/nds32/nds32.md (sibcall_internal): New.
20032         (sibcall_register): Remove.
20033         (sibcall_immediate): Remove.
20034         (sibcall_value_internal): New.
20035         (sibcall_value_register): Remove.
20036         (sibcall_value_immediate): Remove.
20037         * config/nds32/predicates.md (nds32_general_register_operand): New.
20038         (nds32_call_address_operand): New.
20040 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20042         PR rtl-optimization/85167
20043         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20044         bb_defs if *split_p, instead preinitialize it to NULL.
20046         PR tree-optimization/85156
20047         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20048         evaluating the argument multiple times.
20050 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20052         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20053         than vector.
20054         (_mm_cvtpd_ps): Likewise.
20055         (_mm_cvttpd_epi32): Likewise.
20056         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20057         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20058         vector, pixel, and bool following altivec.h include.
20060 2018-04-03  Martin Sebor  <msebor@redhat.com>
20062         * doc/extend.texi (Common Function Attributes): Clarify.
20063         (const attribute): Likewise.
20064         (pure attribute): Likewise.
20066 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20068         PR target/85169
20069         * config/i386/i386.c (ix86_expand_vector_set): Use
20070         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20072 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20074         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20075         instructions when changing rounding bits to preserve precision bits
20076         in the x87 control word.
20078 2018-04-03  Martin Liska  <mliska@suse.cz>
20080         PR tree-optimization/82491
20081         * rtl.h (strip_offset_and_add): Replace += suboffset with
20082         poly_uint64 () + suboffset.
20084 2018-03-29  Martin Liska  <mliska@suse.cz>
20085             Martin Jambor  <mjambor@suse.cz>
20087         PR ipa/84947
20088         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20089         param_type is not an integral or pointer type.
20091 2018-04-03  Richard Biener  <rguenther@suse.de>
20093         * sese.h (recompute_all_dominators): Remove.
20095 2018-04-02  Martin Sebor  <msebor@redhat.com>
20097         * doc/invoke.texi (-Wrestrict): Fix typos.
20099 2018-04-02  Jim Wilson  <jimw@sifive.com>
20101         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20102         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20103         (<optab>di3, <optab>si3_extend): Likewise.
20104         (<optab>si3_mask, <optab>si3_mask_1): New.
20105         (<optab>di3_mask, <optab>di3_mask_1): New.
20106         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20107         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20108         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20110 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20112         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20113         example.
20115 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20117         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20118         (nds32_canonicalize_comparison): New function.
20120 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20121             Kito Cheng  <kito.cheng@gmail.com>
20122             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20124         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20125         * config/nds32/constants.md (unspec_volatile_element): Add
20126         UNSPEC_VOLATILE_RELAX_GROUP.
20127         * config/nds32/nds32-relax-opt.c: New file.
20128         * config/nds32/nds32-predicates.c
20129         (nds32_symbol_load_store_p): New function.
20130         * config/nds32/nds32-protos.h
20131         (nds32_symbol_load_store_p): Declare function.
20132         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20133         * config/nds32/nds32.c
20134         (nds32_register_pass): New function to register pass.
20135         (nds32_register_passes): New function to register passes.
20136         * config/nds32/nds32.md (relax_group): New pattern.
20137         * config/nds32/nds32.opt (mrelax-hint): New option.
20138         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20140 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20142         * config/nds32/t-nds32: Modify files dependency.
20144 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20146         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20147         (PROFILE_HOOK): Define its implementation.
20149 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20151         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20152         type and 32-bit size.
20154 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20156         PR middle-end/85090
20157         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20158         (V_128_256): New mode iterator.
20159         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20160         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20161         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20162         of V.
20163         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20164         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20166 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20168         PR target/83315
20169         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20170         NaN inputs correctly.
20172 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20174         PR target/80546
20175         * config/rs6000/vsx.md (??r): New mode attribute.
20176         (*vsx_mov<mode>_64bit): Use it.
20177         (*vsx_mov<mode>_32bit): Likewise.
20179 2018-03-30  Martin Sebor  <msebor@redhat.com>
20181         PR tree-optimization/84818
20182         * builtins.c (check_access): Use warning_n.
20184 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20186         PR target/83822
20187         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20188         condition.
20189         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20190         condition.
20192 2018-03-30  Julia Koval  <julia.koval@intel.com>
20194         PR target/84413
20195         * x86-tune.def (movx, partial_reg_dependency): Enable for
20196         m_SKYLAKE_AVX512.
20198 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20200         PR inline-asm/84985
20201         * lra-constraints.c (process_alt_operands): Move setting
20202         this_alternative_matches below.
20204 2018-03-29  Martin Liska  <mliska@suse.cz>
20206         PR lto/84995.
20207         * doc/invoke.texi: Document how LTO works with debug info.
20208         Describe auto-load support of binutils.  Mention 'x86-64'
20209         as valid option value of -march option.
20211 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20213         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20215         PR c/85094
20216         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20217         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20218         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20219         checking.
20221 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20223         PR target/84912
20224         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20225         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20226         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20227         for RS6000_BTM_POWERPC64.
20228         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20229         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20230         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20231         definition.
20232         (DIVDE): Use it.
20233         (DIVDEU): Likewise.
20235 2018-03-28  Carl Love  <cel@us.ibm.com>
20237         Revert
20238         2017-09-27  Carl Love  <cel@us.ibm.com>
20240         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20241         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20242         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20243         fctiw instruction.
20245 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20247         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20248         instead of __vector bool.
20249         (_mm_max_pu8): Likewise.
20250         (_mm_min_pi16): Likewise.
20252 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20254         PR target/84912
20255         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20256         (DIVWEUO): Likewise.
20257         (DIVDEO): Likewise.
20258         (DIVDEUO): Likewise.
20259         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20260         DIVWEUO and DIVDEUO.
20261         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20262         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20263         (div_extend): Likewise.
20264         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20265         builtin function.
20266         (__builtin_divweuo): Likewise.
20267         (__builtin_divdeo): Likewise.
20268         (__builtin_divdeuo): Likewise.
20270 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20272         PR target/85095
20273         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20274         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20276         PR tree-optimization/82004
20277         * gimple-match-head.c (optimize_pow_to_exp): New function.
20278         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20279         Don't fold to exp if optimize_pow_to_exp is false.
20281 2018-03-28  Martin Liska  <mliska@suse.cz>
20283         PR other/84819
20284         * calls.c (initialize_argument_information): Fix trailing space.
20285         * common.opt: Fix typo and provide better explanation for
20286         -fsanitize-coverage option.
20287         * config/i386/i386.opt: Fix typo.
20289 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20290             Martin Liska  <mliska@suse.cz>
20292         PR sanitizer/85081
20293         * gimplify.c (asan_poison_variable): Don't do the check for
20294         gimplify_omp_ctxp here.
20295         (gimplify_decl_expr): Do it here.
20296         (gimplify_target_expr): Likewise.
20298 2018-03-28  Martin Liska  <mliska@suse.cz>
20300         PR target/84988
20301         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20302         chkp_type_bounds_count if MPX is not enabled.
20304 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20306         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20308 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20310         PR target/84914
20311         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20312         function to create the function decl for complex long double
20313         multiply and divide for -mabi=ieeelongdouble.
20314         (init_float128_ieee): Call it.
20316 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20318         PR target/85044
20319         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20320         -fcf-protection=branch -mibt.
20321         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20323 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20325         PR target/81863
20326         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20327         arm_word_relocations.
20329 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20331         PR target/85056
20332         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20333         extern array declarations.
20335 2018-03-27  Richard Biener  <rguenther@suse.de>
20337         PR middle-end/84067
20338         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20339         explicit single_use checks.
20341 2018-03-27  Richard Biener  <rguenther@suse.de>
20343         PR tree-optimization/85082
20344         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20345         Valueize the VUSE.
20347 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20349         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20350         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20351         Turn on fasynchronous-unwind-tables and funwind-tables.
20353 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20355         PR target/85073
20356         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20357         (*bmi_blsr_<mode>_ccz): Ditto.
20359 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20361         PR tree-optimization/85063
20362         * omp-general.c (offloading_function_p): New function.  Factor out
20363         of ...
20364         * omp-offload.c (pass_omp_target_link::gate): ... here.
20365         * omp-general.h (offloading_function_p): Declare.
20366         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20367         with attribute omp declare target for offloading functions.
20369 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20371         PR tree-optimization/84005
20372         * tree-data-ref.h (get_base_for_alignment): Declare.
20373         * tree-data-ref.c (get_base_for_alignment_1): New function.
20374         (get_base_for_alignment): Likewise.
20375         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20376         get_base_for_alignment to find a suitable base object, instead
20377         of always using drb->base_address.
20379 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20381         PR inline-asm/85022
20382         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20383         known size by default.
20385 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
20387         PR inline-asm/85030
20388         * lra-constraints.c (process_alt_operands): Don't match BLKmode
20389         and non BLKmode operands.
20391 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20393         PR target/85026
20394         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20395         Clean up attributes.
20397 2018-03-23  Richard Biener  <rguenther@suse.de>
20399         PR debug/85020
20400         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20401         we are going to emit early debug for LTO.
20403 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20405         PR inline-asm/85034
20406         * function.c (match_asm_constraints_1): Don't optimize if input
20407         doesn't satisfy general_operand predicate for output's mode.
20409         PR inline-asm/85022
20410         * alias.c (write_dependence_p): Don't require for x_canonicalized
20411         non-VOIDmode if x has VOIDmode.
20413         PR sanitizer/85029
20414         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20415         just don't try to optimize it rather than assert it never happens.
20417 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20419         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20420         macro expansions for definition of ST_INTERNAL_<mode> and
20421         LD_INTERNAL_<mode> builtins.
20422         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20423         Remove prototype.
20424         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20425         function.
20426         (altivec_expand_st_builtin): Likewise.
20427         (altivec_expand_builtin): Remove calls to deleted functions.
20428         (rs6000_address_for_altivec): Delete this function.
20429         * config/rs6000/vector.md: Remove expands for
20430         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20432 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
20434         PR target/84826
20435         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20436         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20437         re-computing once computed.
20438         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20439         (arm_init_machine_status): Initialize
20440         machine->static_chain_stack_bytes.
20442 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20444         PR target/84760
20445         * doc/extend.texi: Add four new prototypes for vec_ld.
20446         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20447         definitions for more logical presentation.
20448         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20449         entries for V1TI variants of __builtin_altivec_ld builtin.
20450         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20451         handling of V1TI variant of LVX icode pattern.
20452         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20453         (rs6000_gimple_fold_builtin): Likewise.
20454         (altivec_init_builtins): Add code to define
20455         __builtin_altivec_lvx_v1ti function.
20457 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20459         PR inline-asm/84941
20460         * function.c (match_asm_constraints_1): Don't do the optimization
20461         if input isn't a REG, SUBREG, MEM or constant.
20463 2018-03-22  Tom de Vries  <tom@codesourcery.com>
20465         PR tree-optimization/84956
20466         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20467         bb_has_abnormal_pred.
20469 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20471         PR sanitizer/85018
20472         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20473         DECL_INITIAL (decl) to decl at the end.
20474         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20475         adjust the comment.
20477 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
20479         * doc/extend.texi (__builtin_tgmath): Document when complex
20480         integer types are treated as _Complex _Float64.
20482 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20484         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20486 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20488         PR tree-optimization/84960
20489         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20490         if it is ENTRY block, move them into single succ of ENTRY in that case.
20492 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
20494         PR tree-optimization/84811
20495         * poly-int.h (poly_span_traits): Remove the T3 parameter and
20496         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20497         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20498         (known_subrange_p): Update accordingly.  Cast each value involved
20499         in the size comparison, rather than casting the result of the
20500         subtraction.
20502 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20504         PR tree-optimization/84982
20505         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20506         by flipping the least significant bit rather than all bits from
20507         bitpos to bitpos + bitsize - 1.
20509 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20511         * doc/extend.texi (Deprecated Features): Remove mention of
20512         long-deleted deprecations.
20514 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20516         PR jit/84288
20517         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
20518         * configure: Regenerate.
20520 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20522         PR tree-optimization/83126
20523         * tree-parloops.c (num_phis): New function.
20524         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
20526 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20528         * doc/extend.texi (Deprecated Features): Update deprecated flags,
20529         mention anon-struct/union members and trailing attributes.
20531 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
20533         PR tree-optimization/84969
20534         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
20535         builtin memset partitions if they set different rhs values.
20537 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20539         PR rtl-optimization/84989
20540         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
20541         VEC_DUPLICATE with scalar result mode.
20543 2018-03-21  Martin Liska  <mliska@suse.cz>
20545         PR ipa/84963
20546         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
20547         not intended return statement.
20549 2018-03-21  Martin Liska  <mliska@suse.cz>
20551         PR target/84988
20552         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
20553         (chkp_find_bound_slots_1): Limit number of iterations.
20555 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
20557         PR target/84838
20558         * Minor grammar fixes for x86 options.
20560 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20562         PR debug/84875
20563         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
20564         holding REG_CFA_RESTORE notes, instead turn them into a USE.
20566 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
20568         PR target/83789
20569         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
20570         (altivec_lvx_<mode>_1op): Likewise.
20571         (altivec_stvx_<mode>_2op): Likewise.
20572         (altivec_stvx_<mode>_1op): Likewise.
20573         (altivec_lvx_<VM2:mode>): New define_expand.
20574         (altivec_stvx_<VM2:mode>): Likewise.
20575         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
20576         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20577         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
20578         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20579         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
20580         (rs6000_gen_lvx): Likewise.
20581         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
20582         (altivec_expand_stv_builtin): Likewise.
20583         (altivec_expand_builtin): Likewise.
20584         * config/rs6000/vector.md: Likewise.
20586 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20588         PR target/82518
20589         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
20590         BYTES_BIG_ENDIAN.
20592 2018-03-20  Richard Biener  <rguenther@suse.de>
20594         PR target/84986
20595         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
20596         sign-conversions as zero, fall back to standard scalar_stmt
20597         cost for the rest.
20599 2018-03-20  Martin Liska  <mliska@suse.cz>
20601         PR ipa/84825
20602         * predict.c (rebuild_frequencies): Handle case when we have
20603         PROFILE_ABSENT, but flag_guess_branch_prob is false.
20605 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20607         PR target/84990
20608         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
20609         flag_section_anchors.
20610         * varasm.c (use_blocks_for_decl_p): Remove hack for
20611         dw2_force_const_mem.
20613         PR target/84845
20614         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
20615         to ...
20616         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
20617         be created, use lowpart_subreg of operands[0] rather than operands[0]
20618         itself.
20619         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
20620         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
20621         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
20622         and n constraint instead of aarch64_shift_imm_di and Usd.
20623         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
20624         (*aarch64_<optab>_reg_minus<mode>3): ... this.
20626 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
20628         PR target/82989
20629         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
20630         to favor GPR over NEON registers.
20631         (<shift>di3_neon): Likewise.
20633 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20635         PR target/84952
20636         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
20637         (nvptx_process_pars): Emit bar.sync asap and alap.
20639 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20641         PR target/84954
20642         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
20643         seen_label if seen_label is already set.
20645 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20647         PR target/84945
20648         * config/i386/i386.c (fold_builtin_cpu): For features above 31
20649         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
20650         Use 1U instead of 1.  Formatting fixes.
20652         PR c/84953
20653         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
20654         instead of TREE_TYPE (s1) for the return value.
20656 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
20658         PR tree-optimization/84946
20659         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
20660         bitsize + bitsize in poly_uint64 rather than poly_int64.
20662         PR sanitizer/78651
20663         * dwarf2asm.c: Include fold-const.c.
20664         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
20665         of decl rather than decl itself.
20667         PR rtl-optimization/84643
20668         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
20670 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
20672         PR sanitizer/78651
20673         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
20674         calling assemble_variable.
20676 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
20678         PR target/81647
20679         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
20680         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
20682 2018-03-19  Jim Wilson  <jimw@sifive.com>
20684         PR bootstrap/84856
20685         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
20686         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
20687         (riscv_first_stack_step): Likewise.
20688         (riscv_option_override): Use STACK_BOUNDARY instead of
20689         MIN_STACK_BOUNDARY.
20690         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
20691         MIN_STACK_BOUNDARY.
20692         (BIGGEST_ALIGNMENT): Set to 128.
20693         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
20694         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
20695         STACK_BOUNDARY.
20697 2018-03-19  Richard Biener  <rguenther@suse.de>
20699         PR tree-optimization/84933
20700         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
20701         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
20703 2018-03-19  Richard Biener  <rguenther@suse.de>
20705         PR tree-optimization/84859
20706         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
20707         (cond_if_else_store_replacement): Perform sinking operation on
20708         single-store BBs regardless of MAX_STORES_TO_SINK setting.
20709         Generalize what a BB with a single eligible store is.
20711 2018-03-19  Richard Biener  <rguenther@suse.de>
20713         PR tree-optimization/84929
20714         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
20715         chrec_is_positive against non-chrec arg.
20717 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
20719         PR target/84711
20720         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
20722 2018-03-18  Martin Liska  <mliska@suse.cz>
20724         PR rtl-optimization/84635
20725         * regrename.c (build_def_use): Use matches_mode only when
20726         matches >= 0.
20728 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
20730         PR tree-optimization/84913
20731         * tree-vect-loop.c (vectorizable_reduction): Don't try to
20732         vectorize chains of COND_EXPRs.
20734 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20736         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
20738 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20740         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
20742 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20744         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
20746 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20747             Kito Cheng  <kito.cheng@gmail.com>
20749         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
20750         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
20751         (nds32_adjust_reg_alloc_order): New function.
20752         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
20754 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
20756         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
20757         nds32_print_operand, nds32_print_operand_address): Use
20758         HOST_WIDE_INT_PRINT_DEC instead.
20760 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20762         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
20764 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
20766         PR target/84902
20767         * config/i386/i386.c (initial_ix86_tune_features,
20768         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
20769         unsigned long long.
20770         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
20771         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
20772         rather than 1u << ix86_tune.  Formatting fix.
20773         (ix86_option_override_internal): Change ix86_arch_mask from
20774         unsigned int to unsigned HOST_WIDE_INT, initialize to
20775         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
20776         (ix86_function_specific_restore): Likewise.
20778 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20780         PR target/84899
20781         * postreload.c (reload_combine_recognize_pattern): Perform
20782         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
20783         truncate_int_for_mode the result for the destination's mode.
20785         PR c/84909
20786         * hsa-gen.c (mem_type_for_type): Fix comment typo.
20787         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20788         Likewise.
20789         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20790         Likewise.
20792 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
20794         PR target/84876
20795         * lra-assigns.c (lra_split_hard_reg_for): Don't use
20796         regno_allocno_class_array and sorted_pseudos.
20797         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20798         insns where regno is used.
20800 2018-03-16  Martin Liska  <mliska@suse.cz>
20802         PR ipa/84833
20803         * multiple_target.c (create_dispatcher_calls): Redirect
20804         reference in the symbol table.
20806 2018-03-16  Martin Liska  <mliska@suse.cz>
20808         PR ipa/84722
20809         * multiple_target.c (create_dispatcher_calls): Redirect also
20810         an alias.
20812 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20814         PR c++/79937
20815         PR c++/82410
20816         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20817         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20818         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20820 2018-03-16  Julia Koval  <julia.koval@intel.com>
20822         * doc/invoke.texi (Skylake Server): Add CLWB.
20823         Cannonlake): Remove CLWB.
20825 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20827         PR tree-optimization/84841
20828         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20829         1 << 3.
20830         (FLOAT_ONE_CONST_TYPE): Define.
20831         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20832         (sort_by_operand_rank): Put entries with higher constant_type last
20833         rather than first to match comments.
20835 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
20837         * config/nios2/nios2.md (movsi_internal): Fix thinko in
20838         split predicate.
20840 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20842         PR c++/79085
20843         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20844         check and use address of target always.
20846 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
20848         PR target/84574
20849         * config/i386/i386.c (indirect_thunk_needed): Update comments.
20850         (indirect_thunk_bnd_needed): Likewise.
20851         (indirect_thunks_used): Likewise.
20852         (indirect_thunks_bnd_used): Likewise.
20853         (indirect_return_needed): New.
20854         (indirect_return_bnd_needed): Likewise.
20855         (output_indirect_thunk_function): Add a bool argument for
20856         function return.
20857         (output_indirect_thunk_function): Don't generate alias for
20858         function return thunk.
20859         (ix86_code_end): Call output_indirect_thunk_function to generate
20860         function return thunks.
20861         (ix86_output_function_return): Set indirect_return_bnd_needed
20862         and indirect_return_needed instead of indirect_thunk_bnd_needed
20863         and indirect_thunk_needed.
20865 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
20867         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20868         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20869         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20871 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
20872             Paul Hua <paul.hua.gm@gmail.com>
20874         PR c/84852
20875         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20877 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
20879         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20880         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20881         resp. SFmode cases.
20883 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
20885         PR target/84711
20886         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20887         instead of GET_MODE_SIZE when comparing Units.
20889 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20891         PR target/68256
20892         * varasm.c (hash_section): Return an unchangeble hash value
20893         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20894         Return !aarch64_can_use_per_function_literal_pools_p ().
20896 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20898         PR target/84860
20899         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20900         rather than address of cmode as last argument to prepare_cmp_insn.
20902 2018-03-15  Julia Koval  <julia.koval@intel.com>
20904         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20905         F_AVX512VNNI, F_AVX512BITALG): New.
20907 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20909         PR target/83451
20910         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20911         insn for floating-point loads and stores.
20913 2018-03-14  Carl Love  <cel@us.ibm.com>
20915         * config/rs6000/rs6000-c.c: Add macro definitions for
20916         ALTIVEC_BUILTIN_VEC_PERMXOR.
20917         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20918         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20919         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20920         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20921         UNSPEC_VPERMXOR.
20922         * config/doc/extend.texi: Add prototypes for vec_permxor.
20924 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20926         PR c/84852
20927         * diagnostic-show-locus.c (class layout_point): Convert m_line
20928         from int to linenum_type.
20929         (line_span::comparator): Use linenum "compare" function when
20930         comparing line numbers.
20931         (test_line_span): New function.
20932         (layout_range::contains_point): Convert param "row" from int to
20933         linenum_type.
20934         (layout_range::intersects_line_p): Likewise.
20935         (layout::will_show_line_p): Likewise.
20936         (layout::print_source_line): Likewise.
20937         (layout::should_print_annotation_line_p): Likewise.
20938         (layout::print_annotation_line): Likewise.
20939         (layout::print_leading_fixits): Likewise.
20940         (layout::annotation_line_showed_range_p): Likewise.
20941         (struct line_corrections): Likewise for field m_row.
20942         (line_corrections::line_corrections): Likewise for param "row".
20943         (layout::print_trailing_fixits): Likewise.
20944         (layout::get_state_at_point): Likewise.
20945         (layout::get_x_bound_for_row): Likewise.
20946         (layout::print_line): Likewise.
20947         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20948         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20949         * input.c (selftest::test_linenum_comparisons): New function.
20950         (selftest::input_c_tests): Call it.
20951         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20952         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20953         * selftest.h (ASSERT_GT): New macro.
20954         (ASSERT_GT_AT): New macro.
20955         (ASSERT_LT): New macro.
20956         (ASSERT_LT_AT): New macro.
20958 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
20960         PR rtl-optimization/84780
20961         * combine.c (distribute_links): Don't make a link based on pc_rtx.
20963 2018-03-14  Martin Liska  <mliska@suse.cz>
20965         * tree.c (record_node_allocation_statistics): Use
20966         get_stats_node_kind.
20967         (get_stats_node_kind): New function extracted from
20968         record_node_allocation_statistics.
20969         (free_node): Use get_stats_node_kind.
20971 2018-03-14  Richard Biener  <rguenther@suse.de>
20973         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20974         that the value-set of ANTIC_IN doesn't grow.
20976         Revert
20977         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20978         member.
20979         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20980         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20982 2018-03-14  Julia Koval  <julia.koval@intel.com>
20984         * config.gcc (icelake-client, icelake-server): New.
20985         (icelake): Remove.
20986         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20987         (initial_ix86_arch_features): Ditto.
20988         (PTA_SKYLAKE): Add SGX.
20989         (PTA_ICELAKE): Remove.
20990         (PTA_ICELAKE_CLIENT): New.
20991         (PTA_ICELAKE_SERVER): New.
20992         (ix86_option_override_internal): Split up icelake on icelake client and
20993         icelake server.
20994         (get_builtin_code_for_version): Ditto.
20995         (fold_builtin_cpu): Ditto.
20996         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20997         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20998         * config/i386/i386.h (processor_type): Ditto.
20999         * doc/invoke.texi: Ditto.
21001 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21003         PR sanitizer/83392
21004         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21005         INTEGER_CST offset, add it together with bitpos / 8 and
21006         sign extend based on POINTER_SIZE.
21008         PR target/84844
21009         Revert
21010         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21012         PR target/78090
21013         * config/i386/constraints.md (Yc): New register constraint.
21014         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21015         Use Yc constraint for alternative 2 of operand 0.  Remove
21016         preferred_for_speed attribute.
21018 2018-03-14  Richard Biener  <rguenther@suse.de>
21020         PR tree-optimization/84830
21021         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21022         with the old one to avoid oscillations.
21024 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21026         PR target/83712
21027         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21028         pseudos.
21029         (assign_by_spills): Return a flag of reload assignment failure.
21030         Do not process the reload assignment failures.  Do not spill other
21031         reload pseudos if they has the same reg class.  Update n if
21032         necessary.
21033         (lra_assign): Add a return arg.  Set up from the result of
21034         assign_by_spills call.
21035         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21036         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21037         usage_insns if it is not NULL.
21038         (spill_hard_reg_in_range): New function.
21039         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21040         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21041         function prototypes.
21042         (lra_assign): Change prototype.
21043         * lra.c (lra): Add code to deal with fails by splitting hard reg
21044         live ranges.
21046 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21048         * config/riscv/riscv.opt (mrelax): New option.
21049         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21050         "norelax" when riscv_mrelax is disabled.
21051         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21053 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21055         PR target/84743
21056         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21057         reassociation for int modes.
21059 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21061         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21062         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21063         for big-endian.
21064         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21065         * config/aarch64/aarch64-sve.md
21066         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21067         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21068         (*extend<mode><Vwide>2): Rename to...
21069         (aarch64_sve_extend<mode><Vwide>2): ...this.
21070         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21071         renaming the old pattern to...
21072         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21073         unsigned packs.
21074         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21075         define_expand, renaming the old pattern to...
21076         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21077         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21078         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21079         account when deciding which SVE instruction the optab should use.
21080         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21082 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21084         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21085         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21086         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21087         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21088         (tlsdesc_small_<mode>): Turn a define_expand and use
21089         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21090         (tlsdesc_small_advsimd_<mode>): ...this.
21091         (tlsdesc_small_sve_<mode>): New pattern.
21093 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21095         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21096         (UNSPEC_UMUL_HIGHPART): New constants.
21097         (MUL_HIGHPART): New int iteraor.
21098         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21099         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21100         define_expand.
21101         (*<su>mul<mode>3_highpart): New define_insn.
21103 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21105         PR lto/84805
21106         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21107         incomplete types.
21109 2018-03-13  Martin Liska  <mliska@suse.cz>
21111         PR ipa/84658.
21112         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21113         vector.
21114         (sem_item_optimizer::~sem_item_optimizer): Release it.
21115         (sem_item_optimizer::merge_classes): Register variable aliases.
21116         (sem_item_optimizer::fixup_pt_set): New function.
21117         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21118         * ipa-icf.h: Declare new variables and functions.
21120 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21122         PR middle-end/84834
21123         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21124         integer_pow2p@2 and test integer_pow2p in condition.
21125         (A < 0 ? C : 0): Similarly for @1.
21127         PR middle-end/84831
21128         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21129         characters starting at p contain '\0' character, don't look beyond
21130         that.
21132         PR target/84827
21133         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21134         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21136         PR target/84828
21137         * reg-stack.c (change_stack): Change update_end var from int to
21138         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21139         also call set_block_for_insn on the newly added insns and rescan.
21141         PR target/84786
21142         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21143         on the last operand.
21145         PR c++/84704
21146         * tree.c (stabilize_reference_1): Return save_expr (e) for
21147         STATEMENT_LIST even if it doesn't have side-effects.
21149 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21151         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21153 2018-03-12  Renlin Li  <renlin.li@arm.com>
21155         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21156         aarch64_output_scalar_simd_mov_immediate.
21158 2018-03-12  Martin Sebor  <msebor@redhat.com>
21160         PR tree-optimization/83456
21161         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21162         for perfectly overlapping calls to memcpy.
21163         (gimple_fold_builtin_memory_chk): Same.
21164         (gimple_fold_builtin_strcpy): Handle no-warning.
21165         (gimple_fold_builtin_stxcpy_chk): Same.
21166         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21168 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21170         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21171         parameter.  Use it for SFmode.
21172         (rs6000_function_arg_advance_1): Adjust.
21173         (rs6000_function_arg): Adjust.
21174         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21176 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21178         PR rtl-optimization/84169
21179         PR rtl-optimization/84780
21180         * combine.c (can_combine_p): Check for a 2-insn combination whether
21181         the destination register is used between the two insns, too.
21183 2018-03-12  Richard Biener  <rguenther@suse.de>
21185         PR tree-optimization/84803
21186         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21187         for refs DR analysis didn't process.
21189 2018-03-12  Richard Biener  <rguenther@suse.de>
21191         PR tree-optimization/84777
21192         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21193         force-vectorize loops ignore whether we are optimizing for size.
21195 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21197         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21198         (TARGET_MD_ASM_ADJUST): Define.
21200 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21201             Kito Cheng  <kito.cheng@gmail.com>
21202             Chung-Ju Wu  <jasonwucj@gmail.com>
21204         * config/nds32/nds32.c (nds32_compute_stack_frame,
21205         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21206         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21207         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21208         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21209         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21210         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21211         * config/nds32/nds32.md (prologue, epilogue): Use macro
21212         NDS32_V3PUSH_AVAILABLE_P to do checking.
21214 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21216         PR debug/58150
21217         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21218         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21219         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21220         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21221         addition of most attributes on !orig_type_die or the attribute not
21222         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21224 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21225             Chung-Ju Wu  <jasonwucj@gmail.com>
21227         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21228         __NDS32_VH__ macro.
21229         * config/nds32/nds32.opt (mvh): New option.
21231 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21232             Chung-Ju Wu  <jasonwucj@gmail.com>
21234         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21235         function.
21236         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21237         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21238         definition.
21240 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21241             Chung-Ju Wu  <jasonwucj@gmail.com>
21243         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21244         function.
21245         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21246         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21248 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21249             Kito Cheng  <kito.cheng@gmail.com>
21250             Chung-Ju Wu  <jasonwucj@gmail.com>
21252         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21253         UNSPEC_FFMISM and UNSPEC_FLMISM.
21254         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21255         for ffb, ffmism and flmism.
21256         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21257         (unspec_ffmism): Ditto.
21258         (unspec_flmism): Ditto.
21259         (nds32_expand_builtin_impl): Check if string extension is available.
21260         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21261         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21263 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21265         Reverting patch:
21266         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21268         PR target/83712
21269         * lra-assigns.c (assign_by_spills): Return a flag of reload
21270         assignment failure.  Do not process the reload assignment
21271         failures.  Do not spill other reload pseudos if they has the same
21272         reg class.
21273         (lra_assign): Add a return arg.  Set up from the result of
21274         assign_by_spills call.
21275         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21276         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21277         usage_insns if it is not NULL.
21278         (spill_hard_reg_in_range): New function.
21279         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21280         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21281         function prototypes.
21282         (lra_assign): Change prototype.
21283         * lra.c (lra): Add code to deal with fails by splitting hard reg
21284         live ranges.
21286 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21288         PR target/84807
21289         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21291 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21293         PR debug/84620
21294         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21295         (dw_val_node): Add val_symbolic_view.
21296         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21297         (symview_upper_bound): New.
21298         (new_line_info_table): Initialize symviews_since_reset.
21299         (dwarf2out_source_line): Count symviews_since_reset and set
21300         symview_upper_bound.
21301         (dw_val_equal_p): Handle symview.
21302         (add_AT_symview): New.
21303         (print_dw_val): Handle symview.
21304         (attr_checksum, attr_checksum_ordered): Likewise.
21305         (same_dw_val_p, size_of_die): Likewise.
21306         (value_format, output_die): Likewise.
21307         (add_high_low_attributes): Use add_AT_symview for entry_view.
21308         (dwarf2out_finish): Reset symview_upper_bound, clear
21309         zero_view_p.
21311 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21313         PR target/83969
21314         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21315         Add strict argument and use it.
21316         (rs6000_split_multireg_move): Update for new strict argument.
21317         (mem_operand_gpr): Disallow all non-offsettable addresses.
21318         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21320 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21322         PR target/84772
21323         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21324         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21325         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21327         PR c++/84767
21328         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21329         decl, use remap_type if we want to use the type.
21331 2018-03-09  Martin Sebor  <msebor@redhat.com>
21333         PR tree-optimization/84526
21334         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21335         Remove dead code.
21336         (builtin_access::generic_overlap): Be prepared to handle non-array
21337         base objects.
21339 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21341         PR rtl-optimization/84682
21342         * lra-constraints.c (process_address_1): Check is_address flag
21343         for address constraints.
21344         (process_alt_operands): Likewise.
21345         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21346         preprocess_constraints.
21347         * recog.h (preprocess_constraints): Add oploc parameter.
21348         Adjust callers.
21349         * recog.c (preprocess_constraints): Test address_operand for
21350         CT_ADDRESS constraints.
21352 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21354         PR target/83712
21355         * lra-assigns.c (assign_by_spills): Return a flag of reload
21356         assignment failure.  Do not process the reload assignment
21357         failures.  Do not spill other reload pseudos if they has the same
21358         reg class.
21359         (lra_assign): Add a return arg.  Set up from the result of
21360         assign_by_spills call.
21361         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21362         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21363         usage_insns if it is not NULL.
21364         (spill_hard_reg_in_range): New function.
21365         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21366         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21367         function prototypes.
21368         (lra_assign): Change prototype.
21369         * lra.c (lra): Add code to deal with fails by splitting hard reg
21370         live ranges.
21372 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21374         PR target/83193
21375         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21376         Accept complain bool parameter.  Only emit errors if it is true.
21377         (arm_parse_cpu_option_name): Likewise.
21378         (arm_target_thumb_only): Adjust callers of the above.
21379         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21380         prototype to take a default true bool parameter.
21381         (arm_parse_arch_option_name): Likewise.
21383 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
21384             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
21386         PR jit/64089
21387         PR jit/84288
21388         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21389         * configure: Regenerate.
21390         * configure.ac ("linker --version-script option"): New.
21391         ("linker soname option"): New.
21393 2018-03-09  Richard Biener  <rguenther@suse.de>
21395         PR tree-optimization/84775
21396         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21397         immediate uses of predicate stmts and mark them modified.
21399         Revert
21400         PR tree-optimization/84178
21401         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21402         to caller.
21403         (version_loop_for_if_conversion): Delay update_ssa call.
21404         (tree_if_conversion): Delay update_ssa until after predicate
21405         insertion.
21407 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
21409         PR target/84763
21410         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21411         when the function accesses prior frames.
21413 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21415         PR debug/84456
21416         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21417         gen_llsym, otherwise call maybe_gen_llsym.
21419         PR inline-asm/84742
21420         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21421         has ',' character inside of it.
21423 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21425         PR target/84748
21426         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21427         as clobbering CC_REGNUM.
21429 2018-03-08  Richard Biener  <rguenther@suse.de>
21431         PR middle-end/84552
21432         * tree-scalar-evolution.c: Include tree-into-ssa.h.
21433         (follow_copies_to_constant): Do not follow SSA names registered
21434         for update.
21436 2018-03-08  Richard Biener  <rguenther@suse.de>
21438         PR tree-optimization/84178
21439         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21440         to caller.
21441         (version_loop_for_if_conversion): Delay update_ssa call.
21442         (tree_if_conversion): Delay update_ssa until after predicate
21443         insertion.
21445 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
21447         PR tree-optimization/84178
21448         * tree-if-conv.c (release_bb_predicate): Remove the
21449         the assertion that the stmts have NULL use_ops.
21450         Discard the statements, asserting that they haven't
21451         yet been added to a BB.
21453 2018-03-08  Richard Biener  <rguenther@suse.de>
21455         PR tree-optimization/84746
21456         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21457         (phi_translate): Pass in destination ANTIC_OUT set.
21458         (phi_translate_1): Likewise.  For a simplified result lookup
21459         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21460         (phi_translate_set): Adjust.
21461         (do_pre_regular_insertion): Likewise.
21462         (do_pre_partial_partial_insertion): Likewise.
21464 2018-03-08  Martin Liska  <mliska@suse.cz>
21466         PR gcov-profile/84735
21467         * doc/gcov.texi: Document usage of profile files.
21468         * gcov-io.h: Document changes in the format.
21470 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
21472         PR debug/84404
21473         PR debug/84408
21474         * dwarf2out.c (struct dw_line_info_table): Update comments for
21475         view == -1.
21476         (FORCE_RESET_NEXT_VIEW): New.
21477         (FORCE_RESETTING_VIEW_P): New.
21478         (RESETTING_VIEW_P): Check for -1 too.
21479         (ZERO_VIEW_P): Likewise.
21480         (new_line_info_table): Force-reset next view.
21481         (dwarf2out_begin_function): Likewise.
21482         (dwarf2out_source_line): Simplify zero_view_p initialization.
21483         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21484         view directly.  Omit view when omitting .loc at line 0.
21486 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21488         PR tree-optimization/84740
21489         * tree-switch-conversion.c (process_switch): Call build_constructors
21490         only if info.phi_count is non-zero.
21492         PR tree-optimization/84739
21493         * tree-tailcall.c (find_tail_calls): Check call arguments against
21494         DECL_ARGUMENTS (current_function_decl) rather than
21495         DECL_ARGUMENTS (func) when checking for tail recursion.
21497 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21499         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21500         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21501         Volker Reichelt's entry and add entries for people that perform
21502         GCC fuzzy testing and report numerous bugs.
21504 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
21506         PR target/82411
21507         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
21508         readonly data in sdata, if that is disabled.
21509         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
21510         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
21511         -mreadonly-in-sdata option.
21513 2018-03-07  Martin Sebor  <msebor@redhat.com>
21515         PR tree-optimization/84468
21516         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
21517         basic block when looking for nul assignment.
21519 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
21521         PR target/84277
21522         * except.h (output_function_exception_table): Adjust prototype.
21523         * except.c (output_function_exception_table): Remove FNNAME parameter
21524         and add SECTION parameter.  Ouput one part of the table at a time.
21525         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
21526         the first part of the exception table and emit unwind directives.
21527         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
21528         (i386_pe_seh_cold_init): Likewise.
21529         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
21530         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
21531         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
21532         (ix86_output_call_insn): Emit a nop in one more case for SEH.
21533         * config/i386/winnt.c: Include except.h.
21534         (struct seh_frame_state): Add reg_offset, after_prologue and
21535         in_cold_section fields.
21536         (i386_pe_seh_end_prologue): Set seh->after_prologue.
21537         (i386_pe_seh_cold_init): New function.
21538         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
21539         to seh->in_cold_section.
21540         (seh_emit_push): Record the offset of the push.
21541         (seh_emit_save): Record the offet of the save.
21542         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
21543         Test seh->after_prologue to disregard the epilogue.
21544         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
21545         (i386_pe_end_cold_function): New function.
21547 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21549         PR fortran/84565
21550         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
21551         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
21553         PR c++/84704
21554         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
21555         on tmp_var.
21556         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
21557         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
21559         PR middle-end/84723
21560         * multiple_target.c: Include tree-inline.h and intl.h.
21561         (expand_target_clones): Diagnose and fail if node->definition and
21562         !tree_versionable_function_p (node->decl).
21564 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
21566         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
21567         sprint_ul.
21568         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
21569         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
21570         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
21572 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21574         PR target/84710
21575         * combine.c (try_combine): Use reg_or_subregno instead of handling
21576         just paradoxical SUBREGs and REGs.
21578 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
21580         * config/arc/arc.c (arc_finalize_pic): Remove function.
21581         (arc_must_save_register): We use single base PIC register, remove
21582         checks to save/restore the PIC register.
21583         (arc_expand_prologue): Likewise.
21584         * config/arc/arc-protos.h (arc_set_default_type_attributes):
21585         Remove.
21586         (arc_verify_short): Likewise.
21587         (arc_attr_type): Likewise.
21588         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
21589         (walk_stores): Likewise.
21590         (arc_address_cost): Make it static.
21591         (arc_verify_short): Likewise.
21592         (branch_dest): Likewise.
21593         (arc_attr_type): Likewise.
21594         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
21595         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
21596         (arc_final_prescan_insn): Remove inserting the nops due to
21597         hardware hazards.  It is done in reorg step.
21598         (insn_length_variant_t): Remove.
21599         (insn_length_parameters_t): Likewise.
21600         (arc_insn_length_parameters): Likewise.
21601         (arc_get_insn_variants): Likewise.
21602         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
21604 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21606         PR inline-asm/84683
21607         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
21608         assertion failure.
21610         PR tree-optimization/84687
21611         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
21612         on new_node->decl.
21613         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
21615 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21617         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
21618         Rename to ppc_speculation_barrier.
21619         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
21620         __builtin_ppc_speculation_barrier.
21622 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21624         PR target/84700
21625         * combine.c (combine_simplify_rtx): Don't try to simplify if
21626         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
21627         are equal to x.
21629 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
21631         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
21632         to 32 bytes when compiling for POWER9.
21634 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21636         PR target/84564
21637         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
21638         regparm >= 3 with no arg reg available also for calls with
21639         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
21641         PR target/84524
21642         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
21643         orig,vex.
21644         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
21646 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
21648         PR target/84264
21649         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
21651 2018-03-05  Richard Biener  <rguenther@suse.de>
21653         PR tree-optimization/84486
21654         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
21655         When inserting a __builtin_assume_aligned call set the LHS
21656         SSA name alignment info accordingly.
21658 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
21660         PR tree-optimization/84114
21661         * config/aarch64/aarch64.c (aarch64_reassociation_width)
21662         Avoid reassociation of FLOAT_MODE addition.
21664 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
21666         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
21667         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
21668         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
21669         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
21670         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
21671         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
21672         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
21673         and -mwbnoinvd.
21674         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
21675         __builtin_ia32_wbinvd): New builtins.
21676         (SPECIAL_ARGS2): New.
21677         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
21678         (SPECIAL_ARGS2): New.
21679         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
21680         (ix86_valid_target_attribute_inner_p): Ditto.
21681         (ix86_init_mmx_sse_builtins): Add special_args2.
21682         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
21683         TARGET_WBNOINVD_P): New.
21684         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
21685         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
21686         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
21687         * config/i386/immintrin.h (_wbinvd): New intrinsic.
21688         * config/i386/pconfigintrin.h: New file.
21689         * config/i386/wbnoinvdintrin.h: Ditto.
21690         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
21691         wbnoinvdintrin.h.
21692         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
21694 2018-03-05  Richard Biener  <rguenther@suse.de>
21696         PR tree-optimization/84670
21697         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21698         member.
21699         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21700         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21701         (compute_antic_aux): Only assert the number of values in ANTIC_IN
21702         doesn't grow if all successors (recursively) were visited at least
21703         once.
21705 2018-03-05  Richard Biener  <rguenther@suse.de>
21707         PR tree-optimization/84650
21708         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
21709         if executed in the loop pipeline.
21711 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
21713         * doc/configfiles.texi (Configuration Files): Move info about
21714         conditionalizing $target-protos.h to...
21715         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
21716         differs from $target-protos.h.
21718 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
21719             Chung-Ju Wu  <jasonwucj@gmail.com>
21721         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
21722         * config/nds32/nds32-multiple.md (setmemsi): Define.
21723         * config/nds32/nds32-memory-manipulation.c
21724         (nds32_gen_dup_4_byte_to_word_value): New.
21725         (emit_setmem_word_loop): New.
21726         (emit_setmem_byte_loop): New.
21727         (nds32_expand_setmem_loop): New.
21728         (nds32_expand_setmem_loop_v3m): New.
21729         (nds32_expand_setmem_unroll): New.
21730         (nds32_expand_setmem): New.
21732 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21733             Chung-Ju Wu  <jasonwucj@gmail.com>
21735         * config/nds32/nds32-memory-manipulation.c
21736         (nds32_emit_load_store): New.
21737         (nds32_emit_post_inc_load_store): New.
21738         (nds32_emit_mem_move): New.
21739         (nds32_emit_mem_move_block): New.
21740         (nds32_expand_movmemsi_loop_unknown_size): New.
21741         (nds32_expand_movmemsi_loop_known_size): New.
21742         (nds32_expand_movmemsi_loop): New.
21743         (nds32_expand_movmemsi_unroll): New.
21744         (nds32_expand_movmemqi): Rename ...
21745         (nds32_expand_movmemsi): ... to this.
21746         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
21747         (movmemsi): ... to this.
21748         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
21749         (nds32_expand_movmemsi): ... to this.
21751 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21752             Monk Chiang  <sh.chiang04@gmail.com>
21753             Chung-Ju Wu  <jasonwucj@gmail.com>
21755         * config/nds32/nds32-protos.h
21756         (nds32_expand_load_multiple): New arguments.
21757         (nds32_expand_store_multiple): Ditto.
21758         (nds32_valid_multiple_load_store): Rename ...
21759         (nds32_valid_multiple_load_store_p): ... to this.
21760         * config/nds32/nds32-memory-manipulation.c
21761         (nds32_expand_load_multiple): Refine implementation.
21762         (nds32_expand_store_multiple): Ditto.
21763         * config/nds32/nds32-multiple.md
21764         (load_multiple): Update nds32_expand_load_multiple interface.
21765         (store_multiple): Update nds32_expand_store_multiple interface.
21766         * config/nds32/nds32-predicates.c
21767         (nds32_valid_multiple_load_store): Rename ...
21768         (nds32_valid_multiple_load_store_p): ... to this and refine
21769         implementation.
21770         * config/nds32/predicates.md
21771         (nds32_load_multiple_and_update_address_operation): New predicate.
21772         (nds32_store_multiple_and_update_address_operation): New predicate.
21774 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21775             Chung-Ju Wu  <jasonwucj@gmail.com>
21777         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
21778         (combo): New attribute.
21779         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
21781 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21783         * config/nds32/nds32.opt: Change -mcmodel= default value.
21785 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
21786             Monk Chiang  <sh.chiang04@gmail.com>
21787             Chung-Ju Wu  <jasonwucj@gmail.com>
21789         * config/nds32/constants.md (unspec_element): New enum.
21790         * config/nds32/constraints.md (Umw): New constraint.
21791         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21792         * config/nds32/nds32-intrinsic.md: Likewise.
21793         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21794         (nds32_valid_smw_lwm_base_p): New.
21795         (nds32_output_smw_single_word): New.
21796         (nds32_output_lmw_single_word): New.
21797         (nds32_expand_unaligned_load): New.
21798         (nds32_expand_unaligned_store): New.
21799         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21800         (nds32_output_smw_single_word): Declare.
21801         (nds32_output_lmw_single_word): Declare.
21802         (nds32_expand_unaligned_load): Declare.
21803         (nds32_expand_unaligned_store): Declare.
21804         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21805         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21806         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21807         NDS32_BUILTIN_UASTORE_DW.
21808         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21809         predicate.
21811 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
21812             Kito Cheng  <kito.cheng@gmail.com>
21813             Chung-Ju Wu  <jasonwucj@gmail.com>
21815         * config/nds32/nds32-intrinsic.c
21816         (nds32_expand_builtin_null_ftype_reg): Delete.
21817         (nds32_expand_builtin_reg_ftype_imm): Ditto.
21818         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21819         (nds32_read_argument): New.
21820         (nds32_legitimize_target): Ditto.
21821         (nds32_legitimize_argument): Ditto.
21822         (nds32_check_constant_argument): Ditto.
21823         (nds32_expand_unop_builtin): Ditto.
21824         (nds32_expand_unopimm_builtin): Ditto.
21825         (nds32_expand_binop_builtin): Ditto.
21826         (nds32_builtin_decl_impl): Ditto.
21827         (builtin_description): Ditto.
21828         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21829         (nds32_init_builtins_impl): Ditto.
21830         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21831         (nds32_builtin_decl): New.
21832         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21833         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21835 2018-03-02  Jeff Law  <law@redhat.com>
21837         * reorg.c (stop_search_p): Handle DEBUG_INSN.
21838         (redundant_insn, fill_simple_delay_slots): Likewise.
21839         (fill_slots_from_thread): Likewise.
21840         * resource.c (mark_referenced_resources): Likewise.
21841         (mark_set_resources, find_dead_or_set_registers): Likewise.
21843 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21845         * substring-locations.h (format_warning_va): Formatting fix for
21846         ATTRIBUTE_GCC_DIAG.
21847         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21848         argument.
21849         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21850         * substring-locations.c: Include intl.h.
21851         (format_warning_va): Turned into small wrapper around
21852         format_warning_n_va, renamed to ...
21853         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21854         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21855         use ngettext.
21856         (format_warning_at_substring_n): New function.
21857         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21858         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
21859         format_warning_at_substring with just a shorter name instead of
21860         const function pointer.
21861         (fmtwarn_n): New function.
21862         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21863         appropriate, get rid of all the fmtstr temporaries, move conditionals
21864         with G_() wrapped string literals directly into fmtwarn arguments,
21865         cast dir.len to (int), formatting fixes.
21867 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
21869         * doc/invoke.texi: Remove "Cilk Plus" references.
21871 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21872             Richard Biener  <rguenther@suse.de>
21874         PR ipa/84628
21875         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21876         for error or warning attributes if CALL_FROM_THUNK_P is set.
21877         Formatting fixes.
21879 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21881         PR target/56540
21882         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21883         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21885         PR target/56540
21886         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21887         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21889         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21890         instead of -1U in last predictors element's probability member.
21892 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21894         PR ipa/83983
21895         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21896         arguments if they are comparable.
21898 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21900         PR tree-optimization/84634
21901         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21902         masks and masked_loop_p with a single loop_masks, making sure it's
21903         null for bb vectorization.
21905 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21907         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21908         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21909         loop->force_vectorize to check whether there is no alias.
21911 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21913         PR target/84614
21914         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21915         prototypes.
21916         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21917         comments.
21918         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21919         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21920         instead of a loop around prev_real_insn.
21921         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21922         prev_real_insn.
21924         PR inline-asm/84625
21925         * config/i386/i386.c (ix86_print_operand): Use conditional
21926         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21927         zero vector.
21929 2018-03-02  Richard Biener  <rguenther@suse.de>
21931         PR tree-optimization/84427
21932         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21933         (bitmap_set_subtract_values): Rewrite to handle multiple
21934         exprs per value.
21935         (clean): Likewise.
21936         (prune_clobbered_mems): Likewise.
21937         (phi_translate): Take edge instead of pred/phiblock.
21938         (phi_translate_1): Likewise.
21939         (phi_translate_set): Likewise.  Insert all translated
21940         exprs for a value into the set, keeping possibly multiple
21941         expressions per value.
21942         (compute_antic_aux): Adjust for phi_translate changes.
21943         When intersecting union the expressions and prune those
21944         not in the final value set, keeping possibly multiple
21945         expressions per value.  Do not use value-insertion
21946         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21947         all expressions.  Add verification that the value-sets
21948         only shrink during iteration.
21949         (compute_partial_antic_aux): Adjust for the phi_translate changes.
21950         (do_pre_regular_insertion): Likewise.
21951         (do_pre_partial_partial_insertion): Likewise.
21953 2018-03-02  Richard Biener  <rguenther@suse.de>
21955         PR target/82005
21956         * config/darwin.c (saved_debug_info_level): New static global.
21957         (darwin_asm_lto_start): Disable debug info generation for LTO out.
21958         (darwin_asm_lto_end): Restore debug info generation settings.
21960 2018-03-01  Martin Liska  <mliska@suse.cz>
21962         PR sanitizer/82484
21963         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21964         volatile arguments.
21966 2018-03-01  Richard Biener  <rguenther@suse.de>
21968         PR debug/84645
21969         * dwarf2out.c (gen_variable_die): Properly handle late VLA
21970         type annotation with LTO when debug was disabled at compile-time.
21972 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
21974         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21975         XINT with INTVAL.
21976         (mips_final_postscan_insn): Likewise.
21978 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
21980         PR rtl-optimization/84528
21981         * alias.c (init_alias_target): Add commentary.
21982         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21983         a unique base value if the frame pointer is not eliminated
21984         to the stack pointer.
21986 2018-03-01  Tom de Vries  <tom@codesourcery.com>
21988         PR rtl-optimization/83327
21989         * lra-int.h (hard_regs_spilled_into): Declare.
21990         * lra.c (hard_regs_spilled_into): Define.
21991         (init_reg_info): Init hard_regs_spilled_into.
21992         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21993         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21994         (process_bb_lives): Handle hard_regs_spilled_into.
21995         (lra_create_live_ranges_1): Before doing liveness propagation, clear
21996         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21998 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22000         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22001         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22002         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22003         * config/rs6000/aix72.h: New file.
22005 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22007         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22008         instead of warning_at with conditional singular and plural messages
22009         where possible.
22011         PR target/52991
22012         * stor-layout.c (update_alignment_for_field): For
22013         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22014         && !DECL_PACKED (field), do the alignment update, just use
22015         only desired_align instead of MAX (type_align, desired_align)
22016         as the alignment.
22017         (place_field): Don't do known_align < desired_align handling
22018         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22019         is non-NULL, instead do it after rli->prev_field handling and
22020         only if not within a bitfield word.  For DECL_PACKED (field)
22021         use type_align of BITS_PER_UNIT.
22023 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22025         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22026         superfluous parentheses and trailing spaces.
22028 2018-02-28  Richard Biener  <rguenther@suse.de>
22030         PR tree-optimization/84584
22031         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22032         SCoPs with fake exit edge.
22034 2018-02-28  Martin Liska  <mliska@suse.cz>
22036         PR testsuite/84597
22037         * timevar.c (timer::print): Fix format to properly print 100%
22038         values.
22040 2018-02-28  Richard Biener  <rguenther@suse.de>
22042         PR middle-end/84607
22043         * genmatch.c (capture_info::walk_match): Do not mark
22044         captured expressions without operands as expr_p given
22045         they act more like predicates and should be subject to
22046         "lost tail" side-effect preserving.
22048 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22050         PR rtl-optimization/81611
22051         * auto-inc-dec.c (attempt_change): Move dead note from
22052         mem_insn if it's the next use of regno
22053         (find_address): Take address use of reg holding
22054         non-incremented value.  Add parm to limit search to the named
22055         reg only.
22056         (merge_in_block): Attempt to use a mem insn that is the next
22057         use of the original regno.
22059 2018-02-27  Martin Sebor  <msebor@redhat.com>
22061         PR c++/83871
22062         * doc/invoke.texi (-Wmissing-attributes): New option.
22063         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22065 2018-02-27  Martin Sebor  <msebor@redhat.com>
22067         PR translation/84207
22068         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22069         n argument to unsigned HOST_WIDE_INT.
22070         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22071         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22072         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22073         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22075 2018-02-27  Richard Biener  <rguenther@suse.de>
22077         PR tree-optimization/84512
22078         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22079         Do not use the estimate returned from record_stmt_cost for
22080         the scalar iteration cost but sum properly using add_stmt_cost.
22082 2018-02-27  Richard Biener  <rguenther@suse.de>
22084         PR tree-optimization/84466
22085         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22086         Adjust last change to less strictly validate use operands.
22088 2018-02-27  Martin Liska  <mliska@suse.cz>
22090         PR gcov-profile/84548
22091         * gcov.c (process_file): Allow partial overlap and consider it
22092         also as group functions.
22093         (output_lines): Properly calculate range of lines for a group.
22095 2018-02-27  Martin Liska  <mliska@suse.cz>
22097         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22098         'ggc' suffixes.  Change first column width.
22099         (timer::print): Fix formatting of the column.
22101 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22103         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22104         preserve inline entry blocks for the sake of debug inline
22105         entry point markers alone.
22106         (remove_unused_locals): Suggest in comments a better place to
22107         force the preservation of inline entry blocks that are
22108         otherwise unused, but do not preserve them.
22110 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22112         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22114 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22116         PR target/84039
22117         * config/i386/constraints.md (Bs): Replace
22118         ix86_indirect_branch_register with
22119         TARGET_INDIRECT_BRANCH_REGISTER.
22120         (Bw): Likewise.
22121         * config/i386/i386.md (indirect_jump): Likewise.
22122         (tablejump): Likewise.
22123         (*sibcall_memory): Likewise.
22124         (*sibcall_value_memory): Likewise.
22125         Peepholes of indirect call and jump via memory: Likewise.
22126         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22127         (*sibcall_value_GOT_32): Likewise.
22128         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22129         (GOT_memory_operand): Likewise.
22130         (call_insn_operand): Likewise.
22131         (sibcall_insn_operand): Likewise.
22132         (GOT32_symbol_operand): Likewise.
22133         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22135 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22137         PR rtl-optimization/83496
22138         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22139         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22140         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22141         redundant insn, if any.
22142         (relax_delay_slots): Likewise.
22143         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22145 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22147         PR tree-optimization/83965
22148         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22149         that grouped statements are part of a reduction chain.  Return
22150         true if the statement is not marked as a reduction itself but
22151         is part of a group.
22152         (vect_recog_dot_prod_pattern): Don't check whether the statement
22153         is part of a group here.
22154         (vect_recog_sad_pattern): Likewise.
22155         (vect_recog_widen_sum_pattern): Likewise.
22157 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22159         PR debug/84545
22160         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22162 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22164         PR target/84530
22165         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22166         the bool argument.
22167         (ix86_output_indirect_function_return): New prototype.
22168         (ix86_split_simple_return_pop_internal): Likewise.
22169         * config/i386/i386.c (indirect_return_via_cx): New.
22170         (indirect_return_via_cx_bnd): Likewise.
22171         (indirect_thunk_name): Handle return va CX_REG.
22172         (output_indirect_thunk_function): Create alias for
22173         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22174         (ix86_output_indirect_jmp): Remove the bool argument.
22175         (ix86_output_indirect_function_return): New function.
22176         (ix86_split_simple_return_pop_internal): Likewise.
22177         * config/i386/i386.md (*indirect_jump): Don't pass false
22178         to ix86_output_indirect_jmp.
22179         (*tablejump_1): Likewise.
22180         (simple_return_pop_internal): Change it to define_insn_and_split.
22181         Call ix86_split_simple_return_pop_internal to split it for
22182         -mfunction-return=.
22183         (simple_return_indirect_internal): Call
22184         ix86_output_indirect_function_return instead of
22185         ix86_output_indirect_jmp.
22187 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22189         PR bootstrap/84405
22190         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22191         memset and value initialization afterwards.
22193 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22195         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22197 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22199         PR target/84521
22200         * common/config/aarch64/aarch64-common.c
22201         (aarch_option_optimization_table[]): Switch
22202         off fomit-frame-pointer
22204 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22205             Chung-Ju Wu  <jasonwucj@gmail.com>
22207         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22208         volatile memory.
22209         (store_multiple): Ditto.
22211 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22213         * config.gcc: Add --with-cpu support for nds32 target.
22214         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22215         * config/nds32/nds32.opt: Add -mcpu= option.
22217 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22219         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22220         isel=yes): Warn for these deprecated options.
22222 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22224         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22225         ISA_2_5_MASKS_EMBEDDED.
22227 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22229         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22230         p->max as pointers rather than using iterative_hash_expr.
22232 2018-02-23  Carl Love  <cel@us.ibm.com>
22234         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22235         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22236         BU_P8V_OVERLOAD_2.
22237         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22238         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22239         P8V_BUILTIN_VEC_VUNSIGNED2.
22241 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22243         PR target/81572
22244         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22245         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22246         LRA_UNKNOWN_ALT.
22247         * lra-constraints.c (curr_insn_transform): Set up
22248         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22249         LRA_UNKNOWN_ALT.
22250         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22251         * lra-eliminations.c (spill_pseudos): Ditto.
22252         (process_insn_for_elimination): Ditto.
22253         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22254         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22255         LRA_NON_CLOBBERED_ALT.
22257 2018-02-22  Martin Sebor  <msebor@redhat.com>
22259         PR tree-optimization/84480
22260         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22261         to maybe_diag_stxncpy_trunc.  Call it.
22262         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22263         from gimple_fold_builtin_strcpy.  Print inlining stack.
22264         (handle_builtin_stxncpy): Print inlining stack.
22265         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22267 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22269         PR target/84176
22270         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22271         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22272         and -fcheck-pointer-bounds are used together.
22273         (indirect_thunk_prefix): New enum.
22274         (indirect_thunk_need_prefix): New function.
22275         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22276         "_nt" instead of "_bnd" for NOTRACK prefix.
22277         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22278         (output_indirect_thunk_function): Likewise.
22279         (): Likewise.
22280         (ix86_code_end): Update output_indirect_thunk_function calls.
22281         (ix86_output_indirect_branch_via_reg): Replace
22282         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22283         (ix86_output_indirect_branch_via_push): Likewise.
22284         (ix86_output_function_return): Likewise.
22285         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22286         incompatible with -fcf-protection=branch and
22287         -fcheck-pointer-bounds.
22289 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22291         PR target/83335
22292         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22293         Change gcc_assert call to output_operand_lossage.
22295 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22297         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22299 2018-02-22  DJ Delorie  <dj@redhat.com>
22300             Sebastian Perta  <sebastian.perta@renesas.com>
22301             Oleg Endo  <olegendo@gcc.gnu.org>
22303         * config/rx/rx.c (rx_rtx_costs): New function.
22304         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22306 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22308         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22310 2018-02-22  Martin Liska  <mliska@suse.cz>
22312         PR driver/83193
22313         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22314         Add "native" as a possible value.
22316 2018-02-22  Martin Liska  <mliska@suse.cz>
22318         PR driver/83193
22319         * config/i386/i386.c (ix86_option_override_internal):
22320         Add "native" as a possible value for -march and -mtune.
22322 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22324         PR target/84502
22325         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22326         to all type variants.
22328         PR tree-optimization/84503
22329         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22330         width as info->bitpos + info->bitsize - start.
22331         (merged_store_group::merge_overlapping): Simplify width computation.
22332         (check_no_overlap): New function.
22333         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22334         start + width and last_order of the group, fail if check_no_overlap
22335         fails.
22336         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22337         to group if check_no_overlap fails.
22339 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22341         * config/rs6000/altivec.md: Delete contraint arguments to
22342         define_expand, define_split, and define_peephole2, and in
22343         define_insn_and_split if always unused.
22344         * config/rs6000/darwin.md: Ditto.
22345         * config/rs6000/dfp.md: Ditto.
22346         * config/rs6000/rs6000.md: Ditto.
22347         * config/rs6000/sync.md: Ditto.
22348         * config/rs6000/vector.md: Ditto.
22349         * config/rs6000/vsx.md: Ditto.
22351 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22353         * config/rs6000/altivec.md: Write output control strings as braced
22354         blocks instead of double-quoted strings.
22355         * config/rs6000/darwin.md: Ditto.
22356         * config/rs6000/rs6000.md: Ditto.
22357         * config/rs6000/vector.md: Ditto.
22358         * config/rs6000/vsx.md: Ditto.
22360 2018-02-21  Jason Merrill  <jason@redhat.com>
22362         PR c++/84314 - ICE with templates and fastcall attribute.
22363         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22365 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22367         * ipa-cp.c (determine_versionability): Fix comment typos.
22369 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22371         PR c/84229
22372         * ipa-cp.c (determine_versionability): Do not version functions caling
22373         va_arg_pack.
22375 2018-02-21  Martin Liska  <mliska@suse.cz>
22377         PR driver/83193
22378         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22379         Add "native" as a possible value.
22380         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22381         the macro when native cpu detection is available.
22383 2018-02-21  Martin Liska  <mliska@suse.cz>
22385         PR driver/83193
22386         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22387         Add "native" as a possible value.
22388         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22389         when native cpu detection is available.
22391 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
22392             Martin Sebor  <msebor@redhat.com>
22394         PR tree-optimization/84478
22395         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22396         false.
22397         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22398         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22399         support which is conservatively correct, for 2 only stay conservative
22400         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
22401         argument to the 2 argument get_range_strlen, adjust 6 arg
22402         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22403         false.
22404         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22405         (gimple_fold_builtin_strlen): Pass true as last argument to
22406         get_range_strlen.
22408 2018-02-20  Martin Sebor  <msebor@redhat.com>
22410         PR middle-end/84095
22411         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22412         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
22413         (builtin_memref::builtin_memref): Factor out parts into
22414         set_base_and_offset and call it.
22416 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22418         PR middle-end/84406
22419         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22420         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22421         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
22422         search at the associated MODE_INT.
22424 2018-02-20  Jeff Law  <law@redhat.com>
22426         PR middle-end/82123
22427         PR tree-optimization/81592
22428         PR middle-end/79257
22429         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22430         for range data rather than using global data.
22431         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22432         range data rather than using global data.
22433         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22434         pass it to children as needed.
22435         (struct directive::fmtresult): Similarly.
22436         (struct directive::set_width): Similarly.
22437         (struct directive::set_precision): Similarly.
22438         (format_integer, format_directive, parse_directive): Similarly.
22439         (format_none): Accept unnamed vr_values parameter.
22440         (format_percent, format_floating, format_character): Similarly.
22441         (format_string, format_plain): Similarly.
22442         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22443         the EVRP range analyzer for range data rather than using global data.
22444         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22445         gimple-ssa-evrp-analyze.h
22446         (class sprintf_dom_walker): Add after_dom_children member function.
22447         Add evrp_range_analyzer member.
22448         (sprintf_dom_walker::before_dom_children): Call into the EVRP
22449         range analyzer as needed.
22450         (sprintf_dom_walker::after_dom_children): New member function.
22451         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22452         if not optimizing.
22453         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22454         (evrp_range_analyzer::pop_to_marker): Likewise.
22456 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22458         PR tree-optimization/84419
22459         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22460         with the required type if its current type is compatible but
22461         different.
22463 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
22465         PR middle-end/82004
22466         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22467         after vectorization.
22469 2018-02-20  Martin Liska  <mliska@suse.cz>
22471         PR driver/83193
22472         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22473         possible values if we don't have a hint.
22475 2018-02-20  Martin Liska  <mliska@suse.cz>
22477         PR c/84310
22478         PR target/79747
22479         * final.c (shorten_branches): Build align_tab array with one
22480         more element.
22481         * opts.c (finish_options): Add alignment option limit check.
22482         (MAX_CODE_ALIGN): Likewise.
22483         (MAX_CODE_ALIGN_VALUE): Likewise.
22484         * doc/invoke.texi: Document maximum allowed option value for
22485         all -falign-* options.
22487 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
22489         PR target/84146
22490         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22491         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22492         * var-tracking.c (emit_note_insn_var_location): Remove all references
22493         to NOTE_INSN_CALL_ARG_LOCATION.
22494         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22495         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22496         Use copy_rtx_if_shared.
22497         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22498         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22499         (dwarf2out_var_location): Remove handling of
22500         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22501         on call_insn.
22502         * final.c (final_scan_insn): Remove all references to
22503         NOTE_INSN_CALL_ARG_LOCATION.
22504         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
22505         before dumping final insns.
22506         * except.c (emit_note_eh_region_end): Remove all references to
22507         NOTE_INSN_CALL_ARG_LOCATION.
22508         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
22509         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
22510         * config/arc/arc.c (hwloop_optimize): Likewise.
22511         * config/arm/arm.c (create_fix_barrier): Likewise.
22512         * config/s390/s390.c (s390_chunkify_start): Likewise.
22513         * config/sh/sh.c (find_barrier): Likewise.
22514         * config/i386/i386.c (rest_of_insert_endbranch,
22515         ix86_seh_fixup_eh_fallthru): Likewise.
22516         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
22517         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
22518         * config/frv/frv.c (frv_function_prologue): Likewise.
22519         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
22520         reg note.
22521         (note_outside_basic_block_p): Remove all references to
22522         NOTE_INSN_CALL_ARG_LOCATION.
22523         * gengtype.c (adjust_field_rtx_def): Likewise.
22524         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
22525         Likewise.
22526         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
22527         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
22529         PR c++/84444
22530         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
22531         is ADDR_EXPR.
22533         PR tree-optimization/84452
22534         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
22535         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
22536         is NULL.
22538 2018-02-19  Martin Liska  <mliska@suse.cz>
22540         PR sanitizer/82183
22541         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
22543 2018-02-19  Martin Liska  <mliska@suse.cz>
22544             Richard Sandiford  <richard.sandiford@linaro.org>
22546         PR tree-optimization/82491
22547         * gimple-fold.c (get_base_constructor): Make earlier bail out
22548         to prevent ubsan.
22550 2018-02-19  Carl Love  <cel@us.ibm.com>
22552         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
22553         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
22554         BU_P8V_OVERLOAD_1.
22555         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
22556         P8V_BUILTIN_VEC_NEG.
22558 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
22560         * config/rl78/rl78.md (movdf): New define expand.
22562 2018-02-19  Martin Liska  <mliska@suse.cz>
22564         PR other/80589
22565         * doc/invoke.texi: Fix typo.
22566         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
22568 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
22570         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
22571         handle rs6000_single_float and rs6000_double_float specially for
22572         e500 family CPUs.
22574 2018-02-16  Jeff Law  <law@redhat.com>
22576         * config/rx/rx.c (add_pop_cfi_notes): New function.;
22577         (pop_regs): Use it.
22579 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22581         PR ipa/84425
22582         * ipa-inline.c (inline_small_functions): Fix a typo.
22584 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22586         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
22588 2018-02-16  Carl Love  <cel@us.ibm.com>
22590         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
22591         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
22592         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
22593         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
22594         expansion to P8V_BUILTIN_VEC_FLOAT2.
22596 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
22598         PR rtl-optimization/70023
22599         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
22600         src_regno into account.
22602 2018-02-16  Carl Love  <cel@us.ibm.com>
22604         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
22605         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
22606         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
22607         * config/rs6000/rs6000.c: Remove case statements for
22608         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
22609         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
22610         and P9V_BUILTIN_VEC_VINSERT4B.
22611         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
22612         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
22613         * config/rs6000/vsx.md:
22614         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
22615         vec_insert4b.
22617 2018-02-16  Carl Love  <cel@us.ibm.com>
22619         * config/rs6000/altivec.h: Add builtin names vec_extract4b
22620         vec_insert4b.
22621         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
22622         definitions.
22623         * config/rs6000/rs6000-c.c: Add the definitions for
22624         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
22625         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
22626         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
22627         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
22628         definition for insert4b and define insn *insert3b_internal.
22629         * doc/extend.texi: Add documentation for vec_extract4b.
22631 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22633         * doc/extend.texi (Backwards Compatibility): Mention friend
22634         injection.  Note for-scope is deprecated.
22635         * doc/invoke.texi (-ffriend-injection): Deprecate.
22637 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
22639         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
22640         that moved to I2, also allow destinations that are a paradoxical
22641         subreg (instead of a normal reg).
22643 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
22645         PR target/83831
22646         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
22647         to QImode.
22649 2018-02-16  Richard Biener  <rguenther@suse.de>
22651         PR tree-optimization/84037
22652         PR tree-optimization/84016
22653         PR target/82862
22654         * config/i386/i386.c (ix86_builtin_vectorization_cost):
22655         Adjust vec_construct for the fact we need additional higher latency
22656         128bit inserts for AVX256 and AVX512 vector builds.
22657         (ix86_add_stmt_cost): Scale vector construction cost for
22658         elementwise loads.
22660 2018-02-16  Richard Biener  <rguenther@suse.de>
22662         PR tree-optimization/84417
22663         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
22664         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
22665         (non_rewritable_lvalue_p): Likewise, use poly-ints.
22667 2018-02-16  Martin Liska  <mliska@suse.cz>
22669         PR sanitizer/84307
22670         * internal-fn.def (ASAN_CHECK): Set proper flags.
22671         (ASAN_MARK): Likewise.
22673 2018-02-16  Julia Koval  <julia.koval@intel.com>
22675         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
22676         from PTA_CANNONLAKE.
22678 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22680         PR target/84272
22681         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
22682         Use ++iter rather than iter++ for std::list iterators.
22683         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
22684         defer deleting them until all nodes in the forest are processed.  Do
22685         free even leaf nodes.  Change to_process into auto_vec.
22687         PR bootstrap/84405
22688         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
22689         * vec.h (vec_default_construct): Use memset instead of placement new
22690         if BROKEN_VALUE_INITIALIZATION is defined.
22691         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
22692         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
22693         is defined.
22695         PR rtl-optimization/83723
22696         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
22697         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
22698         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
22699         recursive calls.
22700         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
22701         callers.
22702         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
22704 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
22706         PR rtl-optimization/81443
22707         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
22708         from inner REGs to paradoxical SUBREGs.
22710 2018-02-16  Richard Biener  <rguenther@suse.de>
22712         PR tree-optimization/84399
22713         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22714         For operands we can analyze at their definition make sure we can
22715         analyze them at each use as well.
22717 2018-02-16  Richard Biener  <rguenther@suse.de>
22719         PR tree-optimization/84190
22720         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
22721         volatile accesses if the decl isn't volatile.
22723 2018-02-15  Jason Merrill  <jason@redhat.com>
22725         PR c++/84314 - ICE with templates and fastcall attribute.
22726         * attribs.c (build_type_attribute_qual_variant): Don't clobber
22727         TYPE_CANONICAL on an existing type.
22729 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
22731         PR tree-optimization/84383
22732         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
22733         dstoff nor call operand_equal_p if dstbase is NULL.
22735         PR tree-optimization/84334
22736         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
22737         also a CONSTANT_CLASS_P, punt.
22739 2018-02-14  Jim Wilson  <jimw@sifive.com>
22741         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
22742         first SMALL_OPERAND check.  New local min_second_step.  Move assert
22743         to where locals are set.  Add TARGET_RVC support.
22744         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
22746 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
22748         * doc/invoke.texi: Correct -Wformat-overflow code sample.
22750 2018-02-14  Martin Sebor  <msebor@redhat.com>
22752         PR tree-optimization/83698
22753         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
22754         arrays constrain the offset range to their bounds.
22755         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
22756         (builtin_access::overlap): Avoid setting the size of overlap if it's
22757         already been set.
22758         (maybe_diag_overlap): Also consider arrays when deciding what values
22759         of offsets to include in diagnostics.
22761 2018-02-14  Martin Sebor  <msebor@redhat.com>
22763         PR c/84108
22764         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
22765         that correspond to the kind of a declaration.
22767 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
22769         PR target/83984
22770         * config/pa/pa.md: Load address of PIC label using the linkage table
22771         if the label is nonlocal.
22773 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22775         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
22776         warning message if user requests -maltivec=be.
22777         * doc/invoke.texi: Document deprecation of -maltivec=be.
22779 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
22781         PR target/84220
22782         * config/rs6000/rs6000-c.c: Update definitions for
22783         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
22784         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
22786 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22788         PR target/84239
22789         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22790         add _get_ssp intrinsics. Remove argument from
22791         __builtin_ia32_rdssp[d|q].
22792         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22793         * config/i386/i386-builtin.def: Remove argument from
22794         __builtin_ia32_rdssp[d|q].
22795         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22796         ix86_expand_special_args_builtin for _rdssp[d|q].
22797         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22798         Clear register before usage.
22799         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22800         Add documentation for new _get_ssp and _inc_ssp intrinsics.
22802 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
22804         PR tree-optimization/84357
22805         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22806         operand 1 of an ARRAY_REF too.
22808 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
22810         PR target/83831
22811         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22812         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22813         declarations.
22814         (set_of_reg): New struct.
22815         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22816         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22817         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22818         functions.
22819         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22820         Split into bitclr, bitset, bitinvert patterns if appropriate.
22821         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22822         use rx_fuse_in_memory_bitop.
22823         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22824         to named insn, correct maximum insn length.
22826 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
22828         PR target/79242
22829         * machmode.def: Define a complex mode for PARTIAL_INT.
22830         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22831         MODE_PARTIAL_INT.
22832         * doc/rtl.texi: Document CSPImode.
22833         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22834         handling.
22835         (msp430_hard_regno_nregs_with_padding): Likewise.
22837 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
22839         PR target/84279
22840         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22842 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
22844         PR rtl-optimization/84169
22845         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
22846         we generated a parallel as new i3 and we split that to new i2 and i3
22847         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
22848         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
22849         those to i2, not i1.  Partially rewrite this scan code.
22851 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22853         PR c/82210
22854         * stor-layout.c (place_field): For variable length fields, adjust
22855         offset_align afterwards not just based on the field's alignment,
22856         but also on the size.
22858         PR middle-end/84309
22859         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22860         of exps and logs in the use_exp2 case.
22862 2018-02-13  Jeff Law  <law@redhat.com>
22864         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22865         entry for "vector".
22867         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22868         ARGS as unused.
22870 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
22872         PR debug/84342
22873         PR debug/84319
22874         * common.opt (gas-loc-support, gas-locview-support): New.
22875         (ginline-points, ginternal-reset-location-views): New.
22876         * doc/invoke.texi: Document them.  Use @itemx where intended.
22877         (gvariable-location-views): Adjust.
22878         * target.def (reset_location_view): New.
22879         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22880         (TARGET_RESET_LOCATION_VIEW): New.
22881         * doc/tm.texi: Rebuilt.
22882         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22883         (dwarf2out_default_as_locview_support): New.
22884         (output_asm_line_debug_info): Use option variables.
22885         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22886         (output_loc_list): Likewise.
22887         (add_high_low_attributes): Check option variables.
22888         Don't output entry view attribute in strict mode.
22889         (gen_inlined_subroutine_die): Check option variables.
22890         (dwarf2out_inline_entry): Likewise.
22891         (init_sections_and_labels): Likewise.
22892         (dwarf2out_early_finish): Likewise.
22893         (maybe_reset_location_view): New, from...
22894         (dwarf2out_var_location): ... here.  Call it.
22895         * debug.h (dwarf2out_default_as_loc_support): Declare.
22896         (dwarf2out_default_as_locview_support): Declare.
22897         * hooks.c (hook_int_rtx_insn_0): New.
22898         * hooks.h (hook_int_rtx_insn_0): Declare.
22899         * toplev.c (process_options): Take -gas-loc-support and
22900         -gas-locview-support from dwarf2out.  Enable
22901         -gvariable-location-views by default only with locview
22902         assembler support.  Enable -ginternal-reset-location-views by
22903         default only if the target defines the corresponding hook.
22904         Enable -ginline-points by default if location views are
22905         enabled; force it disabled if statement frontiers are
22906         disabled.
22907         * tree-inline.c (expand_call_inline): Check option variables.
22908         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22910 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22912         PR tree-optimization/84321
22913         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22914         handling.  Also check whether the anti-range contains any values
22915         that satisfy the mask; switch to a VR_RANGE if not.
22917 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22919         PR sanitizer/84340
22920         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22922 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22924         PR c++/83990
22925         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22926         of call statements, also set location of a load to a temporary.
22928 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22930         * config/rl78/rl78.c (add_vector_labels): New function.
22931         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22932         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22933         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22934         which checks that no arguments are passed.
22935         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22936         * doc/extend.texi: Documentation for the new attribute.
22938 2018-02-13  Andreas Schwab  <schwab@suse.de>
22940         * config/riscv/linux.h (CPP_SPEC): Define.
22942 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22944         PR target/84335
22945         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22946         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22947         OPTION_MASK_ISA_AES as first argument to def_builtin_const
22948         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22949         instead of OPTION_MASK_ISA_PCLMUL as first argument to
22950         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22951         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22952         temporarily for AES and PCLMUL builtins.
22954         PR tree-optimization/84339
22955         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22956         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22957         Formatting fixes.
22959         PR middle-end/84309
22960         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22961         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22962         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22963         inline function.
22964         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22965         inline function.
22966         * omp-simd-clone.h: New file.
22967         * omp-simd-clone.c: Include omp-simd-clone.h.
22968         (expand_simd_clones): No longer static.
22969         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22970         cgraph.h and omp-simd-clone.h.
22971         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22972         (vect_recog_widen_shift_pattern): Formatting fix.
22973         (vect_pattern_recog_1): Don't check optab for calls.
22975         PR target/84336
22976         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22977         operands[2] into a REG before using gen_lowpart on it.
22979 2018-02-12  Jeff Law  <law@redhat.com>
22981         PR target/83760
22982         * config/sh/sh.c (find_barrier): Consider a sibling call
22983         a barrier as well.
22985         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22986         successfully back substituting a reg.
22988 2018-02-12  Richard Biener  <rguenther@suse.de>
22990         PR tree-optimization/84037
22991         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22992         parameter, move visited init to caller.
22993         (vect_slp_analyze_operations): Separate cost from validity
22994         check, initialize visited once for all instances.
22995         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22996         for all instances.
22997         * tree-vect-stmts.c (vect_model_simple_cost): Make early
22998         out an assert.
22999         (vect_model_promotion_demotion_cost): Likewise.
23000         (vectorizable_bswap): Guard cost modeling with !slp_node
23001         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23002         SLP stmts.
23003         (vectorizable_call): Likewise.
23004         (vectorizable_conversion): Likewise.
23005         (vectorizable_assignment): Likewise.
23006         (vectorizable_shift): Likewise.
23007         (vectorizable_operation): Likewise.
23008         (vectorizable_store): Likewise.
23009         (vectorizable_load): Likewise.
23010         (vectorizable_condition): Likewise.
23011         (vectorizable_comparison): Likewise.
23013 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23015         PR sanitizer/84307
23016         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23017         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23018         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23020 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23022         PR middle-end/83665
23023         * params.def (inline-min-speedup): Increase from 8 to 15.
23024         (max-inline-insns-auto): Decrease from 40 to 30.
23025         * ipa-split.c (consider_split): Add some buffer for function to
23026         be considered inlining candidate.
23027         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23028         default values.
23030 2018-02-12  Richard Biener  <rguenther@suse.de>
23032         PR tree-optimization/84037
23033         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23034         matched stmts if we cannot swap the non-matched ones.
23036 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23038         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23039         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23040         _mm_maskz_scalef_round_ss): New intrinsics.
23041         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23042         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23043         __builtin_ia32_scalefss_round): Remove.
23044         (__builtin_ia32_scalefsd_mask_round,
23045         __builtin_ia32_scalefss_mask_round): New intrinsics.
23046         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23047         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23048         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23049         "<round_constraint>")): Changed to ...
23050         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23051         "<round_scalar_constraint>")): ... this.
23052         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23053         %0, %1, %2<round_op3>}"): Changed to ...
23054         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23055         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23056         %2<round_scalar_mask_op3>}"): ... this.
23057         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23059 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23061         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23062         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23063         (_mm_maskz_sqrt_round_ss): New intrinsics.
23064         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23065         (__builtin_ia32_sqrtsd_mask_round)
23066         (__builtin_ia32_sqrtss_mask_round): New builtins.
23067         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23068         (__builtin_ia32_sqrtss_round): Remove.
23069         (__builtin_ia32_sqrtsd_mask_round)
23070         (__builtin_ia32_sqrtss_mask_round): New builtins.
23071         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23072         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23073         ((match_operand:VF_128 1 "vector_operand"
23074         "xBm,<round_constraint>")): Changed to ...
23075         ((match_operand:VF_128 1 "vector_operand"
23076         "xBm,<round_scalar_constraint>")): ... this.
23077         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23078         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23079         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23080         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23081         %<iptr>1<round_scalar_mask_op3>}): ... this.
23082         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23083         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23085 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23087         PR target/84266
23088         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23089         Cast vec_cmpeq result to correct type.
23090         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23091         Cast vec_cmpgt result to correct type.
23093 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23095         * final.c (final_scan_insn_1): Renamed from...
23096         (final_scan_insn): ... this.  New wrapper, to recover
23097         seen from the outermost call in recursive ones.
23098         * config/sparc/sparc.c (output_return): Drop seen from call.
23099         (output_sibcall): Likewise.
23100         * config/visium/visium.c (output_branch): Likewise.
23102 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23104         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23105         function label.
23107 2018-02-10  Alan Modra  <amodra@gmail.com>
23109         PR target/84300
23110         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23111         Specify LR as an input.
23113 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23115         PR sanitizer/83987
23116         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23117         remove_member_access_dummy_vars): New functions.
23118         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23119         lower_omp_1, execute_lower_omp): Use them.
23121         PR rtl-optimization/84308
23122         * shrink-wrap.c (spread_components): Release todo vector.
23124 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23126         PR rtl-optimization/57193
23127         * ira-color.c (struct allocno_color_data): Add member
23128         conflict_allocno_hard_prefs.
23129         (update_conflict_allocno_hard_prefs): New.
23130         (bucket_allocno_compare_func): Add a preference based on
23131         conflict_allocno_hard_prefs.
23132         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23133         (color_allocnos): Remove a dead code.  Initiate
23134         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23136 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23138         PR target/84226
23139         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23140         constraint from =wa to wa.  Avoid a subreg on the output operand,
23141         instead use a pseudo and subreg it in a move.
23142         (p9_xxbrd_<mode>): Changed to ...
23143         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23144         (p9_xxbrd_v2df): New expander.
23145         (p9_xxbrw_<mode>): Changed to ...
23146         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23147         (p9_xxbrw_v4sf): New expander.
23149 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23151         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23153 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23155         PR target/83926
23156         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23157         multiply in 32-bit mode.
23158         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23159         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23160         mode.
23162 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23164         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23165         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23166         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23167         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23169 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23171         PR lto/84213
23172         * dwarf2out.c (is_trivial_indirect_ref): New function.
23173         (dwarf2out_late_global_decl): Do not generate a location
23174         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23175         and that are not defined in the current unit.
23177 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23179         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23180         instead of a libcall for UNORDERED.
23182 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23184         PR target/82641
23185         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23186         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23188 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23190         PR target/PR84295
23191         * config/s390/s390.c (s390_set_current_function): Invoke
23192         s390_indirect_branch_settings also if fndecl didn't change.
23194 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23196         * config/rs6000/rs6000.md (blockage): Set length to zero.
23198 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23200         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23202 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23204         PR sanitizer/84285
23205         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23206         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23207         -static-lib*san.
23209         PR debug/84252
23210         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23211         PARALLEL incoming that failed vt_get_decl_and_offset check.
23213         PR middle-end/84237
23214         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23215         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23216         TREE_READONLY bit.
23217         (get_variable_section): For decls in named .bss* sections pass true as
23218         second argument to bss_initializer_p.
23220 2018-02-09  Marek Polacek  <polacek@redhat.com>
23221             Jakub Jelinek  <jakub@redhat.com>
23223         PR c++/83659
23224         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23225         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23226         Sync some changes from cxx_fold_indirect_ref.
23228 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23230         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23231         markers.
23232         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23233         (BLOCK_INLINE_ENTRY_LABEL): New.
23234         (dwarf2out_var_location): Disregard inline entry markers.
23235         (inline_entry_data): New struct.
23236         (inline_entry_data_hasher): New hashtable type.
23237         (inline_entry_data_hasher::hash): New.
23238         (inline_entry_data_hasher::equal): New.
23239         (inline_entry_data_table): New variable.
23240         (add_high_low_attributes): Add DW_AT_entry_pc and
23241         DW_AT_GNU_entry_view attributes if a pending entry is found
23242         in inline_entry_data_table.  Add old entry_pc attribute only
23243         if debug nonbinding markers are disabled.
23244         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23245         markers are enabled.
23246         (block_within_block_p, dwarf2out_inline_entry): New.
23247         (dwarf2out_finish): Check that no entries remained in
23248         inline_entry_data_table.
23249         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23250         (final_scan_insn, notice_source_line): Likewise.
23251         (rest_of_clean_state): Skip inline entry markers.
23252         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23253         markers.
23254         * gimple.c (gimple_build_debug_inline_entry): New.
23255         * gimple.h (enum gimple_debug_subcode): Add
23256         GIMPLE_DEBUG_INLINE_ENTRY.
23257         (gimple_build_debug_inline_entry): Declare.
23258         (gimple_debug_inline_entry_p): New.
23259         (gimple_debug_nonbind_marker_p): Adjust.
23260         * insn-notes.def (INLINE_ENTRY): New.
23261         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23262         inline entry marker notes.
23263         (print_insn): Likewise.
23264         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23265         (INSN_DEBUG_MARKER_KIND): Likewise.
23266         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23267         * tree-inline.c (expand_call_inline): Build and insert
23268         debug_inline_entry stmt.
23269         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23270         inline entry blocks early, if nonbind markers are enabled.
23271         (dump_scope_block): Dump fragment info.
23272         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23273         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23274         (gimple_build_debug_inline_entry): New.
23275         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23276         Enable/disable inline entry points too.
23277         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23278         (DEBUG_INSN): Describe inline entry markers.
23280         * common.opt (gvariable-location-views): New.
23281         (gvariable-location-views=incompat5): New.
23282         * config.in: Rebuilt.
23283         * configure: Rebuilt.
23284         * configure.ac: Test assembler for view support.
23285         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23286         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23287         * dwarf2out.c (var_loc_view): New typedef.
23288         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23289         (dwarf2out_locviews_in_attribute): New.
23290         (dwarf2out_locviews_in_loclist): New.
23291         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23292         (enum dw_line_info_opcode): Add LI_adv_address.
23293         (struct dw_line_info_table): Add view.
23294         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23295         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23296         (zero_view_p): New variable.
23297         (ZERO_VIEW_P): New macro.
23298         (output_asm_line_debug_info): New.
23299         (struct var_loc_node): Add view.
23300         (add_AT_view_list, AT_loc_list): New.
23301         (add_var_loc_to_decl): Add view param.  Test it against last.
23302         (new_loc_list): Add view params.  Record them.
23303         (AT_loc_list_ptr): Handle loc and view lists.
23304         (view_list_to_loc_list_val_node): New.
23305         (print_dw_val): Handle dw_val_class_view_list.
23306         (size_of_die): Likewise.
23307         (value_format): Likewise.
23308         (loc_list_has_views): New.
23309         (gen_llsym): Set vl_symbol too.
23310         (maybe_gen_llsym, skip_loc_list_entry): New.
23311         (dwarf2out_maybe_output_loclist_view_pair): New.
23312         (output_loc_list): Output view list or entries too.
23313         (output_view_list_offset): New.
23314         (output_die): Handle dw_val_class_view_list.
23315         (output_dwarf_version): New.
23316         (output_compilation_unit_header): Use it.
23317         (output_skeleton_debug_sections): Likewise.
23318         (output_rnglists, output_line_info): Likewise.
23319         (output_pubnames, output_aranges): Update version comments.
23320         (output_one_line_info_table): Output view numbers in asm comments.
23321         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23322         Call maybe_gen_llsym.
23323         (loc_list_from_tree_1): Adjust.
23324         (add_AT_location_description): Create view list attribute if
23325         needed, check it's absent otherwise.
23326         (convert_cfa_to_fb_loc_list): Adjust.
23327         (maybe_emit_file): Call output_asm_line_debug_info for test.
23328         (dwarf2out_var_location): Reset views as needed.  Precompute
23329         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23330         attribute.  Set view.
23331         (new_line_info_table): Reset next view.
23332         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23333         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23334         the assembler, or select appropriate line info opcodes.
23335         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23336         (optimize_string_length): Catch it.  Adjust.
23337         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23338         dw_val_class_view_list, and remove it if no longer needed.
23339         (hash_loc_list): Hash view numbers.
23340         (loc_list_hasher::equal): Compare them.
23341         (optimize_location_lists): Check whether a view list symbol is
23342         needed, and whether the locview attribute is present, and
23343         whether they match.  Remove the locview attribute if no longer
23344         needed.
23345         (index_location_lists): Call skip_loc_list_entry for test.
23346         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23347         Use output_dwarf_version.
23348         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23349         (struct dw_val_node): Add val_view_list.
23350         * final.c (SEEN_NEXT_VIEW): New.
23351         (set_next_view_needed): New.
23352         (clear_next_view_needed): New.
23353         (maybe_output_next_view): New.
23354         (final_start_function): Rename to...
23355         (final_start_function_1): ... this.  Take pointer to FIRST,
23356         add SEEN parameter.  Emit param bindings in the initial view.
23357         (final_start_function): Reintroduce SEEN-less interface.
23358         (final): Rename to...
23359         (final_1): ... this.  Take SEEN parameter.  Output final pending
23360         next view at the end.
23361         (final): Reintroduce seen-less interface.
23362         (final_scan_insn): Output pending next view before switching
23363         sections or ending a block.  Mark the next view as needed when
23364         outputting variable locations.  Notify debug backend of section
23365         changes, and of location view changes.
23366         (rest_of_handle_final): Adjust.
23367         * toplev.c (process_options): Autodetect value for debug variable
23368         location views option.  Warn on incompat5 without -gdwarf-5.
23369         * doc/invoke.texi (gvariable-location-views): New.
23370         (gvariable-location-views=incompat5): New.
23371         (gno-variable-location-views): New.
23373 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23375         PR tree-optimization/84136
23376         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23377         that the result of find_edge is non-NULL.
23379 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23381         PR target/83008
23382         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23383         storing integer register in SImode.  Fix cost of 256 and 512
23384         byte aligned SSE register store.
23386 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23388         * config/i386/i386.c (ix86_multiplication_cost): Fix
23389         multiplication cost for TARGET_AVX512DQ.
23391 2018-02-08  Marek Polacek  <polacek@redhat.com>
23393         PR tree-optimization/84238
23394         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23395         get_range_strlen.
23397 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23399         PR tree-optimization/84265
23400         * tree-vect-stmts.c (vectorizable_store): Don't treat
23401         VMAT_CONTIGUOUS accesses as grouped.
23402         (vectorizable_load): Likewise.
23404 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23406         PR tree-optimization/81635
23407         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23408         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23409         (test_round_for_mask): New functions.
23410         (wide_int_cc_tests): Call test_round_for_mask.
23411         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23412         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23413         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23414         range returned by get_range_info.
23416 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23418         PR ipa/81360
23419         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23420         * symtab.c: Include builtins.h
23421         (symtab_node::output_to_lto_symbol_table_p): Move here
23422         from lto-streamer-out.c:output_symbol_p.
23423         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23424         (output_symbol_p): Move all logic to symtab.c
23425         (produce_symtab): Update.
23427 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23429         * config/s390/s390-opts.h (enum indirect_branch): Define.
23430         * config/s390/s390-protos.h (s390_return_addr_from_memory)
23431         (s390_indirect_branch_via_thunk)
23432         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23433         (enum s390_indirect_branch_type): Define.
23434         * config/s390/s390.c (struct s390_frame_layout, struct
23435         machine_function): Remove.
23436         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23437         (indirect_branch_table_label_no, indirect_branch_table_name):
23438         Define variables.
23439         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23440         (enum s390_indirect_branch_option): Define.
23441         (s390_return_addr_from_memory): New function.
23442         (s390_handle_string_attribute): New function.
23443         (s390_attribute_table): Add new attribute handler.
23444         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23445         (s390_indirect_branch_via_thunk): New function.
23446         (s390_indirect_branch_via_inline_thunk): New function.
23447         (s390_function_ok_for_sibcall): When jumping via thunk disallow
23448         sibling call optimization for non z10 compiles.
23449         (s390_emit_call): Force indirect branch target to be a single
23450         register.  Add r1 clobber for non-z10 compiles.
23451         (s390_emit_epilogue): Emit return jump via return_use expander.
23452         (s390_reorg): Handle JUMP_INSNs as execute targets.
23453         (s390_option_override_internal): Perform validity checks for the
23454         new command line options.
23455         (s390_indirect_branch_attrvalue): New function.
23456         (s390_indirect_branch_settings): New function.
23457         (s390_set_current_function): Invoke s390_indirect_branch_settings.
23458         (s390_output_indirect_thunk_function):  New function.
23459         (s390_code_end): Implement target hook.
23460         (s390_case_values_threshold): Implement target hook.
23461         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23462         macros.
23463         * config/s390/s390.h (struct s390_frame_layout)
23464         (struct machine_function): Move here from s390.c.
23465         (TARGET_INDIRECT_BRANCH_NOBP_RET)
23466         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23467         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23468         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23469         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23470         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23471         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23472         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23473         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23474         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23475         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23476         (mnemonic attribute): Add values which aren't recognized
23477         automatically.
23478         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23479         pattern for branch conversion.  Fix mnemonic attribute.
23480         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23481         indirect branch via thunk if requested.
23482         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23483         ("*indirect_jump"): Disable for branch conversion using out of
23484         line thunks.
23485         ("indirect_jump_via_thunk<mode>_z10")
23486         ("indirect_jump_via_thunk<mode>")
23487         ("indirect_jump_via_inlinethunk<mode>_z10")
23488         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23489         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23490         ("casesi_jump_via_inlinethunk<mode>_z10")
23491         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23492         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23493         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23494         ("*indirect2_jump"): Disable for branch conversion.
23495         ("casesi_jump"): Turn into expander and expand patterns for branch
23496         conversion.
23497         ("return_use"): New expander.
23498         ("*return"): Emit return via thunk and rename it to ...
23499         ("*return<mode>"): ... this one.
23500         * config/s390/s390.opt: Add new options and and enum for the
23501         option values.
23503 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23505         * lra-constraints.c (match_reload): Unconditionally use
23506         gen_lowpart_SUBREG, rather than selecting between that
23507         and equivalent gen_rtx_SUBREG code.
23509 2018-02-08  Richard Biener  <rguenther@suse.de>
23511         PR tree-optimization/84233
23512         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
23513         changed flag instead of boguously re-using phi_inserted.
23515 2018-02-08  Martin Jambor  <mjambor@suse.cz>
23517         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
23518         static local variables.
23520 2018-02-08  Richard Biener  <rguenther@suse.de>
23522         PR tree-optimization/84278
23523         * tree-vect-stmts.c (vectorizable_store): When looking for
23524         smaller vector types to perform grouped strided loads/stores
23525         make sure the mode is supported by the target.
23526         (vectorizable_load): Likewise.
23528 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23530         * config/aarch64/aarch64.c (aarch64_components_for_bb):
23531         Increase LDP/STP opportunities by adding adjacent callee-saves.
23533 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23535         PR rtl-optimization/84068
23536         PR rtl-optimization/83459
23537         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
23539 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
23541         PR tree-optimization/84224
23542         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
23543         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
23544         non-zero arguments.
23546 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
23548         PR target/84113
23549         * config/rs6000/altivec.md (*restore_world): Remove LR use.
23550         * config/rs6000/predicates.md (restore_world_operation): Adjust op
23551         count, remove one USE.
23553 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
23555         * doc/install.texi (Configuration): Document the
23556         --with-long-double-format={ibm,ieee} PowerPC configuration
23557         options.
23559         PR target/84154
23560         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
23561         Convert from define_expand to be define_insn_and_split.  Rework
23562         float/double/_Float128 conversions to QI/HI/SImode to work with
23563         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
23564         conversions to QI/HImode types did a store and then a load to
23565         truncate the value.  For conversions to VSX registers, don't split
23566         the insn, instead emit the code directly.  Use the code iterator
23567         any_fix to combine signed and unsigned conversions.
23568         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
23569         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
23570         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23571         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
23572         (fix_<mode>di2_hw): Likewise.
23573         (fixuns_<mode>di2_hw): Likewise.
23574         (fix_<mode>si2_hw): Likewise.
23575         (fixuns_<mode>si2_hw): Likewise.
23576         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
23577         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23578         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
23579         fix<uns>_trunc<SFDF:mode>si2_p8.
23580         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
23581         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23582         (fix<uns>_<mode>_mem): Likewise.
23583         (fctiw<u>z_<mode>_mem): Likewise.
23584         (fix<uns>_<mode>_mem): Likewise.
23585         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
23586         the register allocator from doing a direct move to the GPRs to do
23587         a store, and instead use the ISA 3.0 store byte/half-word from
23588         vector register instruction.  For IEEE 128-bit floating point,
23589         also optimize stores of 32-bit ints.
23590         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
23592 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
23594         * genextract.c (push_pathstr_operand): New function to support
23595         [a-zA-Z].
23596         (walk_rtx): Call push_pathstr_operand.
23597         (print_path): Support [a-zA-Z].
23599 2018-02-07  Richard Biener  <rguenther@suse.de>
23601         PR tree-optimization/84037
23602         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
23603         (cse_and_gimplify_to_preheader): Declare.
23604         (vect_get_place_in_interleaving_chain): Likewise.
23605         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23606         ivexpr_map.
23607         (_loop_vec_info::~_loop_vec_info): Delete it.
23608         (cse_and_gimplify_to_preheader): New function.
23609         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
23610         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
23611         (vectorizable_load): Likewise.  For grouped stores always base
23612         the IV on the first element.
23613         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
23614         condition before gimplifying.
23616 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23618         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
23619         *DIV_EXPR and *MOD_EXPR.
23621 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
23623         PR target/84248
23624         * config/i386/i386.c (ix86_option_override_internal): Mask out
23625         the CF_SET bit when checking -fcf-protection.
23627 2018-02-07  Tom de Vries  <tom@codesourcery.com>
23629         PR libgomp/84217
23630         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
23631         enough.
23633 2018-02-07  Richard Biener  <rguenther@suse.de>
23635         PR tree-optimization/84204
23636         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
23637         this place.
23639         PR tree-optimization/84205
23640         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
23641         special-case isl_ast_op_zdiv_r.
23643         PR tree-optimization/84223
23644         * graphite-scop-detection.c (gather_bbs::before_dom_children):
23645         Only add conditions from within the region.
23646         (gather_bbs::after_dom_children): Adjust.
23648 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
23650         PR target/84209
23651         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
23652         * config/avr/avr.md: Only post-reload split REG-REG moves if
23653         either register is GENERAL_REG_P.
23655 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23657         PR tree-optimization/84235
23658         * tree-ssa-scopedtables.c
23659         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
23660         if the subtraction is performed in floating point type where NaNs are
23661         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
23662         build 1.  Formatting fix.
23664 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
23666         PR target/84146
23667         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
23668         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
23669         and skip it regardless of bb boundaries.  Use CALL_P macro,
23670         don't test INSN_P (insn) together with CALL_P or JUMP_P check
23671         unnecessarily, formatting fix.
23673 2018-02-06  Michael Collison  <michael.collison@arm.com>
23675         * config/arm/thumb2.md:
23676         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
23677         (*thumb_mov_notscc): Ditto.
23679 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
23681         PR target/84154
23682         * config/rs6000/rs6000.md (su code attribute): Use "u" for
23683         unsigned_fix, not "s".
23685 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23687         * configure.ac (gcc_fn_eh_frame_ro): New function.
23688         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
23689         correct .eh_frame permissions.
23690         * configure: Regenerate.
23692 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
23694         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
23695         irrelevant options.
23697 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23699         * config/rs6000/rs6000.c (rs6000_option_override_internal):
23700         Display warning message for -mno-speculate-indirect-jumps.
23702 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
23704         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
23705         Undocumented.
23706         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
23708 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
23710         PR tree-optimization/84225
23711         * tree-eh.c (find_trapping_overflow): Only call
23712         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
23714 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23716         PR target/84145
23717         * config/i386/i386.c: Reimplement the check of possible options
23718         -mibt/-mshstk conbination. Change error messages.
23719         * doc/invoke.texi: Fix a typo: remove extra '='.
23721 2018-02-06  Marek Polacek  <polacek@redhat.com>
23723         PR tree-optimization/84228
23724         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
23726 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
23728         PR target/82641
23729         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
23730         emitted arch directives.
23731         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
23732         __ARM_FEATURE_COPROC before changing architectures.
23734 2018-02-06  Richard Biener  <rguenther@suse.de>
23736         * config/i386/i386.c (print_reg): Fix typo.
23737         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
23739 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
23741         * configure: Regenerate.
23743 2018-02-05  Martin Sebor  <msebor@redhat.com>
23745         PR tree-optimization/83369
23746         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
23747         inlining context.
23749 2018-02-05  Martin Liska  <mliska@suse.cz>
23751         * doc/invoke.texi: Cherry-pick upstream r323995.
23753 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
23755         * ira.c (ira_init_register_move_cost): Adjust comment.
23757 2018-02-05  Martin Liska  <mliska@suse.cz>
23759         PR gcov-profile/84137
23760         * doc/gcov.texi: Fix typo in documentation.
23762 2018-02-05  Martin Liska  <mliska@suse.cz>
23764         PR gcov-profile/83879
23765         * doc/gcov.texi: Document necessity of --dynamic-list-data when
23766         using dlopen functionality.
23768 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
23770         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
23771         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
23772         _mm_maskz_range_ss, _mm_mask_range_round_ss,
23773         _mm_maskz_range_round_ss): New intrinsics.
23774         (__builtin_ia32_rangesd128_round)
23775         (__builtin_ia32_rangess128_round): Remove.
23776         (__builtin_ia32_rangesd128_mask_round,
23777         __builtin_ia32_rangess128_mask_round): New builtins.
23778         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
23779         __builtin_ia32_rangess128_round): Remove.
23780         (__builtin_ia32_rangesd128_mask_round,
23781         __builtin_ia32_rangess128_mask_round): New builtins.
23782         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
23783         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
23784         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
23785         "<round_saeonly_constraint>")): Changed to ...
23786         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
23787         "<round_saeonly_scalar_constraint>")): ... this.
23788         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23789         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23790         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23791         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23792         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23794 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
23796         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23797         options.
23798         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23799         Remove all values except native, 8540 and 8548.
23801 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
23803         * config/i386/i386.c (ix86_output_function_return): Pass
23804         INVALID_REGNUM, instead of -1, as invalid register number to
23805         indirect_thunk_name and output_indirect_thunk.
23807 2018-02-02  Julia Koval  <julia.koval@intel.com>
23809         * config.gcc: Add -march=icelake.
23810         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23811         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23812         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23813         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23814         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23815         (processor_target_table): Add icelake.
23816         (ix86_option_override_internal): Handle new PTAs.
23817         (get_builtin_code_for_version): Handle icelake.
23818         (M_INTEL_COREI7_ICELAKE): New.
23819         (fold_builtin_cpu): Handle icelake.
23820         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23821         * doc/invoke.texi: Add -march=icelake.
23823 2018-02-02  Julia Koval  <julia.koval@intel.com>
23825         * config/i386/i386.c (ix86_option_override_internal): Change flags type
23826         to wide_int_bitmask.
23827         * wide-int-bitmask.h: New.
23829 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23831         PR target/84066
23832         * config/i386/i386.md: Replace Pmode with word_mode in
23833         builtin_setjmp_setup and builtin_longjmp to support x32.
23835 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
23837         PR target/56010
23838         PR target/83743
23839         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23840         #include "opts.h".
23841         (rs6000_supported_cpu_names): New static variable.
23842         (linux_cpu_translation_table): Likewise.
23843         (elf_platform) <cpu>: Define new static variable and use it.
23844         Translate kernel AT_PLATFORM name to canonical name if needed.
23845         Error if platform name is unknown.
23847 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
23849         PR target/84089
23850         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23852 2018-02-01  Jeff Law  <law@redhat.com>
23854         PR target/84128
23855         * config/i386/i386.c (release_scratch_register_on_entry): Add new
23856         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
23857         the scratch if RELEASE_VIA_POP is false.
23858         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23859         If we have to save a temporary register, decrement SIZE appropriately.
23860         Pass new arguments to release_scratch_register_on_entry.
23861         (ix86_adjust_stack_and_probe): Likewise.
23862         (ix86_emit_probe_stack_range): Pass new arguments to
23863         release_scratch_register_on_entry.
23865 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
23867         PR rtl-optimization/84157
23868         * combine.c (change_zero_ext): Use REG_P predicate in
23869         front of HARD_REGISTER_P predicate.
23871 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
23873         * config/avr/avr.c (avr_option_override): Move disabling of
23874         -fdelete-null-pointer-checks to...
23875         * common/config/avr/avr-common.c (avr_option_optimization_table):
23876         ...here.
23878 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23880         PR tree-optimization/81635
23881         * tree-data-ref.c (split_constant_offset_1): For types that
23882         wrap on overflow, try to use range info to prove that wrapping
23883         cannot occur.
23885 2018-02-01  Renlin Li  <renlin.li@arm.com>
23887         PR target/83370
23888         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23889         TAILCALL_ADDR_REGS.
23890         (aarch64_register_move_cost): Likewise.
23891         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23892         TAILCALL_ADDR_REGS.
23893         (REG_CLASS_NAMES): Likewise.
23894         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23895         TAILCALL_ADDR_REGS. Remove IP registers.
23896         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23898 2018-02-01  Richard Biener  <rguenther@suse.de>
23900         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23901         for specifying RPO order and allow NULL for that.
23902         * domwalk.c (dom_walker::dom_walker): Likewise.
23903         (dom_walker::walk): Handle NULL RPO order.
23904         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23905         in RPO order.
23906         (rewrite_update_dom_walker): Likewise.
23907         (mark_def_dom_walker): Likewise.
23909 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23911         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23912         (aarch64_maybe_expand_sve_subreg_move): Declare.
23913         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23914         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23915         predicate.
23916         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23917         that are semantically a reverse operation.
23918         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23919         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23920         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23921         functions.
23922         (aarch64_can_change_mode_class): For big-endian, forbid changes
23923         between two SVE modes if they have different element sizes.
23925 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23927         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23928         the TImode handling for big-endian targets.
23930 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23932         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23933         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23934         not just bytes.
23935         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23936         Remove BSWAP handing for big-endian targets and use the form of
23937         LD1RQ appropariate for the mode.
23939 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23941         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23942         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23943         duplicated element.
23945 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23947         PR tearget/83845
23948         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23949         check for operands that need to go through aarch64_sve_reload_be.
23951 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
23953         PR tree-optimization/81661
23954         PR tree-optimization/84117
23955         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23956         * tree-eh.c: Include gimplify.h.
23957         (find_trapping_overflow, replace_trapping_overflow,
23958         rewrite_to_non_trapping_overflow): New functions.
23959         * tree-vect-loop.c: Include tree-eh.h.
23960         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23961         * tree-data-ref.c: Include tree-eh.h.
23962         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23964 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
23966         PR rtl-optimization/84123
23967         * combine.c (change_zero_ext): Check if hard register satisfies
23968         can_change_dest_mode before calling gen_lowpart_SUBREG.
23970 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
23972         PR target/82444
23973         * ira.c (ira_init_register_move_cost): Remove assert.
23975 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23977         PR rtl-optimization/84071
23978         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23979         * doc/tm.texi: Regenerate.
23981 2018-01-31  Richard Biener  <rguenther@suse.de>
23983         PR tree-optimization/84132
23984         * tree-data-ref.c (analyze_miv_subscript): Properly
23985         check whether evolution_function_is_affine_multivariate_p
23986         before calling gcd_of_steps_may_divide_p.
23988 2018-01-31  Julia Koval  <julia.koval@intel.com>
23990         PR target/83618
23991         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23992         * config/i386/i386.md (rdpid_rex64) New.
23993         (rdpid): Make 32bit only.
23995 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
23997         PR lto/84105
23998         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23999         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24001 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24003         Revert
24004         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24006         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24008 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24010         PR rtl-optimization/84071
24011         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24012         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24014 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24016         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24017         (arc_attribute_table): Add 'aux' attribute.
24018         (arc_in_small_data_p): Consider aux like variables.
24019         (arc_is_aux_reg_p): New function.
24020         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24021         (arc_get_aux_arg): New function.
24022         (prepare_move_operands): Handle aux-register access.
24023         (arc_handle_aux_attribute): New function.
24024         * doc/extend.texi (ARC Variable attributes): Add subsection.
24026 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24028         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24029         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24030         (arc_attribute_table): Add 'uncached' attribute.
24031         (arc_print_operand): Print '.di' flag for uncached memory
24032         accesses.
24033         (arc_in_small_data_p): Do not consider for small data the uncached
24034         types.
24035         (arc_is_uncached_mem_p): New function.
24036         * config/arc/predicates.md (compact_store_memory_operand): Check
24037         for uncached memory accesses.
24038         (nonvol_nonimm_operand): Likewise.
24039         * doc/extend.texi (ARC Type Attribute): New subsection.
24041 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24043         PR c/84100
24044         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24045         falign-loops=): Add Optimization flag.
24047 2018-01-30  Jeff Law  <law@redhat.com>
24049         PR target/84064
24050         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24051         INT_REGISTERS_SAVED.  Check it prior to calling
24052         get_scratch_register_on_entry.
24053         (ix86_adjust_stack_and_probe): Similarly.
24054         (ix86_emit_probe_stack_range): Similarly.
24055         (ix86_expand_prologue): Corresponding changes.
24057 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24059         PR target/40411
24060         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24061         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24063 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24065         PR target/84112
24066         * lra-constraints.c (curr_insn_transform): Process AND in the
24067         address.
24069 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24071         PR rtl-optimization/83986
24072         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24073         dependence against last_pending_memory_flush in addition to
24074         pending_jump_insns.
24076 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24078         PR tree-optimization/81611
24079         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24080         copies.
24082 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24084         PR target/83758
24085         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24086         a reg rtx.
24088 2018-01-30  Richard Biener  <rguenther@suse.de>
24089             Jakub Jelinek  <jakub@redhat.com>
24091         PR tree-optimization/84111
24092         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24093         inner loops added during recursion, as they don't have up-to-date
24094         SSA form.
24096 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24098         PR ipa/81360
24099         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24100         (can_inline_edge_by_limits_p): ... here.
24101         (can_early_inline_edge_p, check_callers,
24102         update_caller_keys, update_callee_keys, recursive_inlining,
24103         add_new_edges_to_heap, speculation_useful_p,
24104         inline_small_functions,
24105         inline_small_functions, flatten_function,
24106         inline_to_all_callers_1): Update.
24108 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24110         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24111         zeros correctly.
24113 2018-01-30  Richard Biener  <rguenther@suse.de>
24115         PR tree-optimization/83008
24116         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24117         invariant and constant vector uses in stmts when they need
24118         more than one stmt.
24120 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24122         PR bootstrap/84017
24123         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24124         * configure: Regenerate.
24126 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24128         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24129         pattern.
24130         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24131         Use gen_rtx_REG rather than gen_lowpart.
24133 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24135         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24136         rather than 0 when creating partial subregs.
24138 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24140         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24141         of usage.
24143 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24145         PR target/81550
24146         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24147         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24148         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24149         flags.  This restores the settings used before the 2017-07-24.
24150         Turning off pre increment/decrement/modify allows IVOPTS to
24151         optimize DF/SF loops where the index is an int.
24153 2018-01-29  Richard Biener  <rguenther@suse.de>
24154             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24156         PR bootstrap/80867
24157         * tree-vect-stmts.c (vectorizable_call): Don't call
24158         targetm.vectorize_builtin_md_vectorized_function if callee is
24159         NULL.
24161 2018-01-22  Carl Love  <cel@us.ibm.com>
24163         * doc/extend.tex: Fix typo in second arg in
24164         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24166 2018-01-29  Richard Biener  <rguenther@suse.de>
24168         PR tree-optimization/84086
24169         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24170         (flush_ssaname_freelist): When SSA names were released reset
24171         the SCEV hash table.
24173 2018-01-29  Richard Biener  <rguenther@suse.de>
24175         PR tree-optimization/84057
24176         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24177         removed paths when removing edges.
24179 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24181         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24182         -mfunction-return=@var{choice}.
24184 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24186         PR diagnostic/84034
24187         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24188         Handle CR like TAB.
24189         (layout::print_source_line): Likewise.
24190         (test_get_line_width_without_trailing_whitespace): Add test cases.
24192 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24194         PR middle-end/84040
24195         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24196         debug insns.
24198 2018-01-26  Jim Wilson  <jimw@sifive.com>
24200         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24202         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24203         specified.
24205 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24207         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24208         and CMP + SUB-immediate -> SUBS.
24210 2018-01-26  Martin Sebor  <msebor@redhat.com>
24212         PR tree-optimization/83896
24213         * tree-ssa-strlen.c (get_string_len): Rename...
24214         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24215         Avoid assuming length is constant.
24216         (handle_char_store): Use HOST_WIDE_INT for string length.
24218 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24220         PR target/81763
24221         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24222         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24224 2018-01-26  Richard Biener  <rguenther@suse.de>
24226         PR rtl-optimization/84003
24227         * dse.c (record_store): Only record redundant stores when
24228         the earlier store aliases at least all accesses the later one does.
24230 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24232         PR rtl-optimization/83985
24233         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24234         REG_CFA_RESTORE insns.
24235         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24236         REG_CFA_RESTORE insns here.
24238         PR c/83989
24239         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24240         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24242 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24244         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24245         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24246         (arc_init): Likewise.
24247         (arc_override_options): Likewise.
24248         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24249         value.
24250         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24251         support.
24252         * config/arc/arc.h (TARGET_DBNZ): Define.
24253         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24254         properly set the tune attribute.
24255         (dbnz): Use TARGET_DBNZ guard.
24256         * config/arc/arc.opt (mtune): Add core3 option.
24258 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24260         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24261         recognize new pic like addresses.
24262         (arc_delegitimize_address): Clean up.
24264 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24266         * config/arc/arc-arches.def: Option mrf16 valid for all
24267         architectures.
24268         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24269         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24270         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24271         * config/arc/arc-tables.opt: Regenerate.
24272         * config/arc/arc.c (arc_conditional_register_usage): Handle
24273         reduced register file case.
24274         (arc_file_start): Set must have build attributes.
24275         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24276         mrf16 option value.
24277         * config/arc/arc.opt (mrf16): Add new option.
24278         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24279         * config/arc/genmultilib.awk: Handle new mrf16 option.
24280         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24281         * config/arc/t-multilib: Regenerate.
24282         * doc/invoke.texi (ARC Options): Document mrf16 option.
24284 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24286         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24287         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24288         (arc_attribute_table): Add 'secure_call' attribute.
24289         (arc_print_operand): Print secure call operand.
24290         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24291         secure.
24292         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24293         (call_i): Add support for sjli instruction.
24294         (call_value_i): Likewise.
24295         * config/arc/constraints.md (Csc): New constraint.
24297 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24298             John Eric Martin  <John.Martin@emmicro-us.com>
24300         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24301         * config/arc/arc.c (_arc_jli_section): New struct.
24302         (arc_jli_section): New type.
24303         (rc_jli_sections): New static variable.
24304         (arc_handle_jli_attribute): New function.
24305         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24306         (arc_file_end): New function.
24307         (TARGET_ASM_FILE_END): Define.
24308         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24309         (arc_add_jli_section): New function.
24310         (jli_call_scan): Likewise.
24311         (arc_reorg): Call jli_call_scan.
24312         (arc_output_addsi): Remove 'S' from printing asm operand.
24313         (arc_is_jli_call_p): New function.
24314         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24315         operand.
24316         (movhi_insn): Likewise.
24317         (movsi_insn): Likewise.
24318         (movsi_set_cc_insn): Likewise.
24319         (loadqi_update): Likewise.
24320         (load_zeroextendqisi_update): Likewise.
24321         (load_signextendqisi_update): Likewise.
24322         (loadhi_update): Likewise.
24323         (load_zeroextendhisi_update): Likewise.
24324         (load_signextendhisi_update): Likewise.
24325         (loadsi_update): Likewise.
24326         (loadsf_update): Likewise.
24327         (movsicc_insn): Likewise.
24328         (bset_insn): Likewise.
24329         (bxor_insn): Likewise.
24330         (bclr_insn): Likewise.
24331         (bmsk_insn): Likewise.
24332         (bicsi3_insn): Likewise.
24333         (cmpsi_cc_c_insn): Likewise.
24334         (movsi_ne): Likewise.
24335         (movsi_cond_exec): Likewise.
24336         (clrsbsi2): Likewise.
24337         (norm_f): Likewise.
24338         (normw): Likewise.
24339         (swap): Likewise.
24340         (divaw): Likewise.
24341         (flag): Likewise.
24342         (sr): Likewise.
24343         (kflag): Likewise.
24344         (ffs): Likewise.
24345         (ffs_f): Likewise.
24346         (fls): Likewise.
24347         (call_i): Remove 'S' asm letter, add jli instruction.
24348         (call_value_i): Likewise.
24349         * config/arc/arc.op (mjli-always): New option.
24350         * config/arc/constraints.md (Cji): New constraint.
24351         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24352         operand.
24353         (subsf3_fpx): Likewise.
24354         (mulsf3_fpx): Likewise.
24355         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24356         asm operand.
24357         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24358         function attrbutes.
24359         * doc/invoke.texi (ARC): Document mjli-always option.
24361 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24363         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24364         avoid addition with 0 and use incw and decw where possible.
24366 2018-01-26  Richard Biener  <rguenther@suse.de>
24368         PR tree-optimization/81082
24369         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24370         association if it requires casting to unsigned.
24371         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24372         from fold_plusminus_mult_expr to catch important cases late when
24373         range info is available.
24375 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24377         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24378         * configure.ac (hidden_linkonce): New test.
24379         * configure: Regenerate.
24380         * config.in: Regenerate.
24382 2018-01-26  Julia Koval  <julia.koval@intel.com>
24384         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24385         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24386         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24387         _mm_mask_bitshuffle_epi64_mask): Fix type.
24388         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24389         USI_FTYPE_V4DI_V4DI_USI): Remove.
24390         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24391         __builtin_ia32_vpshufbitqmb256_mask,
24392         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24393         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24394         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24396 2018-01-26  Alan Modra  <amodra@gmail.com>
24398         PR target/84033
24399         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24400         UNSPEC_VBPERMQ.  Sort other unspecs.
24402 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
24404         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24406 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
24408         PR middle-end/83055
24409         * predict.c (drop_profile): Do not push/pop cfun; update also
24410         node->count.
24411         (handle_missing_profiles): Fix logic looking for zero profiles.
24413 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
24415         PR middle-end/83977
24416         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24417         on functions with #pragma omp declare simd or functions with simd
24418         attribute.
24419         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24420         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24421         Remove trailing \n from warning_at calls.
24423 2018-01-25  Tom de Vries  <tom@codesourcery.com>
24425         PR target/84028
24426         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24427         for neutered workers.
24429 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
24431         PR target/68467
24432         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24433         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24435 2018-01-24  Jeff Law  <law@redhat.com>
24437         PR target/83994
24438         * i386.c (get_probe_interval): Move to earlier point.
24439         (ix86_compute_frame_layout): If -fstack-clash-protection and
24440         the frame is larger than the probe interval, then use pushes
24441         to save registers rather than reg->mem moves.
24442         (ix86_expand_prologue): Remove conditional for int_registers_saved
24443         assertion.
24445 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
24447         PR target/84014
24448         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24449         min/max for never referenced object.
24451 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
24453         PR middle-end/83977
24454         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24455         here.
24456         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24457         attributes from DECL_ATTRIBUTES (decl) without affecting
24458         DECL_ATTRIBUTES (current_function_decl).
24459         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24460         functions with non-NULL DECL_ABSTRACT_ORIGIN.
24462 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
24464         PR tree-optimization/83979
24465         * fold-const.c (fold_comparison): Use constant_boolean_node
24466         instead of boolean_{true,false}_node.
24468 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
24470         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24471         with zero counts.
24473 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24475         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24476         Simplify the clause that sets the length attribute.
24477         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24478         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24479         clause that sets the length attribute.
24480         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24482 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24484         PR target/83589
24485         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24486         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24487         Add strict parameter.
24488         (prevent_branch_around_nothing): Insert dummy insn between branch to
24489         label and label with no ptx insn inbetween.
24490         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24492 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24494         PR target/81352
24495         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24496         for neutered threads in warp.
24497         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24499 2018-01-24  Richard Biener  <rguenther@suse.de>
24501         PR tree-optimization/83176
24502         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24503         operands.
24505 2018-01-24  Richard Biener  <rguenther@suse.de>
24507         PR tree-optimization/82819
24508         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
24509         code generating pluses that are no-ops in the target precision.
24511 2018-01-24  Richard Biener  <rguenther@suse.de>
24513         PR middle-end/84000
24514         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
24516 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24518         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
24519         to merge probabilities.
24520         * predict.c (probably_never_executed): Also mark as cold functions
24521         with global 0 profile and guessed local profile.
24522         * profile-count.c (profile_probability::combine_with_count): New
24523         member function.
24524         * profile-count.h (profile_probability::operator*,
24525         profile_probability::operator*=, profile_probability::operator/,
24526         profile_probability::operator/=): Reduce precision to adjusted
24527         and set value to guessed on contradictory divisions.
24528         (profile_probability::combine_with_freq): Remove.
24529         (profile_probability::combine_wiht_count): Declare.
24530         (profile_count::force_nonzero):: Set to adjusted.
24531         (profile_count::probability_in):: Set quality to adjusted.
24532         * tree-ssa-tail-merge.c (replace_block_by): Use
24533         combine_with_count.
24535 2018-01-23  Andrew Waterman  <andrew@sifive.com>
24536             Jim Wilson  <jimw@sifive.com>
24538         * config/riscv/riscv.c (riscv_stack_boundary): New.
24539         (riscv_option_override): Set riscv_stack_boundary.  Handle
24540         riscv_preferred_stack_boundary_arg.
24541         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
24542         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
24543         (STACK_BOUNDARY): Set to riscv_stack_boundary.
24544         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
24545         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
24546         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
24548 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
24550         PR target/83905
24551         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
24552         of struct ix86_frame.
24553         (ix86_expand_epilogue): Likewise.  Add a local variable for
24554         the reg_save_offset field in struct ix86_frame.
24556 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
24558         PR tree-optimization/82604
24559         * tree-loop-distribution.c (enum partition_kind): New enum item
24560         PKIND_PARTIAL_MEMSET.
24561         (partition_builtin_p): Support above new enum item.
24562         (generate_code_for_partition): Ditto.
24563         (compute_access_range): Differentiate cases that equality can be
24564         proven at all loops, the innermost loops or no loops.
24565         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
24566         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
24567         (finalize_partitions, distribute_loop): Don't fuse partition of
24568         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
24569         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
24570         parloop is enabled.
24572 2018-01-23  Martin Liska  <mliska@suse.cz>
24574         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
24575         order to ignore the predictor.
24576         (PRED_POLYMORPHIC_CALL): Likewise.
24577         (PRED_RECURSIVE_CALL): Likewise.
24579 2018-01-23  Martin Liska  <mliska@suse.cz>
24581         * tree-profile.c (tree_profiling): Print function header to
24582         aware reader which function we are working on.
24583         * value-prof.c (gimple_find_values_to_profile): Do not print
24584         not interesting value histograms.
24586 2018-01-23  Martin Liska  <mliska@suse.cz>
24588         * profile-count.h (enum profile_quality): Add
24589         profile_uninitialized as the first value. Do not number values
24590         as they are zero based.
24591         (profile_count::verify): Update sanity check.
24592         (profile_probability::verify): Likewise.
24594 2018-01-23  Nathan Sidwell  <nathan@acm.org>
24596         * doc/invoke.texi (ffor-scope): Deprecate.
24598 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24600         PR tree-optimization/83510
24601         * domwalk.c (set_all_edges_as_executable): New function.
24602         (dom_walker::dom_walker): Convert bool param
24603         "skip_unreachable_blocks" to enum reachability.  Move setup of
24604         edge flags to set_all_edges_as_executable and only do it when
24605         reachability is REACHABLE_BLOCKS.
24606         * domwalk.h (enum dom_walker::reachability): New enum.
24607         (dom_walker::dom_walker): Convert bool param
24608         "skip_unreachable_blocks" to enum reachability.
24609         (set_all_edges_as_executable): New decl.
24610         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
24611         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
24612         "reachability".
24613         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
24614         but converting true to REACHABLE_BLOCKS.
24615         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
24616         * tree-vrp.c
24617         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
24618         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
24619         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
24620         REACHABLE_BLOCKS.
24621         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
24622         if check_all_array_refs will be called.
24624 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24626         * tree.c (selftest::test_location_wrappers): Add more test
24627         coverage.
24629 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24631         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
24632         (selftest::test_bit_in_range): Likewise.
24634 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24636         PR testsuite/83888
24637         * doc/sourcebuild.texi (vect_float): Say that the selector
24638         only describes the situation when -funsafe-math-optimizations is on.
24639         (vect_float_strict): Document.
24641 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24643         PR tree-optimization/83965
24644         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
24645         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
24646         instead of checking only for a reduction.
24647         (vect_recog_widen_sum_pattern): Likewise.
24649 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24651         * predict.c (probably_never_executed): Only use precise profile info.
24652         (compute_function_frequency): Skip after inlining hack since we now
24653         have quality checking.
24655 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24657         * profile-count.h (profile_probability::very_unlikely,
24658         profile_probability::unlikely, profile_probability::even): Set
24659         precision to guessed.
24661 2018-01-23  Richard Biener  <rguenther@suse.de>
24663         PR tree-optimization/83963
24664         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
24665         Properly terminate dominator walk when crossing the exit edge not
24666         when visiting its source block.
24668 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
24670         PR c++/83918
24671         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
24672         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
24674 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
24676         PR tree-optimization/83957
24677         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
24678         semicolon after for body surrounded by braces.
24680         PR tree-optimization/83081
24681         * profile-count.h (profile_probability::split): New method.
24682         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
24683         Use profile_probability::split.
24684         (do_compare_rtx_and_jump): Fix adjustment of probabilities
24685         when splitting a single conditional jump into 2.
24687 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
24689         PR tree-optimization/69452
24690         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
24691         decl.
24693 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24695         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
24696         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
24697         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
24699 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24701         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
24702         declaration.
24703         * config/rl78/rl78.md (movdi): New define_expand.
24704         * config/rl78/rl78.c (rl78_split_movdi): New function.
24706 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
24708         PR target/83862
24709         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
24710         no longer used.
24711         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
24712         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
24713         128-bit to produce an UNSPEC move to get the double word with the
24714         signbit and then a shift directly to do signbit.
24715         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
24716         implementation with a new version that just does either a direct
24717         move or a regular move.  Move memory interface to separate insns.
24718         Move insns so they are next to the expander.
24719         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
24720         with signbit move.  Split big and little endian case.
24721         (signbit<mode>2_dm_mem_le): Likewise.
24722         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
24723         (signbit<mode>2_dm2): Likewise.
24725 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24727         * config/rl78/rl78.md (anddi3): New define_expand.
24729 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24731         * config/rl78/rl78.md (umindi3): New define_expand.
24733 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24735         * config/rl78/rl78.md (smindi3): New define_expand.
24737 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24739         * config/rl78/rl78.md (smaxdi3): New define_expand.
24741 2018-01-22  Carl Love  <cel@us.ibm.com>
24743         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
24744         LVX_V1TI): Add macro expansion.
24745         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
24746         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
24747         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
24748         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
24749         Change check to determine if the instruction is a byte reversing
24750         entry.  Fix typo in comment.
24751         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
24752         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
24753         Add def_builtin calls for new builtins.
24754         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
24755         Add define_insn expansion.
24757 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24759         * config/rl78/rl78.md (umaxdi3): New define_expand.
24761 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24763         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
24764         for non-QImode registers.
24766 2018-01-22  Richard Biener  <rguenther@suse.de>
24768         PR tree-optimization/83963
24769         * graphite-scop-detection.c (scop_detection::get_sese): Delay
24770         including the loop exit block.
24771         (scop_detection::merge_sese): Likewise.
24772         (scop_detection::add_scop): Do it here instead.
24774 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24776         * doc/sourcebuild.texi (arm_softfloat): Document.
24778 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
24780         PR gcc/77734
24781         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
24782         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
24783         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
24785 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24786             David Edelsohn  <dje.gcc@gmail.com>
24788         PR target/83946
24789         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24790         Change "crset eq" to "crset 2".
24791         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24792         (*call_indirect_aix<mode>_nospec): Likewise.
24793         (*call_value_indirect_aix<mode>_nospec): Likewise.
24794         (*call_indirect_elfv2<mode>_nospec): Likewise.
24795         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24796         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24797         change assembly output from . to $.
24798         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24799         (indirect_jump<mode>_nospec): Change assembly output from . to $.
24800         (*tablejump<mode>_internal1_nospec): Likewise.
24802 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
24804         PR target/80870
24805         * config/sh/sh_optimize_sett_clrt.cc:
24806         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24808 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
24810         PR tree-optimization/83940
24811         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24812         offset_dt to vect_constant_def rather than vect_unknown_def_type.
24813         (vect_check_load_store_mask): Add a mask_dt_out parameter and
24814         use it to pass back the definition type.
24815         (vect_check_store_rhs): Likewise rhs_dt_out.
24816         (vect_build_gather_load_calls): Add a mask_dt argument and use
24817         it instead of a call to vect_is_simple_use.
24818         (vectorizable_store): Update calls to vect_check_load_store_mask
24819         and vect_check_store_rhs.  Use the dt returned by the latter instead
24820         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
24821         instead of calls to vect_is_simple_use.  Pass the scalar rather
24822         than the vector operand to vect_is_simple_use when handling
24823         second and subsequent copies of an rhs value.
24824         (vectorizable_load): Update calls to vect_check_load_store_mask
24825         and vect_build_gather_load_calls.  Use the cached mask_dt and
24826         gs_info.offset_dt instead of calls to vect_is_simple_use.
24828 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
24830         PR middle-end/83945
24831         * tree-emutls.c: Include gimplify.h.
24832         (lower_emutls_2): New function.
24833         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24834         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24835         it before further processing.
24837         PR target/83930
24838         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24839         UINTVAL (trueop1) instead of INTVAL (op1).
24841 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
24843         PR debug/81570
24844         PR debug/83728
24845         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24846         INCOMING_FRAME_SP_OFFSET if not defined.
24847         (scan_trace): Add ENTRY argument.  If true and
24848         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24849         emit a note to adjust the CFA offset.
24850         (create_cfi_notes): Adjust scan_trace callers.
24851         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24852         INCOMING_FRAME_SP_OFFSET in the CIE.
24853         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24854         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24855         Likewise.
24856         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24857         * doc/tm.texi: Regenerated.
24859 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24861         PR rtl-optimization/83147
24862         * lra-constraints.c (remove_inheritance_pseudos): Use
24863         lra_substitute_pseudo_within_insn.
24865 2018-01-19  Tom de Vries  <tom@codesourcery.com>
24866             Cesar Philippidis  <cesar@codesourcery.com>
24868         PR target/83920
24869         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24871 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
24873         PR target/83790
24874         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24875         spaces for function labels.
24877 2018-01-19  Martin Liska  <mliska@suse.cz>
24879         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24880         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24881         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24882         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24883         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24884         (PRED_CONST_RETURN): Change from 69 to 65.
24885         (PRED_NULL_RETURN): Change from 91 to 71.
24886         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24887         (PRED_LOOP_GUARD): Change from 66 to 73.
24889 2018-01-19  Martin Liska  <mliska@suse.cz>
24891         * predict.c (predict_insn_def): Add new assert.
24892         (struct branch_predictor): Change type to signed integer.
24893         (test_prediction_value_range): Amend test to cover
24894         PROB_UNINITIALIZED.
24895         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24896         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24897         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24898         (PRED_LOOP_IV_COMPARE): Likewise.
24899         * predict.h (PROB_UNINITIALIZED): Define new constant.
24901 2018-01-19  Martin Liska  <mliska@suse.cz>
24903         * predict.c (dump_prediction): Add new format for
24904         analyze_brprob.py script which is enabled with -details
24905         suboption.
24906         * profile-count.h (precise_p): New function.
24908 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24910         PR tree-optimization/83922
24911         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24912         there are no statements that need masking.
24913         (vect_active_double_reduction_p): New function.
24914         (vect_analyze_loop_operations): Use it when handling phis that
24915         are not in the loop header.
24917 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24919         PR tree-optimization/83914
24920         * tree-vect-loop.c (vectorizable_induction): Don't convert
24921         init_expr or apply the peeling adjustment for inductions
24922         that are nested within the vectorized loop.
24924 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24926         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24927         instead of NEG.
24929 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24931         PR sanitizer/81715
24932         PR testsuite/83882
24933         * function.h (gimplify_parameters): Add gimple_seq * argument.
24934         * function.c: Include gimple.h and options.h.
24935         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24936         for the added local temporaries if needed.
24937         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24938         if there are any parameter cleanups, wrap whole body into a
24939         try/finally with the cleanups.
24941 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
24943         PR target/82964
24944         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24945         Use GET_MODE_CLASS for scalar floating point.
24947 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24949         PR ipa/82256
24950         patch by PaX Team
24951         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24952         Fix call of call_cgraph_insertion_hooks.
24954 2018-01-18  Martin Sebor  <msebor@redhat.com>
24956         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24958 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24960         PR ipa/83619
24961         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24962         frequencies.
24964 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
24966         PR other/70268
24967         * common.opt: (-ffile-prefix-map): New option.
24968         * opts.c (common_handle_option): Defer it.
24969         * opts-global.c (handle_common_deferred_options): Handle it.
24970         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24971         * file-prefix-map.h: New file.
24972         (remap_debug_filename, add_debug_prefix_map): ...here.
24973         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24974         * final.c (debug_prefix_map, add_debug_prefix_map
24975         remap_debug_filename): Move to...
24976         * file-prefix-map.c: New file.
24977         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24978         generalize, get rid of alloca(), use strrchr() instead of strchr().
24979         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24980         Implement in terms of add_prefix_map().
24981         (remap_macro_filename, remap_debug_filename): Implement in term of
24982         remap_filename().
24983         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24984         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24985         * dbxout.c: Include file-prefix-map.h.
24986         * varasm.c: Likewise.
24987         * vmsdbgout.c: Likewise.
24988         * xcoffout.c: Likewise.
24989         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24990         * doc/cppopts.texi (-fmacro-prefix-map): Document.
24991         * doc/invoke.texi (-ffile-prefix-map): Document.
24992         (-fdebug-prefix-map): Update description.
24994 2018-01-18  Martin Liska  <mliska@suse.cz>
24996         * config/i386/i386.c (indirect_thunk_name): Document that also
24997         lfence is emitted.
24998         (output_indirect_thunk): Document why both instructions
24999         (pause and lfence) are generated.
25001 2018-01-18  Richard Biener  <rguenther@suse.de>
25003         PR tree-optimization/83887
25004         * graphite-scop-detection.c
25005         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25006         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25007         (scop_detection::merge_sese): Re-implement with a flood-fill
25008         algorithm that properly finds a SESE region if it exists.
25010 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25012         PR c/61240
25013         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25014         pointer_diff optimizations use view_convert instead of convert.
25016 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25018         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25019         Generate different code for -mno-speculate-indirect-jumps.
25020         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25021         (*call_indirect_aix<mode>): Disable for
25022         -mno-speculate-indirect-jumps.
25023         (*call_indirect_aix<mode>_nospec): New define_insn.
25024         (*call_value_indirect_aix<mode>): Disable for
25025         -mno-speculate-indirect-jumps.
25026         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25027         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25028         -mno-speculate-indirect-jumps.
25029         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25031 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25033         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25034         long double type, set the flags for noting the default long double
25035         type, even if we don't pass or return a long double type.
25037 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25039         PR ipa/83051
25040         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25041         failure.
25043 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25045         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25046         support for merge[hl].
25047         (fold_mergehl_helper): New helper function.
25048         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25049         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25050         (altivec_vmrglw_direct): Add xxmrglw insn.
25052 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25054         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25055         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25057 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25059         PR lto/83121
25060         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25061         call the lto_location_cache before reading the
25062         DECL_SOURCE_LOCATION of the types.
25064 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25065             Richard Sandiford  <richard.sandiford@linaro.org>
25067         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25068         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25069         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25070         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25071         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25072         Add declaration.
25073         * config/aarch64/constraints.md (aarch64_movti_operand):
25074         Limit immediates.
25075         * config/aarch64/predicates.md (Uti): Add new constraint.
25077 2018-01-17  Carl Love  <cel@us.ibm.com>
25079         * config/rs6000/vsx.md (define_expand xl_len_r,
25080         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25081         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25082         lxvll.
25083         (define_expand, define_insn): Move the shift left from  the
25084         define_insn to the define_expand for lxvl and stxvl instructions.
25085         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25086         and XL_LEN_R definitions to PURE.
25088 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25090         * config/i386/i386.c (indirect_thunk_name): Declare regno
25091         as unsigned int.  Compare regno with INVALID_REGNUM.
25092         (output_indirect_thunk): Ditto.
25093         (output_indirect_thunk_function): Ditto.
25094         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25095         in the call to output_indirect_thunk_function.
25097 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25099         PR middle-end/83884
25100         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25101         rather than the size of inner_type to determine the stack slot size
25102         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25104 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25106         PR target/83546
25107         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25108         to PTA_SILVERMONT.
25110 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25112         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25113         endian Linux systems to optionally enable multilibs for selecting
25114         the long double type if the user configured an explicit type.
25115         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25116         have no long double multilibs if not defined.
25117         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25118         warn if the user used -mabi={ieee,ibm}longdouble and we built
25119         multilibs for long double.
25120         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25121         appropriate multilib option.
25122         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25123         multilib options.
25124         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25125         for building long double multilibs.
25126         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25128 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25130         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25131         copies.
25133         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25134         64 bits.
25135         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25136         128 bits.
25138         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25139         variables.
25141         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25142         return value.
25144 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25146         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25147         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25149 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25151         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25152         different rtl trees depending on TARGET_64BIT.
25153         (rs6000_gen_lvx): Likewise.
25155 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25157         * config/visium/visium.md (nop): Tweak comment.
25158         (hazard_nop): Likewise.
25160 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25162         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25163         -mspeculate-indirect-jumps.
25164         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25165         for -mno-speculate-indirect-jumps.
25166         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25167         (*call_value_indirect_elfv2<mode>): Disable for
25168         -mno-speculate-indirect-jumps.
25169         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25170         (indirect_jump): Emit different RTL for
25171         -mno-speculate-indirect-jumps.
25172         (*indirect_jump<mode>): Disable for
25173         -mno-speculate-indirect-jumps.
25174         (*indirect_jump<mode>_nospec): New define_insn.
25175         (tablejump): Emit different RTL for
25176         -mno-speculate-indirect-jumps.
25177         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25178         (tablejumpsi_nospec): New define_expand.
25179         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25180         (tablejumpdi_nospec): New define_expand.
25181         (*tablejump<mode>_internal1): Disable for
25182         -mno-speculate-indirect-jumps.
25183         (*tablejump<mode>_internal1_nospec): New define_insn.
25184         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25185         option.
25187 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25189         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25190         callers updated.
25192 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25193             Richard Biener  <rguenth@suse.de>
25195         PR libgomp/83590
25196         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25197         return early, inline manually is_gimple_sizepos.  Make sure if we
25198         call gimplify_expr we don't end up with a gimple constant.
25199         * tree.c (variably_modified_type_p): Don't return true for
25200         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25201         * gimplify.h (is_gimple_sizepos): Remove.
25203 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25205         PR tree-optimization/83857
25206         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25207         vectorizable_live_operation for pure SLP statements.
25208         (vectorizable_live_operation): Handle PHIs.
25210 2018-01-16  Richard Biener  <rguenther@suse.de>
25212         PR tree-optimization/83867
25213         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25214         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25216 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25218         PR c/83844
25219         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25220         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25221         If off is not INTEGER_CST, issue a may not be aligned warning
25222         rather than isn't aligned.  Use isn%'t rather than isn't.
25223         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25224         into MULT_EXPR.
25225         <case MULT_EXPR>: Improve the case when bottom and one of the
25226         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25227         operand, in that case check if the other operand is multiple of
25228         bottom divided by the INTEGER_CST operand.
25230 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25232         PR target/83858
25233         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25234         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25235         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25236         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25237         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25238         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25239         (pa_function_arg_size): New function.
25241 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25243         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25244         in a separate statement.
25246 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25248         PR tree-optimization/83847
25249         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25250         group gathers and scatters.
25252 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25254         PR rtl-optimization/86620
25255         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25257         PR rtl-optimization/83213
25258         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25259         to last if both are JUMP_INSNs.
25261         PR tree-optimization/83843
25262         * gimple-ssa-store-merging.c
25263         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25264         store_immediate_info for bswap/nop orig_stores.
25266 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25268         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25269         !TARGET_MUL.
25270         <UDIV>: Increase cost if !TARGET_DIV.
25272 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25274         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25275         (define_attr "cr_logical_3op"): New.
25276         (cceq_ior_compare): Adjust.
25277         (cceq_ior_compare_complement): Adjust.
25278         (*cceq_rev_compare): Adjust.
25279         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25280         (is_cracked_insn): Adjust.
25281         (insn_must_be_first_in_group): Adjust.
25282         * config/rs6000/40x.md: Adjust.
25283         * config/rs6000/440.md: Adjust.
25284         * config/rs6000/476.md: Adjust.
25285         * config/rs6000/601.md: Adjust.
25286         * config/rs6000/603.md: Adjust.
25287         * config/rs6000/6xx.md: Adjust.
25288         * config/rs6000/7450.md: Adjust.
25289         * config/rs6000/7xx.md: Adjust.
25290         * config/rs6000/8540.md: Adjust.
25291         * config/rs6000/cell.md: Adjust.
25292         * config/rs6000/e300c2c3.md: Adjust.
25293         * config/rs6000/e500mc.md: Adjust.
25294         * config/rs6000/e500mc64.md: Adjust.
25295         * config/rs6000/e5500.md: Adjust.
25296         * config/rs6000/e6500.md: Adjust.
25297         * config/rs6000/mpc.md: Adjust.
25298         * config/rs6000/power4.md: Adjust.
25299         * config/rs6000/power5.md: Adjust.
25300         * config/rs6000/power6.md: Adjust.
25301         * config/rs6000/power7.md: Adjust.
25302         * config/rs6000/power8.md: Adjust.
25303         * config/rs6000/power9.md: Adjust.
25304         * config/rs6000/rs64.md: Adjust.
25305         * config/rs6000/titan.md: Adjust.
25307 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25309         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25310         ix86_indirect_branch_register logic.
25312 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25314         * config/i386/constraints.md (Bs): Update
25315         ix86_indirect_branch_register check.  Don't check
25316         ix86_indirect_branch_register with GOT_memory_operand.
25317         (Bw): Likewise.
25318         * config/i386/predicates.md (GOT_memory_operand): Don't check
25319         ix86_indirect_branch_register here.
25320         (GOT32_symbol_operand): Likewise.
25322 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25324         * config/i386/predicates.md (constant_call_address_operand):
25325         Rewrite ix86_indirect_branch_register logic.
25326         (sibcall_insn_operand): Likewise.
25328 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25330         * config/i386/constraints.md (Bs): Replace
25331         ix86_indirect_branch_thunk_register with
25332         ix86_indirect_branch_register.
25333         (Bw): Likewise.
25334         * config/i386/i386.md (indirect_jump): Likewise.
25335         (tablejump): Likewise.
25336         (*sibcall_memory): Likewise.
25337         (*sibcall_value_memory): Likewise.
25338         Peepholes of indirect call and jump via memory: Likewise.
25339         * config/i386/i386.opt: Likewise.
25340         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25341         (GOT_memory_operand): Likewise.
25342         (call_insn_operand): Likewise.
25343         (sibcall_insn_operand): Likewise.
25344         (GOT32_symbol_operand): Likewise.
25346 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25348         PR middle-end/83837
25349         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25350         type rather than type addr's type points to.
25351         (expand_omp_atomic_mutex): Likewise.
25352         (expand_omp_atomic): Likewise.
25354 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25356         PR target/83839
25357         * config/i386/i386.c (output_indirect_thunk_function): Use
25358         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25359         for  __x86_return_thunk.
25361 2018-01-15  Richard Biener  <rguenther@suse.de>
25363         PR middle-end/83850
25364         * expmed.c (extract_bit_field_1): Fix typo.
25366 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25368         PR target/83687
25369         * config/arm/iterators.md (VF): New mode iterator.
25370         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25371         Remove integer-related logic from pattern.
25372         (neon_vabd<mode>_3): Likewise.
25374 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25376         PR middle-end/82694
25377         * common.opt (fstrict-overflow): No longer an alias.
25378         (fwrapv-pointer): New option.
25379         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25380         also for pointer types based on flag_wrapv_pointer.
25381         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25382         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25383         opts->x_flag_wrapv got set.
25384         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25385         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25386         POINTER_TYPE_OVERFLOW_UNDEFINED.
25387         * match.pd: Likewise in address comparison pattern.
25388         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25390 2018-01-15  Richard Biener  <rguenther@suse.de>
25392         PR lto/83804
25393         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25394         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
25395         Reset type names to their identifier if their TYPE_DECL doesn't
25396         have linkage (and thus is used for ODR and devirt).
25397         (save_debug_info_for_decl): Remove.
25398         (save_debug_info_for_type): Likewise.
25399         (add_tree_to_fld_list): Adjust.
25400         * tree-pretty-print.c (dump_generic_node): Make dumping of
25401         type names more robust.
25403 2018-01-15  Richard Biener  <rguenther@suse.de>
25405         * BASE-VER: Bump to 8.0.1.
25407 2018-01-14  Martin Sebor  <msebor@redhat.com>
25409         PR other/83508
25410         * builtins.c (check_access): Avoid warning when the no-warning bit
25411         is set.
25413 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
25415         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25416         * ira-color (allocno_hard_regs_compare): Likewise.
25418 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
25420         PR target/83013
25421         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25422         Use .pushsection/.popsection.
25424 2018-01-14  Martin Sebor  <msebor@redhat.com>
25426         PR c++/81327
25427         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25429 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
25431         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25432         entry from extra_headers.
25433         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25434         extra_headers, make the list bitwise identical to the i?86-*-* one.
25436 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25438         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25439         -mcmodel=large with -mindirect-branch=thunk,
25440         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25441         -mfunction-return=thunk-extern.
25442         * doc/invoke.texi: Document -mcmodel=large is incompatible with
25443         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25444         -mfunction-return=thunk and -mfunction-return=thunk-extern.
25446 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25448         * config/i386/i386.c (print_reg): Print the name of the full
25449         integer register without '%'.
25450         (ix86_print_operand): Handle 'V'.
25451         * doc/extend.texi: Document 'V' modifier.
25453 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25455         * config/i386/constraints.md (Bs): Disallow memory operand for
25456         -mindirect-branch-register.
25457         (Bw): Likewise.
25458         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25459         (GOT_memory_operand): Likewise.
25460         (call_insn_operand): Likewise.
25461         (sibcall_insn_operand): Likewise.
25462         (GOT32_symbol_operand): Likewise.
25463         * config/i386/i386.md (indirect_jump): Call convert_memory_address
25464         for -mindirect-branch-register.
25465         (tablejump): Likewise.
25466         (*sibcall_memory): Likewise.
25467         (*sibcall_value_memory): Likewise.
25468         Disallow peepholes of indirect call and jump via memory for
25469         -mindirect-branch-register.
25470         (*call_pop): Replace m with Bw.
25471         (*call_value_pop): Likewise.
25472         (*sibcall_pop_memory): Replace m with Bs.
25473         * config/i386/i386.opt (mindirect-branch-register): New option.
25474         * doc/invoke.texi: Document -mindirect-branch-register option.
25476 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25478         * config/i386/i386-protos.h (ix86_output_function_return): New.
25479         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25480         set function_return_type.
25481         (indirect_thunk_name): Add ret_p to indicate thunk for function
25482         return.
25483         (output_indirect_thunk_function): Pass false to
25484         indirect_thunk_name.
25485         (ix86_output_indirect_branch_via_reg): Likewise.
25486         (ix86_output_indirect_branch_via_push): Likewise.
25487         (output_indirect_thunk_function): Create alias for function
25488         return thunk if regno < 0.
25489         (ix86_output_function_return): New function.
25490         (ix86_handle_fndecl_attribute): Handle function_return.
25491         (ix86_attribute_table): Add function_return.
25492         * config/i386/i386.h (machine_function): Add
25493         function_return_type.
25494         * config/i386/i386.md (simple_return_internal): Use
25495         ix86_output_function_return.
25496         (simple_return_internal_long): Likewise.
25497         * config/i386/i386.opt (mfunction-return=): New option.
25498         (indirect_branch): Mention -mfunction-return=.
25499         * doc/extend.texi: Document function_return function attribute.
25500         * doc/invoke.texi: Document -mfunction-return= option.
25502 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25504         * config/i386/i386-opts.h (indirect_branch): New.
25505         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25506         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
25507         with local indirect jump when converting indirect call and jump.
25508         (ix86_set_indirect_branch_type): New.
25509         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
25510         (indirectlabelno): New.
25511         (indirect_thunk_needed): Likewise.
25512         (indirect_thunk_bnd_needed): Likewise.
25513         (indirect_thunks_used): Likewise.
25514         (indirect_thunks_bnd_used): Likewise.
25515         (INDIRECT_LABEL): Likewise.
25516         (indirect_thunk_name): Likewise.
25517         (output_indirect_thunk): Likewise.
25518         (output_indirect_thunk_function): Likewise.
25519         (ix86_output_indirect_branch_via_reg): Likewise.
25520         (ix86_output_indirect_branch_via_push): Likewise.
25521         (ix86_output_indirect_branch): Likewise.
25522         (ix86_output_indirect_jmp): Likewise.
25523         (ix86_code_end): Call output_indirect_thunk_function if needed.
25524         (ix86_output_call_insn): Call ix86_output_indirect_branch if
25525         needed.
25526         (ix86_handle_fndecl_attribute): Handle indirect_branch.
25527         (ix86_attribute_table): Add indirect_branch.
25528         * config/i386/i386.h (machine_function): Add indirect_branch_type
25529         and has_local_indirect_jump.
25530         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
25531         to true.
25532         (tablejump): Likewise.
25533         (*indirect_jump): Use ix86_output_indirect_jmp.
25534         (*tablejump_1): Likewise.
25535         (simple_return_indirect_internal): Likewise.
25536         * config/i386/i386.opt (mindirect-branch=): New option.
25537         (indirect_branch): New.
25538         (keep): Likewise.
25539         (thunk): Likewise.
25540         (thunk-inline): Likewise.
25541         (thunk-extern): Likewise.
25542         * doc/extend.texi: Document indirect_branch function attribute.
25543         * doc/invoke.texi: Document -mindirect-branch= option.
25545 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
25547         PR ipa/83051
25548         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
25550 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
25552         * ipa-inline.c (want_inline_small_function_p): Return false if
25553         inlining has already failed with CIF_FINAL_ERROR.
25554         (update_caller_keys): Call want_inline_small_function_p before
25555         can_inline_edge_p.
25556         (update_callee_keys): Likewise.
25558 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25560         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
25561         New function.
25562         (rs6000_quadword_masked_address_p): Likewise.
25563         (quad_aligned_load_p): Likewise.
25564         (quad_aligned_store_p): Likewise.
25565         (const_load_sequence_p): Add comment to describe the outer-most loop.
25566         (mimic_memory_attributes_and_flags): New function.
25567         (rs6000_gen_stvx): Likewise.
25568         (replace_swapped_aligned_store): Likewise.
25569         (rs6000_gen_lvx): Likewise.
25570         (replace_swapped_aligned_load): Likewise.
25571         (replace_swapped_load_constant): Capitalize argument name in
25572         comment describing this function.
25573         (rs6000_analyze_swaps): Add a third pass to search for vector loads
25574         and stores that access quad-word aligned addresses and replace
25575         with stvx or lvx instructions when appropriate.
25576         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
25577         New function prototype.
25578         (rs6000_quadword_masked_address_p): Likewise.
25579         (rs6000_gen_lvx): Likewise.
25580         (rs6000_gen_stvx): Likewise.
25581         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
25582         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
25583         when memory address is aligned.
25584         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
25585         this split to select lvx instruction when memory address is aligned.
25586         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
25587         instruction when memory address is aligned.
25588         (*vsx_le_perm_load_v16qi): Likewise.
25589         (four unnamed splitters): Modify to select the stvx instruction
25590         when memory is aligned.
25592 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
25594         * predict.c (determine_unlikely_bbs): Handle correctly BBs
25595         which appears in the queue multiple times.
25597 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25598             Alan Hayward  <alan.hayward@arm.com>
25599             David Sherwood  <david.sherwood@arm.com>
25601         * tree-vectorizer.h (vec_lower_bound): New structure.
25602         (_loop_vec_info): Add check_nonzero and lower_bounds.
25603         (LOOP_VINFO_CHECK_NONZERO): New macro.
25604         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
25605         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
25606         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
25607         fields.  Make seg_len the distance travelled, not including the
25608         access size.
25609         (dr_direction_indicator): Declare.
25610         (dr_zero_step_indicator): Likewise.
25611         (dr_known_forward_stride_p): Likewise.
25612         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
25613         tree-ssanames.h.
25614         (runtime_alias_check_p): Allow runtime alias checks with
25615         variable strides.
25616         (operator ==): Compare access_size and align.
25617         (prune_runtime_alias_test_list): Rework for new distinction between
25618         the access_size and seg_len.
25619         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
25620         segment lengths.
25621         (get_segment_min_max): New function.
25622         (create_intersect_range_checks): Use it.
25623         (dr_step_indicator): New function.
25624         (dr_direction_indicator): Likewise.
25625         (dr_zero_step_indicator): Likewise.
25626         (dr_known_forward_stride_p): Likewise.
25627         * tree-loop-distribution.c (data_ref_segment_size): Return
25628         DR_STEP * (niters - 1).
25629         (compute_alias_check_pairs): Update call to the dr_with_seg_len
25630         constructor.
25631         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
25632         (vect_preserves_scalar_order_p): New function, split out from...
25633         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
25634         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
25635         (vect_vfa_access_size): New function.
25636         (vect_vfa_align): Likewise.
25637         (vect_compile_time_alias): Take access_size_a and access_b arguments.
25638         (dump_lower_bound): New function.
25639         (vect_check_lower_bound): Likewise.
25640         (vect_small_gap_p): Likewise.
25641         (vectorizable_with_step_bound_p): Likewise.
25642         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
25643         depencies if the vectorization factor is 1.  Convert the checks
25644         for nonzero steps into checks on the bounds of DR_STEP.  Try using
25645         a bunds check for variable steps if the minimum required step is
25646         relatively small. Update calls to the dr_with_seg_len
25647         constructor and to vect_compile_time_alias.
25648         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
25649         function.
25650         (vect_loop_versioning): Call it.
25651         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
25652         when retrying.
25653         (vect_estimate_min_profitable_iters): Account for any bounds checks.
25655 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25656             Alan Hayward  <alan.hayward@arm.com>
25657             David Sherwood  <david.sherwood@arm.com>
25659         * doc/sourcebuild.texi (vect_scatter_store): Document.
25660         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
25661         optabs.
25662         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
25663         Document.
25664         * genopinit.c (main): Add supports_vec_scatter_store and
25665         supports_vec_scatter_store_cached to target_optabs.
25666         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
25667         IFN_MASK_SCATTER_STORE.
25668         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
25669         functions.
25670         * internal-fn.h (internal_store_fn_p): Declare.
25671         (internal_fn_stored_value_index): Likewise.
25672         * internal-fn.c (scatter_store_direct): New macro.
25673         (expand_scatter_store_optab_fn): New function.
25674         (direct_scatter_store_optab_supported_p): New macro.
25675         (internal_store_fn_p): New function.
25676         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
25677         IFN_MASK_SCATTER_STORE.
25678         (internal_fn_mask_index): Likewise.
25679         (internal_fn_stored_value_index): New function.
25680         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
25681         for scatter stores.
25682         * optabs-query.h (supports_vec_scatter_store_p): Declare.
25683         * optabs-query.c (supports_vec_scatter_store_p): New function.
25684         * tree-vectorizer.h (vect_get_store_rhs): Declare.
25685         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
25686         true for scatter stores.
25687         (vect_gather_scatter_fn_p): Handle scatter stores too.
25688         (vect_check_gather_scatter): Consider using scatter stores if
25689         supports_vec_scatter_store_p.
25690         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
25691         scatter stores too.
25692         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25693         internal_fn_stored_value_index.
25694         (check_load_store_masking): Handle scatter stores too.
25695         (vect_get_store_rhs): Make public.
25696         (vectorizable_call): Use internal_store_fn_p.
25697         (vectorizable_store): Handle scatter store internal functions.
25698         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
25699         when deciding whether the end of the group has been reached.
25700         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
25701         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
25702         (mask_scatter_store<mode>): New insns.
25704 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25705             Alan Hayward  <alan.hayward@arm.com>
25706             David Sherwood  <david.sherwood@arm.com>
25708         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
25709         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
25710         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
25711         function.
25712         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
25713         Use vect_truncate_gather_scatter_offset if we can't treat the
25714         operation as a normal gather load or scatter store.
25715         (get_group_load_store_type): Take the gather_scatter_info
25716         as argument.  Try using a gather load or scatter store for
25717         single-element groups.
25718         (get_load_store_type): Update calls to get_group_load_store_type
25719         and vect_use_strided_gather_scatters_p.
25721 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25722             Alan Hayward  <alan.hayward@arm.com>
25723             David Sherwood  <david.sherwood@arm.com>
25725         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
25726         optional tree argument.
25727         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
25728         null target hooks.
25729         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
25730         but continue to use the current value as a fallback.
25731         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
25732         to compare the updates.
25733         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
25734         (get_load_store_type): Use it when handling a strided access.
25735         (vect_get_strided_load_store_ops): New function.
25736         (vect_get_data_ptr_increment): Likewise.
25737         (vectorizable_load): Handle strided gather loads.  Always pass
25738         a step to vect_create_data_ref_ptr and bump_vector_ptr.
25740 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25741             Alan Hayward  <alan.hayward@arm.com>
25742             David Sherwood  <david.sherwood@arm.com>
25744         * doc/md.texi (gather_load@var{m}): Document.
25745         (mask_gather_load@var{m}): Likewise.
25746         * genopinit.c (main): Add supports_vec_gather_load and
25747         supports_vec_gather_load_cached to target_optabs.
25748         * optabs-tree.c (init_tree_optimization_optabs): Use
25749         ggc_cleared_alloc to allocate target_optabs.
25750         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
25751         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
25752         functions.
25753         * internal-fn.h (internal_load_fn_p): Declare.
25754         (internal_gather_scatter_fn_p): Likewise.
25755         (internal_fn_mask_index): Likewise.
25756         (internal_gather_scatter_fn_supported_p): Likewise.
25757         * internal-fn.c (gather_load_direct): New macro.
25758         (expand_gather_load_optab_fn): New function.
25759         (direct_gather_load_optab_supported_p): New macro.
25760         (direct_internal_fn_optab): New function.
25761         (internal_load_fn_p): Likewise.
25762         (internal_gather_scatter_fn_p): Likewise.
25763         (internal_fn_mask_index): Likewise.
25764         (internal_gather_scatter_fn_supported_p): Likewise.
25765         * optabs-query.c (supports_at_least_one_mode_p): New function.
25766         (supports_vec_gather_load_p): Likewise.
25767         * optabs-query.h (supports_vec_gather_load_p): Declare.
25768         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
25769         and memory_type field.
25770         (NUM_PATTERNS): Bump to 15.
25771         * tree-vect-data-refs.c: Include internal-fn.h.
25772         (vect_gather_scatter_fn_p): New function.
25773         (vect_describe_gather_scatter_call): Likewise.
25774         (vect_check_gather_scatter): Try using internal functions for
25775         gather loads.  Recognize existing calls to a gather load function.
25776         (vect_analyze_data_refs): Consider using gather loads if
25777         supports_vec_gather_load_p.
25778         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
25779         (vect_get_gather_scatter_offset_type): Likewise.
25780         (vect_convert_mask_for_vectype): Likewise.
25781         (vect_add_conversion_to_patterm): Likewise.
25782         (vect_try_gather_scatter_pattern): Likewise.
25783         (vect_recog_gather_scatter_pattern): New pattern recognizer.
25784         (vect_vect_recog_func_ptrs): Add it.
25785         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25786         internal_fn_mask_index and internal_gather_scatter_fn_p.
25787         (check_load_store_masking): Take the gather_scatter_info as an
25788         argument and handle gather loads.
25789         (vect_get_gather_scatter_ops): New function.
25790         (vectorizable_call): Check internal_load_fn_p.
25791         (vectorizable_load): Likewise.  Handle gather load internal
25792         functions.
25793         (vectorizable_store): Update call to check_load_store_masking.
25794         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25795         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25796         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25797         (aarch64_gather_scale_operand_d): New predicates.
25798         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25799         (mask_gather_load<mode>): New insns.
25801 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25802             Alan Hayward  <alan.hayward@arm.com>
25803             David Sherwood  <david.sherwood@arm.com>
25805         * optabs.def (fold_left_plus_optab): New optab.
25806         * doc/md.texi (fold_left_plus_@var{m}): Document.
25807         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25808         * internal-fn.c (fold_left_direct): Define.
25809         (expand_fold_left_optab_fn): Likewise.
25810         (direct_fold_left_optab_supported_p): Likewise.
25811         * fold-const-call.c (fold_const_fold_left): New function.
25812         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25813         * tree-parloops.c (valid_reduction_p): New function.
25814         (gather_scalar_reductions): Use it.
25815         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25816         (vect_finish_replace_stmt): Declare.
25817         * tree-vect-loop.c (fold_left_reduction_fn): New function.
25818         (needs_fold_left_reduction_p): New function, split out from...
25819         (vect_is_simple_reduction): ...here.  Accept reductions that
25820         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25821         (vect_force_simple_reduction): Also store the reduction type in
25822         the assignment's STMT_VINFO_REDUC_TYPE.
25823         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25824         (merge_with_identity): New function.
25825         (vect_expand_fold_left): Likewise.
25826         (vectorize_fold_left_reduction): Likewise.
25827         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
25828         scalar phi in place for it.  Check for target support and reject
25829         cases that would reassociate the operation.  Defer the transform
25830         phase to vectorize_fold_left_reduction.
25831         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25832         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25833         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25835 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25837         * tree-if-conv.c (predicate_mem_writes): Remove redundant
25838         call to ifc_temp_var.
25840 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25841             Alan Hayward  <alan.hayward@arm.com>
25842             David Sherwood  <david.sherwood@arm.com>
25844         * target.def (legitimize_address_displacement): Take the original
25845         offset as a poly_int.
25846         * targhooks.h (default_legitimize_address_displacement): Update
25847         accordingly.
25848         * targhooks.c (default_legitimize_address_displacement): Likewise.
25849         * doc/tm.texi: Regenerate.
25850         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25851         as an argument, moving assert of ad->disp == ad->disp_term to...
25852         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
25853         Try calling targetm.legitimize_address_displacement before expanding
25854         the address rather than afterwards, and adjust for the new interface.
25855         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25856         Match the new hook interface.  Handle SVE addresses.
25857         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25858         new hook interface.
25860 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25862         * Makefile.in (OBJS): Add early-remat.o.
25863         * target.def (select_early_remat_modes): New hook.
25864         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25865         * doc/tm.texi: Regenerate.
25866         * targhooks.h (default_select_early_remat_modes): Declare.
25867         * targhooks.c (default_select_early_remat_modes): New function.
25868         * timevar.def (TV_EARLY_REMAT): New timevar.
25869         * passes.def (pass_early_remat): New pass.
25870         * tree-pass.h (make_pass_early_remat): Declare.
25871         * early-remat.c: New file.
25872         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25873         function.
25874         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25876 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25877             Alan Hayward  <alan.hayward@arm.com>
25878             David Sherwood  <david.sherwood@arm.com>
25880         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25881         vfm1 with a bound_epilog parameter.
25882         (vect_do_peeling): Update calls accordingly, and move the prologue
25883         call earlier in the function.  Treat the base bound_epilog as 0 for
25884         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
25885         this base when peeling for gaps.
25886         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25887         with fully-masked loops.
25888         (vect_estimate_min_profitable_iters): Handle the single peeled
25889         iteration in that case.
25891 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25892             Alan Hayward  <alan.hayward@arm.com>
25893             David Sherwood  <david.sherwood@arm.com>
25895         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25896         single-element interleaving even if the size is not a power of 2.
25897         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25898         accesses for single-element interleaving if the group size is
25899         not a power of 2.
25901 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25902             Alan Hayward  <alan.hayward@arm.com>
25903             David Sherwood  <david.sherwood@arm.com>
25905         * doc/md.texi (fold_extract_last_@var{m}): Document.
25906         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25907         * optabs.def (fold_extract_last_optab): New optab.
25908         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25909         * internal-fn.c (fold_extract_direct): New macro.
25910         (expand_fold_extract_optab_fn): Likewise.
25911         (direct_fold_extract_optab_supported_p): Likewise.
25912         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25913         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25914         EXTRACT_LAST_REDUCTION.
25915         (get_initial_def_for_reduction): Do not create an initial vector
25916         for EXTRACT_LAST_REDUCTION reductions.
25917         (vectorizable_reduction): Leave the scalar phi in place for
25918         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25919         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25920         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25921         transform phase to vectorizable_condition.
25922         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25923         split out from...
25924         (vect_finish_stmt_generation): ...here.
25925         (vect_finish_replace_stmt): New function.
25926         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25927         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25928         pattern.
25929         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25931 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25932             Alan Hayward  <alan.hayward@arm.com>
25933             David Sherwood  <david.sherwood@arm.com>
25935         * doc/md.texi (extract_last_@var{m}): Document.
25936         * optabs.def (extract_last_optab): New optab.
25937         * internal-fn.def (EXTRACT_LAST): New internal function.
25938         * internal-fn.c (cond_unary_direct): New macro.
25939         (expand_cond_unary_optab_fn): Likewise.
25940         (direct_cond_unary_optab_supported_p): Likewise.
25941         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25942         loops using EXTRACT_LAST.
25943         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25944         (extract_last_<mode>): ...this optab.
25945         (vec_extract<mode><Vel>): Update accordingly.
25947 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25948             Alan Hayward  <alan.hayward@arm.com>
25949             David Sherwood  <david.sherwood@arm.com>
25951         * target.def (empty_mask_is_expensive): New hook.
25952         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25953         * doc/tm.texi: Regenerate.
25954         * targhooks.h (default_empty_mask_is_expensive): Declare.
25955         * targhooks.c (default_empty_mask_is_expensive): New function.
25956         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25957         if the target says that empty masks are expensive.
25958         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25959         New function.
25960         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25962 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25963             Alan Hayward  <alan.hayward@arm.com>
25964             David Sherwood  <david.sherwood@arm.com>
25966         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25967         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25968         (vect_use_loop_mask_for_alignment_p): New function.
25969         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25970         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25971         niters_skip argument.  Make sure that the first niters_skip elements
25972         of the first iteration are inactive.
25973         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25974         Update call to vect_set_loop_masks_directly.
25975         (get_misalign_in_elems): New function, split out from...
25976         (vect_gen_prolog_loop_niters): ...here.
25977         (vect_update_init_of_dr): Take a code argument that specifies whether
25978         the adjustment should be added or subtracted.
25979         (vect_update_init_of_drs): Likewise.
25980         (vect_prepare_for_masked_peels): New function.
25981         (vect_do_peeling): Skip prologue peeling if we're using a mask
25982         instead.  Update call to vect_update_inits_of_drs.
25983         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25984         mask_skip_niters.
25985         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25986         alignment.  Do not include the number of peeled iterations in
25987         the minimum threshold in that case.
25988         (vectorizable_induction): Adjust the start value down by
25989         LOOP_VINFO_MASK_SKIP_NITERS iterations.
25990         (vect_transform_loop): Call vect_prepare_for_masked_peels.
25991         Take the number of skipped iterations into account when calculating
25992         the loop bounds.
25993         * tree-vect-stmts.c (vect_gen_while_not): New function.
25995 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25996             Alan Hayward  <alan.hayward@arm.com>
25997             David Sherwood  <david.sherwood@arm.com>
25999         * doc/sourcebuild.texi (vect_fully_masked): Document.
26000         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26001         default value to 0.
26002         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26003         split out from...
26004         (vect_analyze_loop_2): ...here. Don't check the vectorization
26005         factor against the number of loop iterations if the loop is
26006         fully-masked.
26008 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26009             Alan Hayward  <alan.hayward@arm.com>
26010             David Sherwood  <david.sherwood@arm.com>
26012         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26013         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26014         (dump_groups): Update accordingly.
26015         (iv_use::mem_type): New member variable.
26016         (address_p): New function.
26017         (record_use): Add a mem_type argument and initialize the new
26018         mem_type field.
26019         (record_group_use): Add a mem_type argument.  Use address_p.
26020         Remove obsolete null checks of base_object.  Update call to record_use.
26021         (find_interesting_uses_op): Update call to record_group_use.
26022         (find_interesting_uses_cond): Likewise.
26023         (find_interesting_uses_address): Likewise.
26024         (get_mem_type_for_internal_fn): New function.
26025         (find_address_like_use): Likewise.
26026         (find_interesting_uses_stmt): Try find_address_like_use before
26027         calling find_interesting_uses_op.
26028         (addr_offset_valid_p): Use the iv mem_type field as the type
26029         of the addressed memory.
26030         (add_autoinc_candidates): Likewise.
26031         (get_address_cost): Likewise.
26032         (split_small_address_groups_p): Use address_p.
26033         (split_address_groups): Likewise.
26034         (add_iv_candidate_for_use): Likewise.
26035         (autoinc_possible_for_pair): Likewise.
26036         (rewrite_groups): Likewise.
26037         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26038         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26039         (get_alias_ptr_type_for_ptr_address): New function.
26040         (rewrite_use_address): Rewrite address uses in calls that were
26041         identified by find_address_like_use.
26043 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26044             Alan Hayward  <alan.hayward@arm.com>
26045             David Sherwood  <david.sherwood@arm.com>
26047         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26048         TARGET_MEM_REFs.
26049         * gimple-expr.h (is_gimple_addressable: Likewise.
26050         * gimple-expr.c (is_gimple_address): Likewise.
26051         * internal-fn.c (expand_call_mem_ref): New function.
26052         (expand_mask_load_optab_fn): Use it.
26053         (expand_mask_store_optab_fn): Likewise.
26055 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26056             Alan Hayward  <alan.hayward@arm.com>
26057             David Sherwood  <david.sherwood@arm.com>
26059         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26060         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26061         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26062         (cond_umax@var{mode}): Document.
26063         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26064         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26065         (cond_umin_optab, cond_umax_optab): New optabs.
26066         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26067         (COND_IOR, COND_XOR): New internal functions.
26068         * internal-fn.h (get_conditional_internal_fn): Declare.
26069         * internal-fn.c (cond_binary_direct): New macro.
26070         (expand_cond_binary_optab_fn): Likewise.
26071         (direct_cond_binary_optab_supported_p): Likewise.
26072         (get_conditional_internal_fn): New function.
26073         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26074         Cope with reduction statements that are vectorized as calls rather
26075         than assignments.
26076         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26077         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26078         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26079         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26080         (UNSPEC_COND_EOR): New unspecs.
26081         (optab): Add mappings for them.
26082         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26083         (sve_int_op, sve_fp_op): New int attributes.
26085 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26086             Alan Hayward  <alan.hayward@arm.com>
26087             David Sherwood  <david.sherwood@arm.com>
26089         * optabs.def (while_ult_optab): New optab.
26090         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26091         * internal-fn.def (WHILE_ULT): New internal function.
26092         * internal-fn.h (direct_internal_fn_supported_p): New override
26093         that takes two types as argument.
26094         * internal-fn.c (while_direct): New macro.
26095         (expand_while_optab_fn): New function.
26096         (convert_optab_supported_p): Likewise.
26097         (direct_while_optab_supported_p): New macro.
26098         * wide-int.h (wi::udiv_ceil): New function.
26099         * tree-vectorizer.h (rgroup_masks): New structure.
26100         (vec_loop_masks): New typedef.
26101         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26102         and fully_masked_p.
26103         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26104         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26105         (vect_max_vf): New function.
26106         (slpeel_make_loop_iterate_ntimes): Delete.
26107         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26108         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26109         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26110         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26111         internal-fn.h, stor-layout.h and optabs-query.h.
26112         (vect_set_loop_mask): New function.
26113         (add_preheader_seq): Likewise.
26114         (add_header_seq): Likewise.
26115         (interleave_supported_p): Likewise.
26116         (vect_maybe_permute_loop_masks): Likewise.
26117         (vect_set_loop_masks_directly): Likewise.
26118         (vect_set_loop_condition_masked): Likewise.
26119         (vect_set_loop_condition_unmasked): New function, split out from
26120         slpeel_make_loop_iterate_ntimes.
26121         (slpeel_make_loop_iterate_ntimes): Rename to..
26122         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26123         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26124         (vect_do_peeling): Update call accordingly.
26125         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26126         loops.
26127         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26128         mask_compare_type, can_fully_mask_p and fully_masked_p.
26129         (release_vec_loop_masks): New function.
26130         (_loop_vec_info): Use it to free the loop masks.
26131         (can_produce_all_loop_masks_p): New function.
26132         (vect_get_max_nscalars_per_iter): Likewise.
26133         (vect_verify_full_masking): Likewise.
26134         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26135         retries, and free the mask rgroups before retrying.  Check loop-wide
26136         reasons for disallowing fully-masked loops.  Make the final decision
26137         about whether use a fully-masked loop or not.
26138         (vect_estimate_min_profitable_iters): Do not assume that peeling
26139         for the number of iterations will be needed for fully-masked loops.
26140         (vectorizable_reduction): Disable fully-masked loops.
26141         (vectorizable_live_operation): Likewise.
26142         (vect_halve_mask_nunits): New function.
26143         (vect_double_mask_nunits): Likewise.
26144         (vect_record_loop_mask): Likewise.
26145         (vect_get_loop_mask): Likewise.
26146         (vect_transform_loop): Handle the case in which the final loop
26147         iteration might handle a partial vector.  Call vect_set_loop_condition
26148         instead of slpeel_make_loop_iterate_ntimes.
26149         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26150         (check_load_store_masking): New function.
26151         (prepare_load_store_mask): Likewise.
26152         (vectorizable_store): Handle fully-masked loops.
26153         (vectorizable_load): Likewise.
26154         (supportable_widening_operation): Use vect_halve_mask_nunits for
26155         booleans.
26156         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26157         (vect_gen_while): New function.
26158         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26159         (aarch64_uqdec<mode>): New insn.
26161 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26162             Alan Hayward  <alan.hayward@arm.com>
26163             David Sherwood  <david.sherwood@arm.com>
26165         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26166         (reduc_xor_scal_optab): New optabs.
26167         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26168         (reduc_xor_scal_@var{m}): Document.
26169         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26170         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26171         internal functions.
26172         * fold-const-call.c (fold_const_call): Handle them.
26173         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26174         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26175         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26176         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26177         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26178         (UNSPEC_XORV): New unspecs.
26179         (optab): Add entries for them.
26180         (BITWISEV): New int iterator.
26181         (bit_reduc_op): New int attributes.
26183 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26184             Alan Hayward  <alan.hayward@arm.com>
26185             David Sherwood  <david.sherwood@arm.com>
26187         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26188         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26189         * optabs.def (vec_shl_insert_optab): New optab.
26190         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26191         (duplicate_and_interleave): Likewise.
26192         * tree-vect-loop.c: Include internal-fn.h.
26193         (neutral_op_for_slp_reduction): New function, split out from
26194         get_initial_defs_for_reduction.
26195         (get_initial_def_for_reduction): Handle option 2 for variable-length
26196         vectors by loading the neutral value into a vector and then shifting
26197         the initial value into element 0.
26198         (get_initial_defs_for_reduction): Replace the code argument with
26199         the neutral value calculated by neutral_op_for_slp_reduction.
26200         Use gimple_build_vector for constant-length vectors.
26201         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26202         but the first group_size elements have a neutral value.
26203         Use duplicate_and_interleave otherwise.
26204         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26205         Update call to get_initial_defs_for_reduction.  Handle SLP
26206         reductions for variable-length vectors by creating one vector
26207         result for each scalar result, with the elements associated
26208         with other scalar results stubbed out with the neutral value.
26209         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26210         Require IFN_VEC_SHL_INSERT for double reductions on
26211         variable-length vectors, or SLP reductions that have
26212         a neutral value.  Require can_duplicate_and_interleave_p
26213         support for variable-length unchained SLP reductions if there
26214         is no neutral value, such as for MIN/MAX reductions.  Also require
26215         the number of vector elements to be a multiple of the number of
26216         SLP statements when doing variable-length unchained SLP reductions.
26217         Update call to vect_create_epilog_for_reduction.
26218         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26219         and remove initial values.
26220         (duplicate_and_interleave): Make public.
26221         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26222         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26224 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26225             Alan Hayward  <alan.hayward@arm.com>
26226             David Sherwood  <david.sherwood@arm.com>
26228         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26229         (can_duplicate_and_interleave_p): New function.
26230         (vect_get_and_check_slp_defs): Take the vector of statements
26231         rather than just the current one.  Remove excess parentheses.
26232         Restriction rejectinon of vect_constant_def and vect_external_def
26233         for variable-length vectors to boolean types, or types for which
26234         can_duplicate_and_interleave_p is false.
26235         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26236         (duplicate_and_interleave): New function.
26237         (vect_get_constant_vectors): Use gimple_build_vector for
26238         constant-length vectors and suitable variable-length constant
26239         vectors.  Use duplicate_and_interleave for other variable-length
26240         vectors.  Don't defer the update when inserting new statements.
26242 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26243             Alan Hayward  <alan.hayward@arm.com>
26244             David Sherwood  <david.sherwood@arm.com>
26246         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26247         min_profitable_iters doesn't go negative.
26249 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26250             Alan Hayward  <alan.hayward@arm.com>
26251             David Sherwood  <david.sherwood@arm.com>
26253         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26254         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26255         * optabs.def (vec_mask_load_lanes_optab): New optab.
26256         (vec_mask_store_lanes_optab): Likewise.
26257         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26258         (MASK_STORE_LANES): Likewise.
26259         * internal-fn.c (mask_load_lanes_direct): New macro.
26260         (mask_store_lanes_direct): Likewise.
26261         (expand_mask_load_optab_fn): Handle masked operations.
26262         (expand_mask_load_lanes_optab_fn): New macro.
26263         (expand_mask_store_optab_fn): Handle masked operations.
26264         (expand_mask_store_lanes_optab_fn): New macro.
26265         (direct_mask_load_lanes_optab_supported_p): Likewise.
26266         (direct_mask_store_lanes_optab_supported_p): Likewise.
26267         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26268         parameter.
26269         (vect_load_lanes_supported): Likewise.
26270         * tree-vect-data-refs.c (strip_conversion): New function.
26271         (can_group_stmts_p): Likewise.
26272         (vect_analyze_data_ref_accesses): Use it instead of checking
26273         for a pair of assignments.
26274         (vect_store_lanes_supported): Take a masked_p parameter.
26275         (vect_load_lanes_supported): Likewise.
26276         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26277         vect_store_lanes_supported and vect_load_lanes_supported.
26278         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26279         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26280         parameter.  Don't allow gaps for masked accesses.
26281         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26282         and vect_load_lanes_supported.
26283         (get_load_store_type): Take a masked_p parameter and update
26284         call to get_group_load_store_type.
26285         (vectorizable_store): Update call to get_load_store_type.
26286         Handle IFN_MASK_STORE_LANES.
26287         (vectorizable_load): Update call to get_load_store_type.
26288         Handle IFN_MASK_LOAD_LANES.
26290 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26291             Alan Hayward  <alan.hayward@arm.com>
26292             David Sherwood  <david.sherwood@arm.com>
26294         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26295         modes for SVE.
26296         * config/aarch64/aarch64-protos.h
26297         (aarch64_sve_struct_memory_operand_p): Declare.
26298         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26299         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26300         (VPRED, vpred): Handle SVE structure modes.
26301         * config/aarch64/constraints.md (Utx): New constraint.
26302         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26303         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26304         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26305         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26306         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26307         structure modes.  Split into pieces after RA.
26308         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26309         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26310         New patterns.
26311         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26312         SVE structure modes.
26313         (aarch64_classify_address): Likewise.
26314         (sizetochar): Move earlier in file.
26315         (aarch64_print_operand): Handle SVE register lists.
26316         (aarch64_array_mode): New function.
26317         (aarch64_sve_struct_memory_operand_p): Likewise.
26318         (TARGET_ARRAY_MODE): Redefine.
26320 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26321             Alan Hayward  <alan.hayward@arm.com>
26322             David Sherwood  <david.sherwood@arm.com>
26324         * target.def (array_mode): New target hook.
26325         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26326         * doc/tm.texi: Regenerate.
26327         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26328         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26329         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26330         targetm.array_mode.
26331         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26332         type sizes.
26334 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26335             Alan Hayward  <alan.hayward@arm.com>
26336             David Sherwood  <david.sherwood@arm.com>
26338         * fold-const.c (fold_binary_loc): Check the argument types
26339         rather than the result type when testing for a vector operation.
26341 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26343         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26344         * doc/tm.texi: Regenerate.
26346 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26347             Alan Hayward  <alan.hayward@arm.com>
26348             David Sherwood  <david.sherwood@arm.com>
26350         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26351         (sve): Document new AArch64 extension.
26352         * doc/md.texi (w): Extend the description of the AArch64
26353         constraint to include SVE vectors.
26354         (Upl, Upa): Document new AArch64 predicate constraints.
26355         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26356         enum.
26357         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26358         (msve-vector-bits=): New option.
26359         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26360         SVE when these are disabled.
26361         (sve): New extension.
26362         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26363         modes.  Adjust their number of units based on aarch64_sve_vg.
26364         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26365         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26366         aarch64_addr_query_type.
26367         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26368         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26369         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26370         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26371         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26372         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26373         (aarch64_simd_imm_zero_p): Delete.
26374         (aarch64_check_zero_based_sve_index_immediate): Declare.
26375         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26376         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26377         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26378         (aarch64_sve_float_mul_immediate_p): Likewise.
26379         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26380         rather than an rtx.
26381         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26382         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26383         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26384         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26385         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26386         (aarch64_regmode_natural_size): Likewise.
26387         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26388         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26389         left one place.
26390         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26391         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26392         for VG and the SVE predicate registers.
26393         (V_ALIASES): Add a "z"-prefixed alias.
26394         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26395         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26396         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26397         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26398         (REG_CLASS_NAMES): Add entries for them.
26399         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
26400         and the predicate registers.
26401         (aarch64_sve_vg): Declare.
26402         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26403         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26404         (REGMODE_NATURAL_SIZE): Define.
26405         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26406         SVE macros.
26407         * config/aarch64/aarch64.c: Include cfgrtl.h.
26408         (simd_immediate_info): Add a constructor for series vectors,
26409         and an associated step field.
26410         (aarch64_sve_vg): New variable.
26411         (aarch64_dbx_register_number): Handle VG and the predicate registers.
26412         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26413         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26414         (VEC_ANY_DATA, VEC_STRUCT): New constants.
26415         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26416         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26417         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26418         (aarch64_get_mask_mode): New functions.
26419         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26420         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26421         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
26422         predicate modes and predicate registers.  Explicitly restrict
26423         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
26424         to store a vector mode if it is recognized by
26425         aarch64_classify_vector_mode.
26426         (aarch64_regmode_natural_size): New function.
26427         (aarch64_hard_regno_caller_save_mode): Return the original mode
26428         for predicates.
26429         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26430         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26431         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26432         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26433         functions.
26434         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
26435         does not overlap dest if the function is frame-related.  Handle
26436         SVE constants.
26437         (aarch64_split_add_offset): New function.
26438         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26439         them aarch64_add_offset.
26440         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26441         and update call to aarch64_sub_sp.
26442         (aarch64_add_cfa_expression): New function.
26443         (aarch64_expand_prologue): Pass extra temporary registers to the
26444         functions above.  Handle the case in which we need to emit new
26445         DW_CFA_expressions for registers that were originally saved
26446         relative to the stack pointer, but now have to be expressed
26447         relative to the frame pointer.
26448         (aarch64_output_mi_thunk): Pass extra temporary registers to the
26449         functions above.
26450         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
26451         IP0 and IP1 values for SVE frames.
26452         (aarch64_expand_vec_series): New function.
26453         (aarch64_expand_sve_widened_duplicate): Likewise.
26454         (aarch64_expand_sve_const_vector): Likewise.
26455         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26456         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
26457         into the register, rather than emitting a SET directly.
26458         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26459         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26460         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26461         (offset_9bit_signed_scaled_p): New functions.
26462         (aarch64_replicate_bitmask_imm): New function.
26463         (aarch64_bitmask_imm): Use it.
26464         (aarch64_cannot_force_const_mem): Reject expressions involving
26465         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
26466         (aarch64_classify_index): Handle SVE indices, by requiring
26467         a plain register index with a scale that matches the element size.
26468         (aarch64_classify_address): Handle SVE addresses.  Assert that
26469         the mode of the address is VOIDmode or an integer mode.
26470         Update call to aarch64_classify_symbol.
26471         (aarch64_classify_symbolic_expression): Update call to
26472         aarch64_classify_symbol.
26473         (aarch64_const_vec_all_in_range_p): New function.
26474         (aarch64_print_vector_float_operand): Likewise.
26475         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
26476         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
26477         and the FP immediates 1.0 and 0.5.
26478         (aarch64_print_address_internal): Handle SVE addresses.
26479         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26480         (aarch64_regno_regclass): Handle predicate registers.
26481         (aarch64_secondary_reload): Handle big-endian reloads of SVE
26482         data modes.
26483         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26484         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26485         (aarch64_convert_sve_vector_bits): New function.
26486         (aarch64_override_options): Use it to handle -msve-vector-bits=.
26487         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26488         rather than an rtx.
26489         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26490         Handle SVE vector and predicate modes.  Accept VL-based constants
26491         that need only one temporary register, and VL offsets that require
26492         no temporary registers.
26493         (aarch64_conditional_register_usage): Mark the predicate registers
26494         as fixed if SVE isn't available.
26495         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26496         Return true for SVE vector and predicate modes.
26497         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26498         rather than an unsigned int.  Handle SVE modes.
26499         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
26500         SVE modes.
26501         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26502         if SVE is enabled.
26503         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26504         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26505         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26506         (aarch64_sve_float_mul_immediate_p): New functions.
26507         (aarch64_sve_valid_immediate): New function.
26508         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
26509         Explicitly reject structure modes.  Check for INDEX constants.
26510         Handle PTRUE and PFALSE constants.
26511         (aarch64_check_zero_based_sve_index_immediate): New function.
26512         (aarch64_simd_imm_zero_p): Delete.
26513         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
26514         vector modes.  Accept constants in the range of CNT[BHWD].
26515         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
26516         ask for an Advanced SIMD mode.
26517         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
26518         (aarch64_simd_vector_alignment): Handle SVE predicates.
26519         (aarch64_vectorize_preferred_vector_alignment): New function.
26520         (aarch64_simd_vector_alignment_reachable): Use it instead of
26521         the vector size.
26522         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
26523         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
26524         functions.
26525         (MAX_VECT_LEN): Delete.
26526         (expand_vec_perm_d): Add a vec_flags field.
26527         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
26528         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
26529         (aarch64_evpc_ext): Don't apply a big-endian lane correction
26530         for SVE modes.
26531         (aarch64_evpc_rev): Rename to...
26532         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
26533         (aarch64_evpc_rev_global): New function.
26534         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
26535         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
26536         MAX_VECT_LEN.
26537         (aarch64_evpc_sve_tbl): New function.
26538         (aarch64_expand_vec_perm_const_1): Update after rename of
26539         aarch64_evpc_rev.  Handle SVE permutes too, trying
26540         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
26541         than aarch64_evpc_tbl.
26542         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
26543         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
26544         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
26545         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
26546         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
26547         (aarch64_expand_sve_vcond): New functions.
26548         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
26549         of aarch64_vector_mode_p.
26550         (aarch64_dwarf_poly_indeterminate_value): New function.
26551         (aarch64_compute_pressure_classes): Likewise.
26552         (aarch64_can_change_mode_class): Likewise.
26553         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
26554         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
26555         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
26556         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
26557         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
26558         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
26559         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
26560         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
26561         constraints.
26562         (Dn, Dl, Dr): Accept const as well as const_vector.
26563         (Dz): Likewise.  Compare against CONST0_RTX.
26564         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
26565         of "vector" where appropriate.
26566         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
26567         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
26568         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
26569         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
26570         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
26571         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
26572         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
26573         (v_int_equiv): Extend to SVE modes.
26574         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
26575         mode attributes.
26576         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
26577         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
26578         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
26579         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
26580         (SVE_COND_FP_CMP): New int iterators.
26581         (perm_hilo): Handle the new unpack unspecs.
26582         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
26583         attributes.
26584         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
26585         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
26586         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
26587         (aarch64_equality_operator, aarch64_constant_vector_operand)
26588         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
26589         (aarch64_sve_nonimmediate_operand): Likewise.
26590         (aarch64_sve_general_operand): Likewise.
26591         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
26592         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
26593         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
26594         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
26595         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
26596         (aarch64_sve_float_arith_immediate): Likewise.
26597         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
26598         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
26599         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
26600         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
26601         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
26602         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
26603         (aarch64_sve_float_arith_operand): Likewise.
26604         (aarch64_sve_float_arith_with_sub_operand): Likewise.
26605         (aarch64_sve_float_mul_operand): Likewise.
26606         (aarch64_sve_vec_perm_operand): Likewise.
26607         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
26608         (aarch64_mov_operand): Accept const_poly_int and const_vector.
26609         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
26610         as well as const_vector.
26611         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
26612         in file.  Use CONST0_RTX and CONSTM1_RTX.
26613         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
26614         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
26615         Use aarch64_simd_imm_zero.
26616         * config/aarch64/aarch64-sve.md: New file.
26617         * config/aarch64/aarch64.md: Include it.
26618         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
26619         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
26620         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
26621         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
26622         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
26623         (sve): New attribute.
26624         (enabled): Disable instructions with the sve attribute unless
26625         TARGET_SVE.
26626         (movqi, movhi): Pass CONST_POLY_INT operaneds through
26627         aarch64_expand_mov_immediate.
26628         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
26629         CNT[BHSD] immediates.
26630         (movti): Split CONST_POLY_INT moves into two halves.
26631         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
26632         Split additions that need a temporary here if the destination
26633         is the stack pointer.
26634         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
26635         (*add<mode>3_poly_1): New instruction.
26636         (set_clobber_cc): New expander.
26638 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26640         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
26641         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
26642         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
26643         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
26644         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
26645         Change innermode from fixed_mode_size to machine_mode.
26646         (simplify_subreg): Update call accordingly.  Handle a constant-sized
26647         subreg of a variable-length CONST_VECTOR.
26649 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26650             Alan Hayward  <alan.hayward@arm.com>
26651             David Sherwood  <david.sherwood@arm.com>
26653         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
26654         (add_offset_to_base): New function, split out from...
26655         (create_mem_ref): ...here.  When handling a scale other than 1,
26656         check first whether the address is valid without the offset.
26657         Add it into the base if so, leaving the index and scale as-is.
26659 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26661         PR c++/83778
26662         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
26663         fold_for_warn before checking if arg2 is INTEGER_CST.
26665 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
26667         * config/rs6000/predicates.md (load_multiple_operation): Delete.
26668         (store_multiple_operation): Delete.
26669         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
26670         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
26671         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
26672         guarded by TARGET_STRING.
26673         (rs6000_output_load_multiple): Delete.
26674         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
26675         OPTION_MASK_STRING / TARGET_STRING handling.
26676         (print_operand) <'N', 'O'>: Add comment that these are unused now.
26677         (const rs6000_opt_masks) <"string">: Change mask to 0.
26678         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
26679         (MASK_STRING): Delete.
26680         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
26681         parts.  Simplify.
26682         (load_multiple): Delete.
26683         (*ldmsi8): Delete.
26684         (*ldmsi7): Delete.
26685         (*ldmsi6): Delete.
26686         (*ldmsi5): Delete.
26687         (*ldmsi4): Delete.
26688         (*ldmsi3): Delete.
26689         (store_multiple): Delete.
26690         (*stmsi8): Delete.
26691         (*stmsi7): Delete.
26692         (*stmsi6): Delete.
26693         (*stmsi5): Delete.
26694         (*stmsi4): Delete.
26695         (*stmsi3): Delete.
26696         (movmemsi_8reg): Delete.
26697         (corresponding unnamed define_insn): Delete.
26698         (movmemsi_6reg): Delete.
26699         (corresponding unnamed define_insn): Delete.
26700         (movmemsi_4reg): Delete.
26701         (corresponding unnamed define_insn): Delete.
26702         (movmemsi_2reg): Delete.
26703         (corresponding unnamed define_insn): Delete.
26704         (movmemsi_1reg): Delete.
26705         (corresponding unnamed define_insn): Delete.
26706         * config/rs6000/rs6000.opt (mno-string): New.
26707         (mstring): Replace by deprecation warning stub.
26708         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
26710 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26712         * regrename.c (regrename_do_replace): If replacing the same
26713         reg multiple times, try to reuse last created gen_raw_REG.
26715         PR debug/81155
26716         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
26717         main to workaround a bug in GDB.
26719 2018-01-12  Tom de Vries  <tom@codesourcery.com>
26721         PR target/83737
26722         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
26724 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
26726         PR rtl-optimization/80481
26727         * ira-color.c (get_cap_member): New function.
26728         (allocnos_conflict_by_live_ranges_p): Use it.
26729         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
26730         (setup_slot_coalesced_allocno_live_ranges): Ditto.
26732 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
26734         PR target/83628
26735         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
26736         (*saddl_se_1): Ditto.
26737         (*ssubsi_1): Ditto.
26738         (*ssubl_se_1): Ditto.
26740 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26742         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
26743         rather than wi::to_widest for DR_INITs.
26744         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
26745         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
26746         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
26747         INTEGER_CSTs.
26748         (vect_analyze_group_access_1): Note that here.
26750 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26752         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
26753         polynomial type sizes.
26755 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26757         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
26758         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
26759         (gimple_add_tmp_var): Likewise.
26761 2018-01-12  Martin Liska  <mliska@suse.cz>
26763         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
26764         (gimple_alloc_sizes): Likewise.
26765         (dump_gimple_statistics): Use PRIu64 in printf format.
26766         * gimple.h: Change uint64_t to int.
26768 2018-01-12  Martin Liska  <mliska@suse.cz>
26770         * tree-core.h: Use uint64_t instead of int.
26771         * tree.c (tree_node_counts): Likewise.
26772         (tree_node_sizes): Likewise.
26773         (dump_tree_statistics): Use PRIu64 in printf format.
26775 2018-01-12  Martin Liska  <mliska@suse.cz>
26777         * Makefile.in: As qsort_chk is implemented in vec.c, add
26778         vec.o to linkage of gencfn-macros.
26779         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
26780         passing the info to record_node_allocation_statistics.
26781         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
26782         and pass the info.
26783         * ggc-common.c (struct ggc_usage): Add operator== and use
26784         it in operator< and compare function.
26785         * mem-stats.h (struct mem_usage): Likewise.
26786         * vec.c (struct vec_usage): Remove operator< and compare
26787         function. Can be simply inherited.
26789 2018-01-12  Martin Jambor  <mjambor@suse.cz>
26791         PR target/81616
26792         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26793         * tree-ssa-math-opts.c: Include domwalk.h.
26794         (convert_mult_to_fma_1): New function.
26795         (fma_transformation_info): New type.
26796         (fma_deferring_state): Likewise.
26797         (cancel_fma_deferring): New function.
26798         (result_of_phi): Likewise.
26799         (last_fma_candidate_feeds_initial_phi): Likewise.
26800         (convert_mult_to_fma): Added deferring logic, split actual
26801         transformation to convert_mult_to_fma_1.
26802         (math_opts_dom_walker): New type.
26803         (math_opts_dom_walker::after_dom_children): New method, body moved
26804         here from pass_optimize_widening_mul::execute, added deferring logic
26805         bits.
26806         (pass_optimize_widening_mul::execute): Moved most of code to
26807         math_opts_dom_walker::after_dom_children.
26808         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26809         * config/i386/i386.c (ix86_option_override_internal): Added
26810         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26812 2018-01-12  Richard Biener  <rguenther@suse.de>
26814         PR debug/83157
26815         * dwarf2out.c (gen_variable_die): Do not reset old_die for
26816         inline instance vars.
26818 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
26820         PR target/81819
26821         * config/rx/rx.c (rx_is_restricted_memory_address):
26822         Handle SUBREG case.
26824 2018-01-12  Richard Biener  <rguenther@suse.de>
26826         PR tree-optimization/80846
26827         * target.def (split_reduction): New target hook.
26828         * targhooks.c (default_split_reduction): New function.
26829         * targhooks.h (default_split_reduction): Declare.
26830         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26831         target requests first reduce vectors by combining low and high
26832         parts.
26833         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26834         (get_vectype_for_scalar_type_and_size): Export.
26835         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26836         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26837         * doc/tm.texi: Regenerate.
26838         * config/i386/i386.c (ix86_split_reduction): Implement
26839         TARGET_VECTORIZE_SPLIT_REDUCTION.
26841 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26843         PR target/83368
26844         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26845         in PIC mode except for TARGET_VXWORKS_RTP.
26846         * config/sparc/sparc.c: Include cfgrtl.h.
26847         (TARGET_INIT_PIC_REG): Define.
26848         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26849         (sparc_pic_register_p): New predicate.
26850         (sparc_legitimate_address_p): Use it.
26851         (sparc_legitimize_pic_address): Likewise.
26852         (sparc_delegitimize_address): Likewise.
26853         (sparc_mode_dependent_address_p): Likewise.
26854         (gen_load_pcrel_sym): Remove 4th parameter.
26855         (load_got_register): Adjust call to above.  Remove obsolete stuff.
26856         (sparc_expand_prologue): Do not call load_got_register here.
26857         (sparc_flat_expand_prologue): Likewise.
26858         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26859         (sparc_use_pseudo_pic_reg): New function.
26860         (sparc_init_pic_reg): Likewise.
26861         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26862         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26864 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
26866         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26867         Add item for branch_cost.
26869 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26871         PR rtl-optimization/83565
26872         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26873         not extend the result to a larger mode for rotate operations.
26874         (num_sign_bit_copies1): Likewise.
26876 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26878         PR target/40411
26879         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26880         -symbolic.
26881         Use values-Xc.o for -pedantic.
26882         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26884 2018-01-12  Martin Liska  <mliska@suse.cz>
26886         PR ipa/83054
26887         * ipa-devirt.c (final_warning_record::grow_type_warnings):
26888         New function.
26889         (possible_polymorphic_call_targets): Use it.
26890         (ipa_devirt): Likewise.
26892 2018-01-12  Martin Liska  <mliska@suse.cz>
26894         * profile-count.h (enum profile_quality): Use 0 as invalid
26895         enum value of profile_quality.
26897 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26899         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26900         -mext-string options.
26902 2018-01-12  Richard Biener  <rguenther@suse.de>
26904         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26905         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26906         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26907         Likewise.
26908         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26910 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26912         * configure.ac (--with-long-double-format): Add support for the
26913         configuration option to change the default long double format on
26914         PowerPC systems.
26915         * config.gcc (powerpc*-linux*-*): Likewise.
26916         * configure: Regenerate.
26917         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26918         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26919         used without modification.
26921 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26923         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26924         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26925         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26926         MISC_BUILTIN_SPEC_BARRIER.
26927         (rs6000_init_builtins): Likewise.
26928         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26929         enum value.
26930         (speculation_barrier): New define_insn.
26931         * doc/extend.texi: Document __builtin_speculation_barrier.
26933 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26935         PR target/83203
26936         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26937         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26938         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26939         iterators.
26940         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26941         integral modes instead of "ss" and "sd".
26942         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26943         vectors with 32-bit and 64-bit elements.
26944         (vecdupssescalarmodesuffix): New mode attribute.
26945         (vec_dup<mode>): Use it.
26947 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
26949         PR target/83330
26950         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26951         frame if argument is passed on stack.
26953 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26955         PR target/82682
26956         * ree.c (combine_reaching_defs): Optimize also
26957         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26958         reg2=any_extend(exp); reg1=reg2;, formatting fix.
26960 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26962         PR middle-end/83189
26963         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26965 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26967         PR middle-end/83718
26968         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26969         after they are computed.
26971 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
26973         PR tree-optimization/83695
26974         * gimple-loop-linterchange.cc
26975         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26976         reset cached scev information after interchange.
26977         (pass_linterchange::execute): Remove call to scev_reset_htab.
26979 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26981         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26982         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26983         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26984         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26985         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26986         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26987         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26988         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26989         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26990         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26991         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26992         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26993         (V_lane_reg): Likewise.
26994         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26995         New define_expand.
26996         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26997         (vfmal_lane_low<mode>_intrinsic,
26998         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26999         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27000         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27001         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27002         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27003         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27005 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27007         * config/arm/arm-cpus.in (fp16fml): New feature.
27008         (ALL_SIMD): Add fp16fml.
27009         (armv8.2-a): Add fp16fml as an option.
27010         (armv8.3-a): Likewise.
27011         (armv8.4-a): Add fp16fml as part of fp16.
27012         * config/arm/arm.h (TARGET_FP16FML): Define.
27013         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27014         when appropriate.
27015         * config/arm/arm-modes.def (V2HF): Define.
27016         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27017         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27018         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27019         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27020         vfmsl_low, vfmsl_high): New set of builtins.
27021         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27022         (vfml_op): New code attribute.
27023         (VFMLHALVES): New int iterator.
27024         (VFML, VFMLSEL): New mode attributes.
27025         (V_reg): Define mapping for V2HF.
27026         (V_hi, V_lo): New mode attributes.
27027         (VF_constraint): Likewise.
27028         (vfml_half, vfml_half_selector): New int attributes.
27029         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27030         define_expand.
27031         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27032         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27033         New define_insn.
27034         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27035         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27036         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27037         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27038         documentation.
27039         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27040         Document new effective target and option set.
27042 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27044         * config/arm/arm-cpus.in (armv8_4): New feature.
27045         (ARMv8_4a): New fgroup.
27046         (armv8.4-a): New arch.
27047         * config/arm/arm-tables.opt: Regenerate.
27048         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27049         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27050         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27051         Add matching rules for -march=armv8.4-a and extensions.
27052         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27054 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27056         PR target/81821
27057         * config/rx/rx.md (BW): New mode attribute.
27058         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27060 2018-01-11  Richard Biener  <rguenther@suse.de>
27062         PR tree-optimization/83435
27063         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27064         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27065         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27067 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27068             Alan Hayward  <alan.hayward@arm.com>
27069             David Sherwood  <david.sherwood@arm.com>
27071         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27072         field.
27073         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27074         (aarch64_print_address_internal): Use it to check for a zero offset.
27076 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27077             Alan Hayward  <alan.hayward@arm.com>
27078             David Sherwood  <david.sherwood@arm.com>
27080         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27081         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27082         Return a poly_int64 rather than a HOST_WIDE_INT.
27083         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27084         rather than a HOST_WIDE_INT.
27085         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27086         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27087         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27088         final_offset from HOST_WIDE_INT to poly_int64.
27089         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27090         to_constant when getting the number of units in an Advanced SIMD
27091         mode.
27092         (aarch64_builtin_vectorized_function): Check for a constant number
27093         of units.
27094         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27095         GET_MODE_SIZE.
27096         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27097         attribute instead of GET_MODE_NUNITS.
27098         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27099         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27100         GET_MODE_SIZE for fixed-size registers.
27101         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27102         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27103         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27104         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27105         (aarch64_print_operand, aarch64_print_address_internal)
27106         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27107         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27108         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27109         Handle polynomial GET_MODE_SIZE.
27110         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27111         wider than SImode without modification.
27112         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27113         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27114         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27115         passing and returning SVE modes.
27116         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27117         rather than GEN_INT.
27118         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27119         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27120         (aarch64_allocate_and_probe_stack_space): Likewise.
27121         (aarch64_layout_frame): Cope with polynomial offsets.
27122         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27123         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27124         polynomial offsets.
27125         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27126         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27127         poly_int64 rather than a HOST_WIDE_INT.
27128         (aarch64_get_separate_components, aarch64_process_components)
27129         (aarch64_expand_prologue, aarch64_expand_epilogue)
27130         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27131         (aarch64_anchor_offset): New function, split out from...
27132         (aarch64_legitimize_address): ...here.
27133         (aarch64_builtin_vectorization_cost): Handle polynomial
27134         TYPE_VECTOR_SUBPARTS.
27135         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27136         GET_MODE_NUNITS.
27137         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27138         number of elements from the PARALLEL rather than the mode.
27139         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27140         rather than GET_MODE_BITSIZE.
27141         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27142         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27143         (aarch64_expand_vec_perm_const_1): Handle polynomial
27144         d->perm.length () and d->perm elements.
27145         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27146         Apply to_constant to d->perm elements.
27147         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27148         polynomial CONST_VECTOR_NUNITS.
27149         (aarch64_move_pointer): Take amount as a poly_int64 rather
27150         than an int.
27151         (aarch64_progress_pointer): Avoid temporary variable.
27152         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27153         the mode attribute instead of GET_MODE.
27155 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27156             Alan Hayward  <alan.hayward@arm.com>
27157             David Sherwood  <david.sherwood@arm.com>
27159         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27160         x exists before using it.
27161         (aarch64_add_constant_internal): Rename to...
27162         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27163         src and dest rtxes.  Handle the case in which they're different,
27164         including when the offset is zero.  Replace scratchreg with an rtx.
27165         Use 2 additions if there is no spare register into which we can
27166         move a 16-bit constant.
27167         (aarch64_add_constant): Delete.
27168         (aarch64_add_offset): Replace reg with separate src and dest
27169         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27170         Use aarch64_add_offset_1.
27171         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27172         an rtx rather than an int.  Take the delta as a poly_int64
27173         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27174         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27175         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27176         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27177         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27178         and aarch64_add_sp.
27179         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27180         aarch64_add_constant.
27182 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27184         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27185         Use scalar_float_mode.
27187 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27189         * config/aarch64/aarch64-simd.md
27190         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27191         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27192         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27193         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27194         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27195         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27196         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27197         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27198         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27199         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27201 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27203         PR target/83514
27204         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27205         targ_options->x_arm_arch_string is non NULL.
27207 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27209         * config/aarch64/aarch64.h
27210         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27212 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27214         PR target/82096
27215         * expmed.c (emit_store_flag_force): Swap if const op0
27216         and change VOIDmode to mode of op0.
27218 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27220         PR rtl-optimization/83761
27221         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27222         than bytes to mode_for_size.
27224 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27226         PR middle-end/83189
27227         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27228         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27229         profile.
27231 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27233         PR middle-end/83575
27234         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27235         when in layout mode.
27236         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27237         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27238         partition fixup.
27240 2018-01-10  Michael Collison  <michael.collison@arm.com>
27242         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27243         * config/aarch64/aarch64-option-extension.def: Add
27244         AARCH64_OPT_EXTENSION of 'fp16fml'.
27245         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27246         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27247         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27248         * config/aarch64/constraints.md (Ui7): New constraint.
27249         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27250         (VFMLA_SEL_W): Ditto.
27251         (f16quad): Ditto.
27252         (f16mac1): Ditto.
27253         (VFMLA16_LOW): New int iterator.
27254         (VFMLA16_HIGH): Ditto.
27255         (UNSPEC_FMLAL): New unspec.
27256         (UNSPEC_FMLSL): Ditto.
27257         (UNSPEC_FMLAL2): Ditto.
27258         (UNSPEC_FMLSL2): Ditto.
27259         (f16mac): New code attribute.
27260         * config/aarch64/aarch64-simd-builtins.def
27261         (aarch64_fmlal_lowv2sf): Ditto.
27262         (aarch64_fmlsl_lowv2sf): Ditto.
27263         (aarch64_fmlalq_lowv4sf): Ditto.
27264         (aarch64_fmlslq_lowv4sf): Ditto.
27265         (aarch64_fmlal_highv2sf): Ditto.
27266         (aarch64_fmlsl_highv2sf): Ditto.
27267         (aarch64_fmlalq_highv4sf): Ditto.
27268         (aarch64_fmlslq_highv4sf): Ditto.
27269         (aarch64_fmlal_lane_lowv2sf): Ditto.
27270         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27271         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27272         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27273         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27274         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27275         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27276         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27277         (aarch64_fmlal_lane_highv2sf): Ditto.
27278         (aarch64_fmlsl_lane_highv2sf): Ditto.
27279         (aarch64_fmlal_laneq_highv2sf): Ditto.
27280         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27281         (aarch64_fmlalq_lane_highv4sf): Ditto.
27282         (aarch64_fmlsl_lane_highv4sf): Ditto.
27283         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27284         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27285         * config/aarch64/aarch64-simd.md:
27286         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27287         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27288         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27289         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27290         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27291         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27292         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27293         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27294         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27295         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27296         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27297         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27298         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27299         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27300         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27301         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27302         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27303         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27304         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27305         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27306         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27307         (vfmlsl_low_u32): Ditto.
27308         (vfmlalq_low_u32): Ditto.
27309         (vfmlslq_low_u32): Ditto.
27310         (vfmlal_high_u32): Ditto.
27311         (vfmlsl_high_u32): Ditto.
27312         (vfmlalq_high_u32): Ditto.
27313         (vfmlslq_high_u32): Ditto.
27314         (vfmlal_lane_low_u32): Ditto.
27315         (vfmlsl_lane_low_u32): Ditto.
27316         (vfmlal_laneq_low_u32): Ditto.
27317         (vfmlsl_laneq_low_u32): Ditto.
27318         (vfmlalq_lane_low_u32): Ditto.
27319         (vfmlslq_lane_low_u32): Ditto.
27320         (vfmlalq_laneq_low_u32): Ditto.
27321         (vfmlslq_laneq_low_u32): Ditto.
27322         (vfmlal_lane_high_u32): Ditto.
27323         (vfmlsl_lane_high_u32): Ditto.
27324         (vfmlal_laneq_high_u32): Ditto.
27325         (vfmlsl_laneq_high_u32): Ditto.
27326         (vfmlalq_lane_high_u32): Ditto.
27327         (vfmlslq_lane_high_u32): Ditto.
27328         (vfmlalq_laneq_high_u32): Ditto.
27329         (vfmlslq_laneq_high_u32): Ditto.
27330         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27331         (AARCH64_FL_FOR_ARCH8_4): New.
27332         (AARCH64_ISA_F16FML): New ISA flag.
27333         (TARGET_F16FML): New feature flag for fp16fml.
27334         (doc/invoke.texi): Document new fp16fml option.
27336 2018-01-10  Michael Collison  <michael.collison@arm.com>
27338         * config/aarch64/aarch64-builtins.c:
27339         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27340         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27341         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27342         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27343         (AARCH64_ISA_SHA3): New ISA flag.
27344         (TARGET_SHA3): New feature flag for sha3.
27345         * config/aarch64/iterators.md (sha512_op): New int attribute.
27346         (CRYPTO_SHA512): New int iterator.
27347         (UNSPEC_SHA512H): New unspec.
27348         (UNSPEC_SHA512H2): Ditto.
27349         (UNSPEC_SHA512SU0): Ditto.
27350         (UNSPEC_SHA512SU1): Ditto.
27351         * config/aarch64/aarch64-simd-builtins.def
27352         (aarch64_crypto_sha512hqv2di): New builtin.
27353         (aarch64_crypto_sha512h2qv2di): Ditto.
27354         (aarch64_crypto_sha512su0qv2di): Ditto.
27355         (aarch64_crypto_sha512su1qv2di): Ditto.
27356         (aarch64_eor3qv8hi): Ditto.
27357         (aarch64_rax1qv2di): Ditto.
27358         (aarch64_xarqv2di): Ditto.
27359         (aarch64_bcaxqv8hi): Ditto.
27360         * config/aarch64/aarch64-simd.md:
27361         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27362         (aarch64_crypto_sha512su0qv2di): Ditto.
27363         (aarch64_crypto_sha512su1qv2di): Ditto.
27364         (aarch64_eor3qv8hi): Ditto.
27365         (aarch64_rax1qv2di): Ditto.
27366         (aarch64_xarqv2di): Ditto.
27367         (aarch64_bcaxqv8hi): Ditto.
27368         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27369         (vsha512h2q_u64): Ditto.
27370         (vsha512su0q_u64): Ditto.
27371         (vsha512su1q_u64): Ditto.
27372         (veor3q_u16): Ditto.
27373         (vrax1q_u64): Ditto.
27374         (vxarq_u64): Ditto.
27375         (vbcaxq_u16): Ditto.
27376         * config/arm/types.md (crypto_sha512): New type attribute.
27377         (crypto_sha3): Ditto.
27378         (doc/invoke.texi): Document new sha3 option.
27380 2018-01-10  Michael Collison  <michael.collison@arm.com>
27382         * config/aarch64/aarch64-builtins.c:
27383         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27384         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27385         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27386         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27387         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27388         (AARCH64_ISA_SM4): New ISA flag.
27389         (TARGET_SM4): New feature flag for sm4.
27390         * config/aarch64/aarch64-simd-builtins.def
27391         (aarch64_sm3ss1qv4si): Ditto.
27392         (aarch64_sm3tt1aq4si): Ditto.
27393         (aarch64_sm3tt1bq4si): Ditto.
27394         (aarch64_sm3tt2aq4si): Ditto.
27395         (aarch64_sm3tt2bq4si): Ditto.
27396         (aarch64_sm3partw1qv4si): Ditto.
27397         (aarch64_sm3partw2qv4si): Ditto.
27398         (aarch64_sm4eqv4si): Ditto.
27399         (aarch64_sm4ekeyqv4si): Ditto.
27400         * config/aarch64/aarch64-simd.md:
27401         (aarch64_sm3ss1qv4si): Ditto.
27402         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27403         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27404         (aarch64_sm4eqv4si): Ditto.
27405         (aarch64_sm4ekeyqv4si): Ditto.
27406         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27407         (sm3part_op): Ditto.
27408         (CRYPTO_SM3TT): Ditto.
27409         (CRYPTO_SM3PART): Ditto.
27410         (UNSPEC_SM3SS1): New unspec.
27411         (UNSPEC_SM3TT1A): Ditto.
27412         (UNSPEC_SM3TT1B): Ditto.
27413         (UNSPEC_SM3TT2A): Ditto.
27414         (UNSPEC_SM3TT2B): Ditto.
27415         (UNSPEC_SM3PARTW1): Ditto.
27416         (UNSPEC_SM3PARTW2): Ditto.
27417         (UNSPEC_SM4E): Ditto.
27418         (UNSPEC_SM4EKEY): Ditto.
27419         * config/aarch64/constraints.md (Ui2): New constraint.
27420         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27421         * config/arm/types.md (crypto_sm3): New type attribute.
27422         (crypto_sm4): Ditto.
27423         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27424         (vsm3tt1aq_u32): Ditto.
27425         (vsm3tt1bq_u32): Ditto.
27426         (vsm3tt2aq_u32): Ditto.
27427         (vsm3tt2bq_u32): Ditto.
27428         (vsm3partw1q_u32): Ditto.
27429         (vsm3partw2q_u32): Ditto.
27430         (vsm4eq_u32): Ditto.
27431         (vsm4ekeyq_u32): Ditto.
27432         (doc/invoke.texi): Document new sm4 option.
27434 2018-01-10  Michael Collison  <michael.collison@arm.com>
27436         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27437         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27438         (AARCH64_FL_FOR_ARCH8_4): New.
27439         (AARCH64_FL_V8_4): New flag.
27440         (doc/invoke.texi): Document new armv8.4-a option.
27442 2018-01-10  Michael Collison  <michael.collison@arm.com>
27444         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27445         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27446         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27447         * config/aarch64/aarch64-option-extension.def: Add
27448         AARCH64_OPT_EXTENSION of 'sha2'.
27449         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27450         (crypto): Disable sha2 and aes if crypto disabled.
27451         (crypto): Enable aes and sha2 if enabled.
27452         (simd): Disable sha2 and aes if simd disabled.
27453         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27454         New flags.
27455         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27456         (TARGET_SHA2): New feature flag for sha2.
27457         (TARGET_AES): New feature flag for aes.
27458         * config/aarch64/aarch64-simd.md:
27459         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27460         conditional on TARGET_AES.
27461         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27462         (aarch64_crypto_sha1hsi): Make pattern conditional
27463         on TARGET_SHA2.
27464         (aarch64_crypto_sha1hv4si): Ditto.
27465         (aarch64_be_crypto_sha1hv4si): Ditto.
27466         (aarch64_crypto_sha1su1v4si): Ditto.
27467         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27468         (aarch64_crypto_sha1su0v4si): Ditto.
27469         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27470         (aarch64_crypto_sha256su0v4si): Ditto.
27471         (aarch64_crypto_sha256su1v4si): Ditto.
27472         (doc/invoke.texi): Document new aes and sha2 options.
27474 2018-01-10  Martin Sebor  <msebor@redhat.com>
27476         PR tree-optimization/83781
27477         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27478         as string arrays.
27480 2018-01-11  Martin Sebor  <msebor@gmail.com>
27481             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27483         PR tree-optimization/83501
27484         PR tree-optimization/81703
27486         * tree-ssa-strlen.c (get_string_cst): Rename...
27487         (get_string_len): ...to this.  Handle global constants.
27488         (handle_char_store): Adjust.
27490 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
27491             Jim Wilson  <jimw@sifive.com>
27493         * config/riscv/riscv-protos.h (riscv_output_return): New.
27494         * config/riscv/riscv.c (struct machine_function): New naked_p field.
27495         (riscv_attribute_table, riscv_output_return),
27496         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27497         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27498         (riscv_compute_frame_info): Only compute frame->mask if not a naked
27499         function.
27500         (riscv_expand_prologue): Add early return for naked function.
27501         (riscv_expand_epilogue): Likewise.
27502         (riscv_function_ok_for_sibcall): Return false for naked function.
27503         (riscv_set_current_function): New.
27504         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27505         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27506         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
27507         * doc/extend.texi (RISC-V Function Attributes): New.
27509 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
27511         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
27512         check for 128-bit long double before checking TCmode.
27513         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
27514         128-bit long doubles before checking TFmode or TCmode.
27515         (FLOAT128_IBM_P): Likewise.
27517 2018-01-10  Martin Sebor  <msebor@redhat.com>
27519         PR tree-optimization/83671
27520         * builtins.c (c_strlen): Unconditionally return zero for the empty
27521         string.
27522         Use -Warray-bounds for warnings.
27523         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
27524         for non-constant array indices with COMPONENT_REF, arrays of
27525         arrays, and pointers to arrays.
27526         (gimple_fold_builtin_strlen): Determine and set length range for
27527         non-constant character arrays.
27529 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
27531         PR middle-end/81897
27532         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
27533         empty blocks.
27535 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
27537         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
27539 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27541         PR target/83399
27542         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
27543         VECTOR_MEM_ALTIVEC_OR_VSX_P.
27544         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
27545         indexed_or_indirect_operand predicate.
27546         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
27547         (*vsx_le_perm_load_v8hi): Likewise.
27548         (*vsx_le_perm_load_v16qi): Likewise.
27549         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
27550         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
27551         (*vsx_le_perm_store_v8hi): Likewise.
27552         (*vsx_le_perm_store_v16qi): Likewise.
27553         (eight unnamed splitters): Likewise.
27555 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27557         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
27558         * config/rs6000/emmintrin.h: Likewise.
27559         * config/rs6000/mmintrin.h: Likewise.
27560         * config/rs6000/xmmintrin.h: Likewise.
27562 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
27564         PR c++/43486
27565         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
27566         "public_flag".
27567         * tree.c (tree_nop_conversion): Return true for location wrapper
27568         nodes.
27569         (maybe_wrap_with_location): New function.
27570         (selftest::check_strip_nops): New function.
27571         (selftest::test_location_wrappers): New function.
27572         (selftest::tree_c_tests): Call it.
27573         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
27574         (maybe_wrap_with_location): New decl.
27575         (EXPR_LOCATION_WRAPPER_P): New macro.
27576         (location_wrapper_p): New inline function.
27577         (tree_strip_any_location_wrapper): New inline function.
27579 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
27581         PR target/83735
27582         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
27583         stack_realign_offset for the largest alignment of stack slot
27584         actually used.
27585         (ix86_find_max_used_stack_alignment): New function.
27586         (ix86_finalize_stack_frame_flags): Use it.  Set
27587         max_used_stack_alignment if we don't realign stack.
27588         * config/i386/i386.h (machine_function): Add
27589         max_used_stack_alignment.
27591 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
27593         * config/arm/arm.opt (-mbranch-cost): New option.
27594         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
27595         account.
27597 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
27599         PR target/83629
27600         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
27601         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
27603 2018-01-10  Richard Biener  <rguenther@suse.de>
27605         PR debug/83765
27606         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
27607         early out so it also covers the case where we have a non-NULL
27608         origin.
27610 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27612         PR tree-optimization/83753
27613         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
27614         for non-strided grouped accesses if the number of elements is 1.
27616 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27618         PR target/81616
27619         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
27620         * i386.h (TARGET_USE_GATHER): Define.
27621         * x86-tune.def (X86_TUNE_USE_GATHER): New.
27623 2018-01-10  Martin Liska  <mliska@suse.cz>
27625         PR bootstrap/82831
27626         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
27627         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
27628         partitioning.
27629         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
27630         CLEANUP_NO_PARTITIONING is not set.
27632 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27634         * doc/rtl.texi: Remove documentation of (const ...) wrappers
27635         for vectors, as a partial revert of r254296.
27636         * rtl.h (const_vec_p): Delete.
27637         (const_vec_duplicate_p): Don't test for vector CONSTs.
27638         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
27639         * expmed.c (make_tree): Likewise.
27641         Revert:
27642         * common.md (E, F): Use CONSTANT_P instead of checking for
27643         CONST_VECTOR.
27644         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
27645         checking for CONST_VECTOR.
27647 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27649         PR middle-end/83575
27650         * predict.c (force_edge_cold): Handle in more sane way edges
27651         with no prediction.
27653 2018-01-09  Carl Love  <cel@us.ibm.com>
27655         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
27656         V4SI, V4SF types.
27657         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
27658         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
27659         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
27660         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
27661         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
27662         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
27663         * config/rs6000/rs6000-protos.h: Add extern defition for
27664         rs6000_generate_float2_double_code.
27665         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
27666         function.
27667         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
27668         (float2_v2df): Add define_expand.
27670 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
27672         PR target/83628
27673         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
27674         op_mode in the force_to_mode call.
27676 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27678         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
27679         instead of checking each element individually.
27680         (aarch64_evpc_uzp): Likewise.
27681         (aarch64_evpc_zip): Likewise.
27682         (aarch64_evpc_ext): Likewise.
27683         (aarch64_evpc_rev): Likewise.
27684         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
27685         instead of checking each element individually.  Return true without
27686         generating rtl if
27687         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
27688         whether all selected elements come from the same input, instead of
27689         checking each element individually.  Remove calls to gen_rtx_REG,
27690         start_sequence and end_sequence and instead assert that no rtl is
27691         generated.
27693 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27695         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
27696         order of HIGH and CONST checks.
27698 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27700         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
27701         if the destination isn't an SSA_NAME.
27703 2018-01-09  Richard Biener  <rguenther@suse.de>
27705         PR tree-optimization/83668
27706         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
27707         move prologue...
27708         (canonicalize_loop_form): ... here, renamed from ...
27709         (canonicalize_loop_closed_ssa_form): ... this and amended to
27710         swap successor edges for loop exit blocks to make us use
27711         the RPO order we need for initial schedule generation.
27713 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
27715         PR tree-optimization/64811
27716         * match.pd: When optimizing comparisons with Inf, avoid
27717         introducing or losing exceptions from comparisons with NaN.
27719 2018-01-09  Martin Liska  <mliska@suse.cz>
27721         PR sanitizer/82517
27722         * asan.c (shadow_mem_size): Add gcc_assert.
27724 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
27726         Don't save registers in main().
27728         PR target/83738
27729         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
27730         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
27731         * config/avr/avr.c (avr_set_current_function): Don't error if
27732         naked, OS_task or OS_main are specified at the same time.
27733         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
27734         OS_main.
27735         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
27736         attribute.
27737         * common/config/avr/avr-common.c (avr_option_optimization_table):
27738         Switch on -mmain-is-OS_task for optimizing compilations.
27740 2018-01-09  Richard Biener  <rguenther@suse.de>
27742         PR tree-optimization/83572
27743         * graphite.c: Include cfganal.h.
27744         (graphite_transform_loops): Connect infinite loops to exit
27745         and remove fake edges at the end.
27747 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27749         * ipa-inline.c (edge_badness): Revert accidental checkin.
27751 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27753         PR ipa/80763
27754         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
27755         symbols; not inline clones.
27757 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
27759         PR target/83507
27760         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
27761         hard registers.  Formatting fixes.
27763         PR preprocessor/83722
27764         * gcc.c (try_generate_repro): Pass
27765         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
27766         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
27767         do_report_bug.
27769 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
27770             Kito Cheng  <kito.cheng@gmail.com>
27772         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
27773         (riscv_leaf_function_p): Delete.
27774         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
27776 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27778         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
27779         function.
27780         (do_ifelse): New function.
27781         (do_isel): New function.
27782         (do_sub3): New function.
27783         (do_add3): New function.
27784         (do_load_mask_compare): New function.
27785         (do_overlap_load_compare): New function.
27786         (expand_compare_loop): New function.
27787         (expand_block_compare): Call expand_compare_loop() when appropriate.
27788         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27789         option description.
27790         (-mblock-compare-inline-loop-limit): New option.
27792 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27794         PR target/83677
27795         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27796         Reverse order of second and third operands in first alternative.
27797         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27798         of first and second elements in UNSPEC_VPERMR vector.
27799         (altivec_expand_vec_perm_le): Likewise.
27801 2018-01-08  Jeff Law  <law@redhat.com>
27803         PR rtl-optimizatin/81308
27804         * tree-switch-conversion.c (cfg_altered): New file scoped static.
27805         (process_switch): If group_case_labels makes a change, then set
27806         cfg_altered.
27807         (pass_convert_switch::execute): If a switch is converted, then
27808         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
27810         PR rtl-optimization/81308
27811         * recog.c (split_all_insns): Conditionally cleanup the CFG after
27812         splitting insns.
27814 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
27816         PR target/83663 - Revert r255946
27817         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27818         generation for cases where splatting a value is not useful.
27819         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27820         across a vec_duplicate and a paradoxical subreg forming a vector
27821         mode to a vec_concat.
27823 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27825         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27826         -march=armv8.3-a variants.
27827         * config/arm/t-multilib: Likewise.
27828         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
27830 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27832         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27833         to generate rtl.
27834         (cceq_ior_compare_complement): Give it a name so I can use it, and
27835         change boolean_or_operator predicate to boolean_operator so it can
27836         be used to generate a crand.
27837         (eqne): New code iterator.
27838         (bd/bd_neg): New code_attrs.
27839         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27840         a single define_insn.
27841         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27842         decrement (bdnzt/bdnzf/bdzt/bdzf).
27843         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27844         with the new names of the branch decrement patterns, and added the
27845         names of the branch decrement conditional patterns.
27847 2018-01-08  Richard Biener  <rguenther@suse.de>
27849         PR tree-optimization/83563
27850         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27851         cache.
27853 2018-01-08  Richard Biener  <rguenther@suse.de>
27855         PR middle-end/83713
27856         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27858 2018-01-08  Richard Biener  <rguenther@suse.de>
27860         PR tree-optimization/83685
27861         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27862         references to abnormals.
27864 2018-01-08  Richard Biener  <rguenther@suse.de>
27866         PR lto/83719
27867         * dwarf2out.c (output_indirect_strings): Handle empty
27868         skeleton_debug_str_hash.
27869         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27871 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27873         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27874         (emit_store_direct): Likewise.
27875         (arc_trampoline_adjust_address): Likewise.
27876         (arc_asm_trampoline_template): New function.
27877         (arc_initialize_trampoline): Use asm_trampoline_template.
27878         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27879         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27880         * config/arc/arc.md (flush_icache): Delete pattern.
27882 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27884         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27885         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27886         munaligned-access.
27888 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27890         PR target/83681
27891         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27892         by not USED_FOR_TARGET.
27893         (make_pass_resolve_sw_modes): Likewise.
27895 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27897         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27898         USED_FOR_TARGET.
27900 2018-01-08  Richard Biener  <rguenther@suse.de>
27902         PR middle-end/83580
27903         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27905 2018-01-08  Richard Biener  <rguenther@suse.de>
27907         PR middle-end/83517
27908         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27910 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27912         PR middle-end/81897
27913         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27914         basic blocks with a small number of successors.
27915         (convert_control_dep_chain_into_preds): Improve handling of
27916         forwarder blocks.
27917         (dump_predicates): Split apart into...
27918         (dump_pred_chain): ...here...
27919         (dump_pred_info): ...and here.
27920         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27921         (can_chain_union_be_invalidated_p): Improve check for invalidation
27922         of paths.
27923         (uninit_uses_cannot_happen): Avoid unnecessary if
27924         convert_control_dep_chain_into_preds yielded nothing.
27926 2018-01-06  Martin Sebor  <msebor@redhat.com>
27928         PR tree-optimization/83640
27929         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27930         subtracting negative offset from size.
27931         (builtin_access::overlap): Adjust offset bounds of the access to fall
27932         within the size of the object if possible.
27934 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27936         PR rtl-optimization/83699
27937         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27938         extract_bit_field_as_subreg to cases in which the extracted
27939         value is also a vector.
27941         * lra-constraints.c (process_alt_operands): Test for the equivalence
27942         substitutions when detecting a possible reload cycle.
27944 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
27946         PR debug/83480
27947         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27948         by default if flag_selective_schedling{,2}.  Formatting fixes.
27950         PR rtl-optimization/83682
27951         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27952         if it has non-VECTOR_MODE element mode.
27953         (vec_duplicate_p): Likewise.
27955         PR middle-end/83694
27956         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27957         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27959 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27961         PR target/83604
27962         * config/i386/i386-builtin.def
27963         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27964         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27965         Require also OPTION_MASK_ISA_AVX512F in addition to
27966         OPTION_MASK_ISA_GFNI.
27967         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27968         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27969         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27970         to OPTION_MASK_ISA_GFNI.
27971         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27972         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27973         OPTION_MASK_ISA_AVX512BW.
27974         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27975         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27976         addition to OPTION_MASK_ISA_GFNI.
27977         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27978         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27979         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27980         to OPTION_MASK_ISA_GFNI.
27981         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27982         a requirement for all ISAs rather than any of them with a few
27983         exceptions.
27984         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27985         processing.
27986         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27987         bitmasks to be enabled with 3 exceptions, instead of requiring any
27988         enabled ISA with lots of exceptions.
27989         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27990         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27991         Change avx512bw in isa attribute to avx512f.
27992         * config/i386/sgxintrin.h: Add license boilerplate.
27993         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
27994         to __AVX512F__ and __AVX512VL to __AVX512VL__.
27995         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27996         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27997         defined.
27998         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27999         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28000         temporarily sse2 rather than sse if not enabled already.
28002         PR target/83604
28003         * config/i386/sse.md (VI248_VLBW): Rename to ...
28004         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28005         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28006         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28007         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28008         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28009         mode iterator instead of VI248_VLBW.
28011 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28013         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28014         (record_modified): Skip clobbers; add debug output.
28015         (param_change_prob): Use sreal frequencies.
28017 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28019         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28020         punt for user-aligned variables.
28022 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28024         * tree-chrec.c (chrec_contains_symbols): Return true for
28025         POLY_INT_CST.
28027 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28029         PR target/82439
28030         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28031         of (x|y) == x for BICS pattern.
28033 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28035         PR tree-optimization/83605
28036         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28037         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28038         can throw.
28040 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28042         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28043         * config/epiphany/rtems.h: New file.
28045 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28046             Uros Bizjak  <ubizjak@gmail.com>
28048         PR target/83554
28049         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28050         QIreg_operand instead of register_operand predicate.
28051         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28052         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28053         comments instead of -fmitigate[-_]rop.
28055 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28057         PR bootstrap/81926
28058         * cgraphunit.c (symbol_table::compile): Switch to text_section
28059         before calling assembly_start debug hook.
28060         * run-rtl-passes.c (run_rtl_passes): Likewise.
28061         Include output.h.
28063 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28065         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28066         range_int_cst_p rather than !symbolic_range_p before calling
28067         extract_range_from_multiplicative_op_1.
28069 2018-01-04  Jeff Law  <law@redhat.com>
28071         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28072         redundant test in assertion.
28074 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28076         * doc/rtl.texi: Document machine_mode wrapper classes.
28078 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28080         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28081         using tree_to_uhwi.
28083 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28085         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28086         the VEC_PERM_EXPR fold to fail.
28088 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28090         PR debug/83585
28091         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28092         to switched_sections.
28094 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28096         PR target/83680
28097         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28098         test for d.testing.
28100 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28102         PR target/83387
28103         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28104         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28106 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28108         PR debug/83666
28109         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28110         is BLKmode and bitpos not zero or mode change is needed.
28112 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28114         PR target/83675
28115         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28116         TARGET_VIS2.
28118 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28120         PR target/83628
28121         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28122         instead of MULT rtx.  Update all corresponding splitters.
28123         (*saddl_se): Ditto.
28124         (*ssub<modesuffix>): Ditto.
28125         (*ssubl_se): Ditto.
28126         (*cmp_sadd_di): Update split patterns.
28127         (*cmp_sadd_si): Ditto.
28128         (*cmp_sadd_sidi): Ditto.
28129         (*cmp_ssub_di): Ditto.
28130         (*cmp_ssub_si): Ditto.
28131         (*cmp_ssub_sidi): Ditto.
28132         * config/alpha/predicates.md (const23_operand): New predicate.
28133         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28134         Look for ASHIFT, not MULT inner operand.
28135         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28137 2018-01-04  Martin Liska  <mliska@suse.cz>
28139         PR gcov-profile/83669
28140         * gcov.c (output_intermediate_file): Add version to intermediate
28141         gcov file.
28142         * doc/gcov.texi: Document new field 'version' in intermediate
28143         file format. Fix location of '-k' option of gcov command.
28145 2018-01-04  Martin Liska  <mliska@suse.cz>
28147         PR ipa/82352
28148         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28150 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28152         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28154 2018-01-03  Martin Sebor  <msebor@redhat.com>
28156         PR tree-optimization/83655
28157         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28158         checking calls with invalid arguments.
28160 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28162         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28163         (vectorizable_mask_load_store): Delete.
28164         (vectorizable_call): Return false for masked loads and stores.
28165         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28166         instead of gimple_assign_rhs1.
28167         (vectorizable_load): Handle IFN_MASK_LOAD.
28168         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28170 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28172         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28173         split out from..,
28174         (vectorizable_mask_load_store): ...here.
28175         (vectorizable_load): ...and here.
28177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28179         * tree-vect-stmts.c (vect_build_all_ones_mask)
28180         (vect_build_zero_merge_argument): New functions, split out from...
28181         (vectorizable_load): ...here.
28183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28185         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28186         split out from...
28187         (vectorizable_mask_load_store): ...here.
28188         (vectorizable_store): ...and here.
28190 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28192         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28193         split out from...
28194         (vectorizable_mask_load_store): ...here.
28196 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28198         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28199         (vect_model_store_cost): Take a vec_load_store_type instead of a
28200         vect_def_type.
28201         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28202         (vect_model_store_cost): Take a vec_load_store_type instead of a
28203         vect_def_type.
28204         (vectorizable_mask_load_store): Update accordingly.
28205         (vectorizable_store): Likewise.
28206         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28208 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28210         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28211         IFN_MASK_LOAD calls here rather than...
28212         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28214 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28215             Alan Hayward  <alan.hayward@arm.com>
28216             David Sherwood  <david.sherwood@arm.com>
28218         * expmed.c (extract_bit_field_1): For vector extracts,
28219         fall back to extract_bit_field_as_subreg if vec_extract
28220         isn't available.
28222 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28223             Alan Hayward  <alan.hayward@arm.com>
28224             David Sherwood  <david.sherwood@arm.com>
28226         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28227         they are variable or constant sized.
28228         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28229         slots for constant-sized data.
28231 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28232             Alan Hayward  <alan.hayward@arm.com>
28233             David Sherwood  <david.sherwood@arm.com>
28235         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28236         handling COND_EXPRs with boolean comparisons, try to find a better
28237         basis for the mask type than the boolean itself.
28239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28241         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28242         is calculated and how it can be overridden.
28243         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28244         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28245         if defined.
28246         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28247         if nonzero.
28249 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28250             Alan Hayward  <alan.hayward@arm.com>
28251             David Sherwood  <david.sherwood@arm.com>
28253         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28254         Remove the mode argument.
28255         (aarch64_simd_valid_immediate): Remove the mode and inverse
28256         arguments.
28257         * config/aarch64/iterators.md (bitsize): New iterator.
28258         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28259         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28260         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28261         aarch64_simd_valid_immediate.
28262         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28263         (aarch64_reg_or_bic_imm): Likewise.
28264         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28265         with an insn_type enum and msl with a modifier_type enum.
28266         Replace element_width with a scalar_mode.  Change the shift
28267         to unsigned int.  Add constructors for scalar_float_mode and
28268         scalar_int_mode elements.
28269         (aarch64_vect_float_const_representable_p): Delete.
28270         (aarch64_can_const_movi_rtx_p)
28271         (aarch64_simd_scalar_immediate_valid_for_move)
28272         (aarch64_simd_make_constant): Update call to
28273         aarch64_simd_valid_immediate.
28274         (aarch64_advsimd_valid_immediate_hs): New function.
28275         (aarch64_advsimd_valid_immediate): Likewise.
28276         (aarch64_simd_valid_immediate): Remove mode and inverse
28277         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28278         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28279         and aarch64_float_const_representable_p on the result.
28280         (aarch64_output_simd_mov_immediate): Remove mode argument.
28281         Update call to aarch64_simd_valid_immediate and use of
28282         simd_immediate_info.
28283         (aarch64_output_scalar_simd_mov_immediate): Update call
28284         accordingly.
28286 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28287             Alan Hayward  <alan.hayward@arm.com>
28288             David Sherwood  <david.sherwood@arm.com>
28290         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28291         (mode_nunits): Likewise CONST_MODE_NUNITS.
28292         * machmode.def (ADJUST_NUNITS): Document.
28293         * genmodes.c (mode_data::need_nunits_adj): New field.
28294         (blank_mode): Update accordingly.
28295         (adj_nunits): New variable.
28296         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28297         parameter.
28298         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28299         listed in adj_nunits.
28300         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28301         listed in adj_nunits.  Don't emit case statements for such modes.
28302         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28303         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28304         nothing if adj_nunits is nonnull.
28305         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28306         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28307         (emit_mode_fbit): Update use of print_maybe_const_decl.
28308         (emit_move_size): Likewise.  Treat the array as non-const
28309         if adj_nunits.
28310         (emit_mode_adjustments): Handle adj_nunits.
28312 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28314         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28315         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28316         (VECTOR_MODES): Use it.
28317         (make_vector_modes): Take the prefix as an argument.
28319 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28320             Alan Hayward  <alan.hayward@arm.com>
28321             David Sherwood  <david.sherwood@arm.com>
28323         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28324         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28325         for MODE_VECTOR_BOOL.
28326         * machmode.def (VECTOR_BOOL_MODE): Document.
28327         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28328         (make_vector_bool_mode): New function.
28329         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28330         MODE_VECTOR_BOOL.
28331         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28332         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28333         Likewise.
28334         * stor-layout.c (int_mode_for_mode): Likewise.
28335         * tree.c (build_vector_type_for_mode): Likewise.
28336         * varasm.c (output_constant_pool_2): Likewise.
28337         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28338         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28339         for MODE_VECTOR_BOOL.
28340         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28341         of mode class checks.
28342         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28343         instead of a list of mode class checks.
28344         (expand_vector_scalar_condition): Likewise.
28345         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28347 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28348             Alan Hayward  <alan.hayward@arm.com>
28349             David Sherwood  <david.sherwood@arm.com>
28351         * machmode.h (mode_size): Change from unsigned short to
28352         poly_uint16_pod.
28353         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28354         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28355         or if measurement_type is not polynomial.
28356         (fixed_size_mode::includes_p): Check for constant-sized modes.
28357         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28358         return a poly_uint16 rather than an unsigned short.
28359         (emit_mode_size): Change the type of mode_size from unsigned short
28360         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28361         (emit_mode_adjustments): Cope with polynomial vector sizes.
28362         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28363         for GET_MODE_SIZE.
28364         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28365         for GET_MODE_SIZE.
28366         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28367         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28368         * caller-save.c (setup_save_areas): Likewise.
28369         (replace_reg_with_saved_mem): Likewise.
28370         * calls.c (emit_library_call_value_1): Likewise.
28371         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28372         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28373         (gen_lowpart_for_combine): Likewise.
28374         * convert.c (convert_to_integer_1): Likewise.
28375         * cse.c (equiv_constant, cse_insn): Likewise.
28376         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28377         (cselib_subst_to_values): Likewise.
28378         * dce.c (word_dce_process_block): Likewise.
28379         * df-problems.c (df_word_lr_mark_ref): Likewise.
28380         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28381         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28382         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28383         (rtl_for_decl_location): Likewise.
28384         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28385         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28386         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28387         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28388         (expand_expr_real_1): Likewise.
28389         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28390         (pad_below): Likewise.
28391         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28392         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28393         * ira.c (get_subreg_tracking_sizes): Likewise.
28394         * ira-build.c (ira_create_allocno_objects): Likewise.
28395         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28396         (ira_sort_regnos_for_alter_reg): Likewise.
28397         * ira-costs.c (record_operand_costs): Likewise.
28398         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28399         (resolve_simple_move): Likewise.
28400         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28401         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28402         (lra_constraints): Likewise.
28403         (CONST_POOL_OK_P): Reject variable-sized modes.
28404         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28405         (add_pseudo_to_slot, lra_spill): Likewise.
28406         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28407         * optabs-query.c (get_best_extraction_insn): Likewise.
28408         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28409         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28410         (expand_mult_highpart, valid_multiword_target_p): Likewise.
28411         * recog.c (offsettable_address_addr_space_p): Likewise.
28412         * regcprop.c (maybe_mode_change): Likewise.
28413         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28414         * regrename.c (build_def_use): Likewise.
28415         * regstat.c (dump_reg_info): Likewise.
28416         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28417         (find_reloads, find_reloads_subreg_address): Likewise.
28418         * reload1.c (eliminate_regs_1): Likewise.
28419         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28420         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28421         (simplify_binary_operation_1, simplify_subreg): Likewise.
28422         * targhooks.c (default_function_arg_padding): Likewise.
28423         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28424         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28425         (verify_gimple_assign_ternary): Likewise.
28426         * tree-inline.c (estimate_move_cost): Likewise.
28427         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28428         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28429         (get_address_cost_ainc): Likewise.
28430         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28431         (vect_supportable_dr_alignment): Likewise.
28432         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28433         (vectorizable_reduction): Likewise.
28434         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28435         (vectorizable_operation, vectorizable_load): Likewise.
28436         * tree.c (build_same_sized_truth_vector_type): Likewise.
28437         * valtrack.c (cleanup_auto_inc_dec): Likewise.
28438         * var-tracking.c (emit_note_insn_var_location): Likewise.
28439         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28440         (ADDR_VEC_ALIGN): Likewise.
28442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28443             Alan Hayward  <alan.hayward@arm.com>
28444             David Sherwood  <david.sherwood@arm.com>
28446         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28447         unsigned short.
28448         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28449         or if measurement_type is polynomial.
28450         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28451         * combine.c (make_extraction): Likewise.
28452         * dse.c (find_shift_sequence): Likewise.
28453         * dwarf2out.c (mem_loc_descriptor): Likewise.
28454         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28455         (extract_bit_field, extract_low_bits): Likewise.
28456         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28457         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28458         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28459         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28460         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28461         * reload.c (find_reloads): Likewise.
28462         * reload1.c (alter_reg): Likewise.
28463         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28464         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28465         * tree-if-conv.c (predicate_mem_writes): Likewise.
28466         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28467         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28468         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28469         * valtrack.c (dead_debug_insert_temp): Likewise.
28470         * varasm.c (mergeable_constant_section): Likewise.
28471         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28473 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28474             Alan Hayward  <alan.hayward@arm.com>
28475             David Sherwood  <david.sherwood@arm.com>
28477         * expr.c (expand_assignment): Cope with polynomial mode sizes
28478         when assigning to a CONCAT.
28480 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28481             Alan Hayward  <alan.hayward@arm.com>
28482             David Sherwood  <david.sherwood@arm.com>
28484         * machmode.h (mode_precision): Change from unsigned short to
28485         poly_uint16_pod.
28486         (mode_to_precision): Return a poly_uint16 rather than an unsigned
28487         short.
28488         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28489         or if measurement_type is not polynomial.
28490         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
28491         in which the mode is already known to be a scalar_int_mode.
28492         * genmodes.c (emit_mode_precision): Change the type of mode_precision
28493         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
28494         initializer.
28495         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28496         for GET_MODE_PRECISION.
28497         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28498         for GET_MODE_PRECISION.
28499         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28500         as polynomial.
28501         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28502         (expand_field_assignment, make_extraction): Likewise.
28503         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28504         (get_last_value): Likewise.
28505         * convert.c (convert_to_integer_1): Likewise.
28506         * cse.c (cse_insn): Likewise.
28507         * expr.c (expand_expr_real_1): Likewise.
28508         * lra-constraints.c (simplify_operand_subreg): Likewise.
28509         * optabs-query.c (can_atomic_load_p): Likewise.
28510         * optabs.c (expand_atomic_load): Likewise.
28511         (expand_atomic_store): Likewise.
28512         * ree.c (combine_reaching_defs): Likewise.
28513         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
28514         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
28515         * tree.h (type_has_mode_precision_p): Likewise.
28516         * ubsan.c (instrument_si_overflow): Likewise.
28518 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28519             Alan Hayward  <alan.hayward@arm.com>
28520             David Sherwood  <david.sherwood@arm.com>
28522         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
28523         polynomial numbers of units.
28524         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
28525         (valid_vector_subparts_p): New function.
28526         (build_vector_type): Remove temporary shim and take the number
28527         of units as a poly_uint64 rather than an int.
28528         (build_opaque_vector_type): Take the number of units as a
28529         poly_uint64 rather than an int.
28530         * tree.c (build_vector_from_ctor): Handle polynomial
28531         TYPE_VECTOR_SUBPARTS.
28532         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
28533         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
28534         (build_vector_from_val): If the number of units is variable,
28535         use build_vec_duplicate_cst for constant operands and
28536         VEC_DUPLICATE_EXPR otherwise.
28537         (make_vector_type): Remove temporary is_constant ().
28538         (build_vector_type, build_opaque_vector_type): Take the number of
28539         units as a poly_uint64 rather than an int.
28540         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
28541         VECTOR_CST_NELTS.
28542         * cfgexpand.c (expand_debug_expr): Likewise.
28543         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
28544         (store_constructor, expand_expr_real_1): Likewise.
28545         (const_scalar_mask_from_tree): Likewise.
28546         * fold-const-call.c (fold_const_reduction): Likewise.
28547         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
28548         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
28549         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
28550         (fold_relational_const): Likewise.
28551         (native_interpret_vector): Likewise.  Change the size from an
28552         int to an unsigned int.
28553         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
28554         TYPE_VECTOR_SUBPARTS.
28555         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
28556         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
28557         duplicating a non-constant operand into a variable-length vector.
28558         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
28559         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
28560         * ipa-icf.c (sem_variable::equals): Likewise.
28561         * match.pd: Likewise.
28562         * omp-simd-clone.c (simd_clone_subparts): Likewise.
28563         * print-tree.c (print_node): Likewise.
28564         * stor-layout.c (layout_type): Likewise.
28565         * targhooks.c (default_builtin_vectorization_cost): Likewise.
28566         * tree-cfg.c (verify_gimple_comparison): Likewise.
28567         (verify_gimple_assign_binary): Likewise.
28568         (verify_gimple_assign_ternary): Likewise.
28569         (verify_gimple_assign_single): Likewise.
28570         * tree-pretty-print.c (dump_generic_node): Likewise.
28571         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28572         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
28573         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
28574         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
28575         (vect_shift_permute_load_chain): Likewise.
28576         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
28577         (expand_vector_condition, optimize_vector_constructor): Likewise.
28578         (lower_vec_perm, get_compute_type): Likewise.
28579         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28580         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
28581         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
28582         (vect_recog_mask_conversion_pattern): Likewise.
28583         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
28584         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
28585         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28586         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
28587         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
28588         (vectorizable_shift, vectorizable_operation, vectorizable_store)
28589         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
28590         (supportable_widening_operation): Likewise.
28591         (supportable_narrowing_operation): Likewise.
28592         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
28593         Likewise.
28594         * varasm.c (output_constant): Likewise.
28596 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28597             Alan Hayward  <alan.hayward@arm.com>
28598             David Sherwood  <david.sherwood@arm.com>
28600         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
28601         so that both the length == 3 and length != 3 cases set up their
28602         own permute vectors.  Add comments explaining why we know the
28603         number of elements is constant.
28604         (vect_permute_load_chain): Likewise.
28606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28607             Alan Hayward  <alan.hayward@arm.com>
28608             David Sherwood  <david.sherwood@arm.com>
28610         * machmode.h (mode_nunits): Change from unsigned char to
28611         poly_uint16_pod.
28612         (ONLY_FIXED_SIZE_MODES): New macro.
28613         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
28614         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
28615         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
28616         New typedefs.
28617         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
28618         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
28619         or if measurement_type is not polynomial.
28620         * genmodes.c (ZERO_COEFFS): New macro.
28621         (emit_mode_nunits_inline): Make mode_nunits_inline return a
28622         poly_uint16.
28623         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
28624         Use ZERO_COEFFS when emitting initializers.
28625         * data-streamer.h (bp_pack_poly_value): New function.
28626         (bp_unpack_poly_value): Likewise.
28627         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28628         for GET_MODE_NUNITS.
28629         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28630         for GET_MODE_NUNITS.
28631         * tree.c (make_vector_type): Remove temporary shim and make
28632         the real function take the number of units as a poly_uint64
28633         rather than an int.
28634         (build_vector_type_for_mode): Handle polynomial nunits.
28635         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
28636         * emit-rtl.c (const_vec_series_p_1): Likewise.
28637         (gen_rtx_CONST_VECTOR): Likewise.
28638         * fold-const.c (test_vec_duplicate_folding): Likewise.
28639         * genrecog.c (validate_pattern): Likewise.
28640         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
28641         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28642         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
28643         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
28644         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
28645         * rtlanal.c (subreg_get_info): Likewise.
28646         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28647         (vect_grouped_load_supported): Likewise.
28648         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
28649         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
28650         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
28651         (simplify_const_unary_operation, simplify_binary_operation_1)
28652         (simplify_const_binary_operation, simplify_ternary_operation)
28653         (test_vector_ops_duplicate, test_vector_ops): Likewise.
28654         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
28655         instead of CONST_VECTOR_NUNITS.
28656         * varasm.c (output_constant_pool_2): Likewise.
28657         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
28658         explicit-encoded elements in the XVEC for variable-length vectors.
28660 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28662         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
28664 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28665             Alan Hayward  <alan.hayward@arm.com>
28666             David Sherwood  <david.sherwood@arm.com>
28668         * coretypes.h (fixed_size_mode): Declare.
28669         (fixed_size_mode_pod): New typedef.
28670         * builtins.h (target_builtins::x_apply_args_mode)
28671         (target_builtins::x_apply_result_mode): Change type to
28672         fixed_size_mode_pod.
28673         * builtins.c (apply_args_size, apply_result_size, result_vector)
28674         (expand_builtin_apply_args_1, expand_builtin_apply)
28675         (expand_builtin_return): Update accordingly.
28677 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28679         * cse.c (hash_rtx_cb): Hash only the encoded elements.
28680         * cselib.c (cselib_hash_rtx): Likewise.
28681         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
28682         CONST_VECTOR encoding.
28684 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28685             Jeff Law  <law@redhat.com>
28687         PR target/83641
28688         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
28689         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
28690         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
28691         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
28693         PR target/83641
28694         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
28695         explicitly probe *sp in a noreturn function if there were any callee
28696         register saves or frame pointer is needed.
28698 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28700         PR debug/83621
28701         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
28702         BLKmode for ternary, binary or unary expressions.
28704         PR debug/83645
28705         * var-tracking.c (delete_vta_debug_insn): New inline function.
28706         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
28707         insns from get_insns () to NULL instead of each bb separately.
28708         Use delete_vta_debug_insn.  No longer static.
28709         (vt_debug_insns_local, variable_tracking_main_1): Adjust
28710         delete_vta_debug_insns callers.
28711         * rtl.h (delete_vta_debug_insns): Declare.
28712         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
28713         instead of variable_tracking_main.
28715 2018-01-03  Martin Sebor  <msebor@redhat.com>
28717         PR tree-optimization/83603
28718         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
28719         arguments past the endof the argument list in functions declared
28720         without a prototype.
28721         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
28722         Avoid checking when arguments are null.
28724 2018-01-03  Martin Sebor  <msebor@redhat.com>
28726         PR c/83559
28727         * doc/extend.texi (attribute const): Fix a typo.
28728         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
28729         issuing -Wsuggest-attribute for void functions.
28731 2018-01-03  Martin Sebor  <msebor@redhat.com>
28733         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
28734         offset_int::from instead of wide_int::to_shwi.
28735         (maybe_diag_overlap): Remove assertion.
28736         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
28737         * gimple-ssa-sprintf.c (format_directive): Same.
28738         (parse_directive): Same.
28739         (sprintf_dom_walker::compute_format_length): Same.
28740         (try_substitute_return_value): Same.
28742 2018-01-03  Jeff Law  <law@redhat.com>
28744         PR middle-end/83654
28745         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
28746         non-constant residual for zero at runtime and avoid probing in
28747         that case.  Reorganize code for trailing problem to mirror handling
28748         of the residual.
28750 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28752         PR tree-optimization/83501
28753         * tree-ssa-strlen.c (get_string_cst): New.
28754         (handle_char_store): Call get_string_cst.
28756 2018-01-03  Martin Liska  <mliska@suse.cz>
28758         PR tree-optimization/83593
28759         * tree-ssa-strlen.c: Include tree-cfg.h.
28760         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
28761         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
28762         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
28763         to false.
28764         (strlen_dom_walker::before_dom_children): Call
28765         gimple_purge_dead_eh_edges. Dump tranformation with details
28766         dump flags.
28767         (strlen_dom_walker::before_dom_children): Update call by adding
28768         new argument cleanup_eh.
28769         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
28771 2018-01-03  Martin Liska  <mliska@suse.cz>
28773         PR ipa/83549
28774         * cif-code.def (VARIADIC_THUNK): New enum value.
28775         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
28776         thunks.
28778 2018-01-03  Jan Beulich  <jbeulich@suse.com>
28780         * sse.md (mov<mode>_internal): Tighten condition for when to use
28781         vmovdqu<ssescalarsize> for TI and OI modes.
28783 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28785         Update copyright years.
28787 2018-01-03  Martin Liska  <mliska@suse.cz>
28789         PR ipa/83594
28790         * ipa-visibility.c (function_and_variable_visibility): Skip
28791         functions with noipa attribure.
28793 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28795         * gcc.c (process_command): Update copyright notice dates.
28796         * gcov-dump.c (print_version): Ditto.
28797         * gcov.c (print_version): Ditto.
28798         * gcov-tool.c (print_version): Ditto.
28799         * gengtype.c (create_file): Ditto.
28800         * doc/cpp.texi: Bump @copying's copyright year.
28801         * doc/cppinternals.texi: Ditto.
28802         * doc/gcc.texi: Ditto.
28803         * doc/gccint.texi: Ditto.
28804         * doc/gcov.texi: Ditto.
28805         * doc/install.texi: Ditto.
28806         * doc/invoke.texi: Ditto.
28808 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28810         * vector-builder.h (vector_builder::m_full_nelts): Change from
28811         unsigned int to poly_uint64.
28812         (vector_builder::full_nelts): Update prototype accordingly.
28813         (vector_builder::new_vector): Likewise.
28814         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28815         (vector_builder::operator ==): Likewise.
28816         (vector_builder::finalize): Likewise.
28817         * int-vector-builder.h (int_vector_builder::int_vector_builder):
28818         Take the number of elements as a poly_uint64 rather than an
28819         unsigned int.
28820         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28821         from unsigned int to poly_uint64.
28822         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28823         (vec_perm_indices::new_vector): Likewise.
28824         (vec_perm_indices::length): Likewise.
28825         (vec_perm_indices::nelts_per_input): Likewise.
28826         (vec_perm_indices::input_nelts): Likewise.
28827         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28828         number of elements per input as a poly_uint64 rather than an
28829         unsigned int.  Use the original encoding for variable-length
28830         vectors, rather than clamping each individual element.
28831         For the second and subsequent elements in each pattern,
28832         clamp the step and base before clamping their sum.
28833         (vec_perm_indices::series_p): Handle polynomial element counts.
28834         (vec_perm_indices::all_in_range_p): Likewise.
28835         (vec_perm_indices_to_tree): Likewise.
28836         (vec_perm_indices_to_rtx): Likewise.
28837         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28838         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28839         (tree_vector_builder::new_binary_operation): Handle polynomial
28840         element counts.  Return false if we need to know the number
28841         of elements at compile time.
28842         * fold-const.c (fold_vec_perm): Punt if the number of elements
28843         isn't known at compile time.
28845 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28847         * vec-perm-indices.h (vec_perm_builder): Change element type
28848         from HOST_WIDE_INT to poly_int64.
28849         (vec_perm_indices::element_type): Update accordingly.
28850         (vec_perm_indices::clamp): Handle polynomial element_types.
28851         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28852         (vec_perm_indices::all_in_range_p): Likewise.
28853         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28854         than shwi trees.
28855         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28856         polynomial vec_perm_indices element types.
28857         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28858         * fold-const.c (fold_vec_perm): Likewise.
28859         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28860         * tree-vect-generic.c (lower_vec_perm): Likewise.
28861         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28862         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28863         element type to HOST_WIDE_INT.
28865 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28866             Alan Hayward  <alan.hayward@arm.com>
28867             David Sherwood  <david.sherwood@arm.com>
28869         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28870         rather than an int.  Use plus_constant.
28871         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28872         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28874 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28875             Alan Hayward  <alan.hayward@arm.com>
28876             David Sherwood  <david.sherwood@arm.com>
28878         * calls.c (emit_call_1, expand_call): Change struct_value_size from
28879         a HOST_WIDE_INT to a poly_int64.
28881 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28882             Alan Hayward  <alan.hayward@arm.com>
28883             David Sherwood  <david.sherwood@arm.com>
28885         * calls.c (load_register_parameters): Cope with polynomial
28886         mode sizes.  Require a constant size for BLKmode parameters
28887         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
28888         forces a parameter to be padded at the lsb end in order to
28889         fill a complete number of words, require the parameter size
28890         to be ordered wrt UNITS_PER_WORD.
28892 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28893             Alan Hayward  <alan.hayward@arm.com>
28894             David Sherwood  <david.sherwood@arm.com>
28896         * reload1.c (spill_stack_slot_width): Change element type
28897         from unsigned int to poly_uint64_pod.
28898         (alter_reg): Treat mode sizes as polynomial.
28900 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28901             Alan Hayward  <alan.hayward@arm.com>
28902             David Sherwood  <david.sherwood@arm.com>
28904         * reload.c (complex_word_subreg_p): New function.
28905         (reload_inner_reg_of_subreg, push_reload): Use it.
28907 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28908             Alan Hayward  <alan.hayward@arm.com>
28909             David Sherwood  <david.sherwood@arm.com>
28911         * lra-constraints.c (process_alt_operands): Reject matched
28912         operands whose sizes aren't ordered.
28913         (match_reload): Refer to this check here.
28915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28916             Alan Hayward  <alan.hayward@arm.com>
28917             David Sherwood  <david.sherwood@arm.com>
28919         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28920         that the mode size is in the set {1, 2, 4, 8, 16}.
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         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28927         Use plus_constant instead of gen_rtx_PLUS.
28929 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28930             Alan Hayward  <alan.hayward@arm.com>
28931             David Sherwood  <david.sherwood@arm.com>
28933         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28934         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28935         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28936         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28937         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28938         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28939         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28940         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28941         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28942         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28943         a poly_int64.
28944         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28945         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28946         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28947         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28948         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28949         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28950         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28951         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28952         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28953         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28954         function.
28955         * expr.c (emit_move_resolve_push): Treat the input and result
28956         of PUSH_ROUNDING as a poly_int64.
28957         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28958         (emit_push_insn): Likewise.
28959         * lra-eliminations.c (mark_not_eliminable): Likewise.
28960         * recog.c (push_operand): Likewise.
28961         * reload1.c (elimination_effects): Likewise.
28962         * rtlanal.c (nonzero_bits1): Likewise.
28963         * calls.c (store_one_arg): Likewise.  Require the padding to be
28964         known at compile time.
28966 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28967             Alan Hayward  <alan.hayward@arm.com>
28968             David Sherwood  <david.sherwood@arm.com>
28970         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28971         Use plus_constant instead of gen_rtx_PLUS.
28973 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28974             Alan Hayward  <alan.hayward@arm.com>
28975             David Sherwood  <david.sherwood@arm.com>
28977         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28978         rather than an int.
28980 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28981             Alan Hayward  <alan.hayward@arm.com>
28982             David Sherwood  <david.sherwood@arm.com>
28984         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28985         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28986         via stack temporaries.  Treat the mode size as polynomial too.
28988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28989             Alan Hayward  <alan.hayward@arm.com>
28990             David Sherwood  <david.sherwood@arm.com>
28992         * expr.c (expand_expr_real_2): When handling conversions involving
28993         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28994         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
28995         as a poly_uint64 too.
28997 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28998             Alan Hayward  <alan.hayward@arm.com>
28999             David Sherwood  <david.sherwood@arm.com>
29001         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29003 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29004             Alan Hayward  <alan.hayward@arm.com>
29005             David Sherwood  <david.sherwood@arm.com>
29007         * combine.c (can_change_dest_mode): Handle polynomial
29008         REGMODE_NATURAL_SIZE.
29009         * expmed.c (store_bit_field_1): Likewise.
29010         * expr.c (store_constructor): Likewise.
29011         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29012         and polynomial REGMODE_NATURAL_SIZE.
29013         (gen_lowpart_common): Likewise.
29014         * reginfo.c (record_subregs_of_mode): Likewise.
29015         * rtlanal.c (read_modify_subreg_p): Likewise.
29017 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29018             Alan Hayward  <alan.hayward@arm.com>
29019             David Sherwood  <david.sherwood@arm.com>
29021         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29022         numbers of elements.
29024 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29025             Alan Hayward  <alan.hayward@arm.com>
29026             David Sherwood  <david.sherwood@arm.com>
29028         * match.pd: Cope with polynomial numbers of vector elements.
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         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29035         in a POINTER_PLUS_EXPR.
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         * omp-simd-clone.c (simd_clone_subparts): New function.
29042         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29043         (ipa_simd_modify_function_body): Likewise.
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         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29050         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29051         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29052         (expand_vector_condition, vector_element): Likewise.
29053         (subparts_gt): New function.
29054         (get_compute_type): Use subparts_gt.
29055         (count_type_subparts): Delete.
29056         (expand_vector_operations_1): Use subparts_gt instead of
29057         count_type_subparts.
29059 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29060             Alan Hayward  <alan.hayward@arm.com>
29061             David Sherwood  <david.sherwood@arm.com>
29063         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29064         (vect_compile_time_alias): ...this new function.  Do the calculation
29065         on poly_ints rather than trees.
29066         (vect_prune_runtime_alias_test_list): Update call accordingly.
29068 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29069             Alan Hayward  <alan.hayward@arm.com>
29070             David Sherwood  <david.sherwood@arm.com>
29072         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29073         numbers of units.
29074         (vect_schedule_slp_instance): Likewise.
29076 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29077             Alan Hayward  <alan.hayward@arm.com>
29078             David Sherwood  <david.sherwood@arm.com>
29080         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29081         constant and extern definitions for variable-length vectors.
29082         (vect_get_constant_vectors): Note that the number of units
29083         is known to be constant.
29085 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29086             Alan Hayward  <alan.hayward@arm.com>
29087             David Sherwood  <david.sherwood@arm.com>
29089         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29090         of units as polynomial.  Choose between WIDE and NARROW based
29091         on multiple_p.
29093 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29094             Alan Hayward  <alan.hayward@arm.com>
29095             David Sherwood  <david.sherwood@arm.com>
29097         * tree-vect-stmts.c (simd_clone_subparts): New function.
29098         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29100 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29101             Alan Hayward  <alan.hayward@arm.com>
29102             David Sherwood  <david.sherwood@arm.com>
29104         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29105         vectors as polynomial.  Use build_index_vector for
29106         IFN_GOMP_SIMD_LANE.
29108 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29109             Alan Hayward  <alan.hayward@arm.com>
29110             David Sherwood  <david.sherwood@arm.com>
29112         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29113         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29114         for variable-length vectors.
29115         (vectorizable_mask_load_store): Treat the number of units as
29116         polynomial, asserting that it is constant if the condition has
29117         already been enforced.
29118         (vectorizable_store, vectorizable_load): Likewise.
29120 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29121             Alan Hayward  <alan.hayward@arm.com>
29122             David Sherwood  <david.sherwood@arm.com>
29124         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29125         of units as polynomial.  Punt if we can't tell at compile time
29126         which vector contains the final result.
29128 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29129             Alan Hayward  <alan.hayward@arm.com>
29130             David Sherwood  <david.sherwood@arm.com>
29132         * tree-vect-loop.c (vectorizable_induction): Treat the number
29133         of units as polynomial.  Punt on SLP inductions.  Use an integer
29134         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29135         cast of such a series for variable-length floating-point
29136         reductions.
29138 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29139             Alan Hayward  <alan.hayward@arm.com>
29140             David Sherwood  <david.sherwood@arm.com>
29142         * tree.h (build_index_vector): Declare.
29143         * tree.c (build_index_vector): New function.
29144         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29145         of units as polynomial, forcibly converting it to a constant if
29146         vectorizable_reduction has already enforced the condition.
29147         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29148         to create a {1,2,3,...} vector.
29149         (vectorizable_reduction): Treat the number of units as polynomial.
29150         Choose vectype_in based on the largest scalar element size rather
29151         than the smallest number of units.  Enforce the restrictions
29152         relied on above.
29154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29155             Alan Hayward  <alan.hayward@arm.com>
29156             David Sherwood  <david.sherwood@arm.com>
29158         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29159         number of units as polynomial.
29161 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29162             Alan Hayward  <alan.hayward@arm.com>
29163             David Sherwood  <david.sherwood@arm.com>
29165         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29166         * target.def (autovectorize_vector_sizes): Return the vector sizes
29167         by pointer, using vector_sizes rather than a bitmask.
29168         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29169         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29170         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29171         Likewise.
29172         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29173         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29174         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29175         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29176         * omp-general.c (omp_max_vf): Likewise.
29177         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29178         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29179         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29180         * tree-vect-slp.c (vect_slp_bb): Likewise.
29181         * doc/tm.texi: Regenerate.
29182         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29183         to a poly_uint64.
29184         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29185         the vector size as a poly_uint64 rather than an unsigned int.
29186         (current_vector_size): Change from an unsigned int to a poly_uint64.
29187         (get_vectype_for_scalar_type): Update accordingly.
29188         * tree.h (build_truth_vector_type): Take the size and number of
29189         units as a poly_uint64 rather than an unsigned int.
29190         (build_vector_type): Add a temporary overload that takes
29191         the number of units as a poly_uint64 rather than an unsigned int.
29192         * tree.c (make_vector_type): Likewise.
29193         (build_truth_vector_type): Take the number of units as a poly_uint64
29194         rather than an unsigned int.
29196 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29197             Alan Hayward  <alan.hayward@arm.com>
29198             David Sherwood  <david.sherwood@arm.com>
29200         * target.def (get_mask_mode): Take the number of units and length
29201         as poly_uint64s rather than unsigned ints.
29202         * targhooks.h (default_get_mask_mode): Update accordingly.
29203         * targhooks.c (default_get_mask_mode): Likewise.
29204         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29205         * doc/tm.texi: Regenerate.
29207 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29208             Alan Hayward  <alan.hayward@arm.com>
29209             David Sherwood  <david.sherwood@arm.com>
29211         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29212         * omp-general.c (omp_max_vf): Likewise.
29213         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29214         (expand_omp_simd): Handle polynomial safelen.
29215         * omp-low.c (omplow_simd_context): Add a default constructor.
29216         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29217         (lower_rec_simd_input_clauses): Update accordingly.
29218         (lower_rec_input_clauses): Likewise.
29220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29221             Alan Hayward  <alan.hayward@arm.com>
29222             David Sherwood  <david.sherwood@arm.com>
29224         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29225         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29226         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29227         (vect_analyze_slp_cost): Likewise.
29228         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29229         (vect_model_load_cost): Likewise.
29231 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29232             Alan Hayward  <alan.hayward@arm.com>
29233             David Sherwood  <david.sherwood@arm.com>
29235         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29236         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29237         from an unsigned int * to a poly_uint64_pod *.
29238         (calculate_unrolling_factor): New function.
29239         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29242             Alan Hayward  <alan.hayward@arm.com>
29243             David Sherwood  <david.sherwood@arm.com>
29245         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29246         from an unsigned int to a poly_uint64.
29247         (_loop_vec_info::slp_unrolling_factor): Likewise.
29248         (_loop_vec_info::vectorization_factor): Change from an int
29249         to a poly_uint64.
29250         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29251         (vect_get_num_vectors): New function.
29252         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29253         (vect_get_num_copies): Use vect_get_num_vectors.
29254         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29255         to an unsigned int *.
29256         (vect_analyze_data_refs): Change min_vf from an int * to a
29257         poly_uint64 *.
29258         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29259         than an unsigned HOST_WIDE_INT.
29260         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29261         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29262         to an unsigned int *.
29263         (vect_analyze_data_ref_dependences): Likewise.
29264         (vect_compute_data_ref_alignment): Handle polynomial vf.
29265         (vect_enhance_data_refs_alignment): Likewise.
29266         (vect_prune_runtime_alias_test_list): Likewise.
29267         (vect_shift_permute_load_chain): Likewise.
29268         (vect_supportable_dr_alignment): Likewise.
29269         (dependence_distance_ge_vf): Take the vectorization factor as a
29270         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29271         (vect_analyze_data_refs): Change min_vf from an int * to a
29272         poly_uint64 *.
29273         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29274         vfm1 as a poly_uint64 rather than an int.  Make the same change
29275         for the returned bound_scalar.
29276         (vect_gen_vector_loop_niters): Handle polynomial vf.
29277         (vect_do_peeling): Likewise.  Update call to
29278         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29279         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29280         be constant.
29281         * tree-vect-loop.c (vect_determine_vectorization_factor)
29282         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29283         (vect_get_known_peeling_cost): Likewise.
29284         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29285         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29286         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29287         updating the upper bounds of the loop.
29288         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29289         rather than an int.
29290         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29291         polynomial unroll factors.
29292         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29293         (vect_make_slp_decision): Likewise.
29294         (vect_supported_load_permutation_p): Likewise, and polynomial
29295         vf too.
29296         (vect_analyze_slp_cost): Handle polynomial vf.
29297         (vect_slp_analyze_node_operations): Likewise.
29298         (vect_slp_analyze_bb_1): Likewise.
29299         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29300         than an unsigned HOST_WIDE_INT.
29301         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29302         (vectorizable_load): Handle polynomial vf.
29303         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29304         a poly_uint64.
29305         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29307 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29308             Alan Hayward  <alan.hayward@arm.com>
29309             David Sherwood  <david.sherwood@arm.com>
29311         * match.pd: Handle bit operations involving three constants
29312         and try to fold one pair.
29314 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29316         * tree-vect-loop-manip.c: Include gimple-fold.h.
29317         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29318         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29319         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29320         Add a path that uses a step of VF instead of 1, but disable it
29321         for now.
29322         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29323         and niters_no_overflow parameters.  Update calls to
29324         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29325         Create a new SSA name if the latter choses to use a ste other
29326         than zero, and return it via niters_vector_mult_vf_var.
29327         * tree-vect-loop.c (vect_transform_loop): Update calls to
29328         vect_do_peeling, vect_gen_vector_loop_niters and
29329         slpeel_make_loop_iterate_ntimes.
29330         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29331         (vect_gen_vector_loop_niters): Update declarations after above changes.
29333 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29335         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29336         128-bit round to integer instructions.
29337         (ceil<mode>2): Likewise.
29338         (btrunc<mode>2): Likewise.
29339         (round<mode>2): Likewise.
29341 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29343         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29344         unaligned VSX load/store on P8/P9.
29345         (expand_block_clear): Allow the use of unaligned VSX
29346         load/store on P8/P9.
29348 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29350         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29351         New function.
29352         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29353         swap associated with both a load and a store.
29355 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29357         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29358         * config/riscv/riscv.md (clear_cache): Use it.
29360 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29362         * web.c: Remove out-of-date comment.
29364 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29366         * expr.c (fixup_args_size_notes): Check that any existing
29367         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29368         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29369         (emit_single_push_insn): ...here.
29371 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29373         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29374         (const_vector_encoded_nelts): New function.
29375         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29376         (const_vector_int_elt, const_vector_elt): Declare.
29377         * emit-rtl.c (const_vector_int_elt_1): New function.
29378         (const_vector_elt): Likewise.
29379         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29380         of CONST_VECTOR_ELT.
29382 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29384         * expr.c: Include rtx-vector-builder.h.
29385         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29386         directly on the tree encoding.
29387         (const_vector_from_tree): Likewise.
29388         * optabs.c: Include rtx-vector-builder.h.
29389         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29390         sequence of "u" values.
29391         * vec-perm-indices.c: Include rtx-vector-builder.h.
29392         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29393         directly on the vec_perm_indices encoding.
29395 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29397         * doc/rtl.texi (const_vector): Describe new encoding scheme.
29398         * Makefile.in (OBJS): Add rtx-vector-builder.o.
29399         * rtx-vector-builder.h: New file.
29400         * rtx-vector-builder.c: Likewise.
29401         * rtl.h (rtx_def::u2): Add a const_vector field.
29402         (CONST_VECTOR_NPATTERNS): New macro.
29403         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29404         (CONST_VECTOR_DUPLICATE_P): Likewise.
29405         (CONST_VECTOR_STEPPED_P): Likewise.
29406         (CONST_VECTOR_ENCODED_ELT): Likewise.
29407         (const_vec_duplicate_p): Check for a duplicated vector encoding.
29408         (unwrap_const_vec_duplicate): Likewise.
29409         (const_vec_series_p): Check for a non-duplicated vector encoding.
29410         Say that the function only returns true for integer vectors.
29411         * emit-rtl.c: Include rtx-vector-builder.h.
29412         (gen_const_vec_duplicate_1): Delete.
29413         (gen_const_vector): Call gen_const_vec_duplicate instead of
29414         gen_const_vec_duplicate_1.
29415         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29416         (gen_const_vec_duplicate): Use rtx_vector_builder.
29417         (gen_const_vec_series): Likewise.
29418         (gen_rtx_CONST_VECTOR): Likewise.
29419         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29420         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29421         Build a new vector rather than modifying a CONST_VECTOR in-place.
29422         (handle_special_swappables): Update call accordingly.
29423         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29424         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29425         Build a new vector rather than modifying a CONST_VECTOR in-place.
29426         (handle_special_swappables): Update call accordingly.
29428 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29430         * simplify-rtx.c (simplify_const_binary_operation): Use
29431         CONST_VECTOR_ELT instead of XVECEXP.
29433 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29435         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29436         the selector elements to be different from the data elements
29437         if the selector is a VECTOR_CST.
29438         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29439         ssizetype for the selector.
29441 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29443         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29444         before testing each element individually.
29445         * tree-vect-generic.c (lower_vec_perm): Likewise.
29447 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29449         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29450         * selftest-run-tests.c (selftest::run_tests): Call it.
29451         * vector-builder.h (vector_builder::operator ==): New function.
29452         (vector_builder::operator !=): Likewise.
29453         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29454         (vec_perm_indices::all_from_input_p): New function.
29455         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29456         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29457         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29458         instead of reading the VECTOR_CST directly.  Detect whether both
29459         vector inputs are the same before constructing the vec_perm_indices,
29460         and update the number of inputs argument accordingly.  Use the
29461         utility functions added above.  Only construct sel2 if we need to.
29463 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29465         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29466         the broadcast of the low byte.
29467         (expand_mult_highpart): Use an explicit encoding for the permutes.
29468         * optabs-query.c (can_mult_highpart_p): Likewise.
29469         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29470         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29471         (vectorizable_bswap): Likewise.
29472         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29473         explicit encoding for the power-of-2 permutes.
29474         (vect_permute_store_chain): Likewise.
29475         (vect_grouped_load_supported): Likewise.
29476         (vect_permute_load_chain): Likewise.
29478 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29480         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29481         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29482         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29483         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29484         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29485         (vect_gen_perm_mask_any): Likewise.
29487 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29489         * int-vector-builder.h: New file.
29490         * vec-perm-indices.h: Include int-vector-builder.h.
29491         (vec_perm_indices): Redefine as an int_vector_builder.
29492         (auto_vec_perm_indices): Delete.
29493         (vec_perm_builder): Redefine as a stand-alone class.
29494         (vec_perm_indices::vec_perm_indices): New function.
29495         (vec_perm_indices::clamp): Likewise.
29496         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29497         (vec_perm_indices::new_vector): New function.
29498         (vec_perm_indices::new_expanded_vector): Update for new
29499         vec_perm_indices class.
29500         (vec_perm_indices::rotate_inputs): New function.
29501         (vec_perm_indices::all_in_range_p): Operate directly on the
29502         encoded form, without computing elided elements.
29503         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29504         encoding.  Update for new vec_perm_indices class.
29505         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29506         the given vec_perm_builder.
29507         (expand_vec_perm_var): Update vec_perm_builder constructor.
29508         (expand_mult_highpart): Use vec_perm_builder instead of
29509         auto_vec_perm_indices.
29510         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
29511         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
29512         or double series encoding as appropriate.
29513         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
29514         vec_perm_indices instead of auto_vec_perm_indices.
29515         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29516         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29517         (vect_permute_store_chain): Likewise.
29518         (vect_grouped_load_supported): Likewise.
29519         (vect_permute_load_chain): Likewise.
29520         (vect_shift_permute_load_chain): Likewise.
29521         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29522         (vect_transform_slp_perm_load): Likewise.
29523         (vect_schedule_slp_instance): Likewise.
29524         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29525         (vectorizable_mask_load_store): Likewise.
29526         (vectorizable_bswap): Likewise.
29527         (vectorizable_store): Likewise.
29528         (vectorizable_load): Likewise.
29529         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
29530         vec_perm_indices instead of auto_vec_perm_indices.  Use
29531         tree_to_vec_perm_builder to read the vector from a tree.
29532         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
29533         vec_perm_builder instead of a vec_perm_indices.
29534         (have_whole_vector_shift): Use vec_perm_builder and
29535         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
29536         truncation to calc_vec_perm_mask_for_shift.
29537         (vect_create_epilog_for_reduction): Likewise.
29538         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
29539         from auto_vec_perm_indices to vec_perm_indices.
29540         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29541         instead of changing individual elements.
29542         (aarch64_vectorize_vec_perm_const): Use new_vector to install
29543         the vector in d.perm.
29544         * config/arm/arm.c (expand_vec_perm_d::perm): Change
29545         from auto_vec_perm_indices to vec_perm_indices.
29546         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29547         instead of changing individual elements.
29548         (arm_vectorize_vec_perm_const): Use new_vector to install
29549         the vector in d.perm.
29550         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
29551         Update vec_perm_builder constructor.
29552         (rs6000_expand_interleave): Likewise.
29553         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
29554         (rs6000_expand_interleave): Likewise.
29556 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29558         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
29559         to qimode could truncate the indices.
29560         * optabs.c (expand_vec_perm_var): Likewise.
29562 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29564         * Makefile.in (OBJS): Add vec-perm-indices.o.
29565         * vec-perm-indices.h: New file.
29566         * vec-perm-indices.c: Likewise.
29567         * target.h (vec_perm_indices): Replace with a forward class
29568         declaration.
29569         (auto_vec_perm_indices): Move to vec-perm-indices.h.
29570         * optabs.h: Include vec-perm-indices.h.
29571         (expand_vec_perm): Delete.
29572         (selector_fits_mode_p, expand_vec_perm_var): Declare.
29573         (expand_vec_perm_const): Declare.
29574         * target.def (vec_perm_const_ok): Replace with...
29575         (vec_perm_const): ...this new hook.
29576         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
29577         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
29578         * doc/tm.texi: Regenerate.
29579         * optabs.def (vec_perm_const): Delete.
29580         * doc/md.texi (vec_perm_const): Likewise.
29581         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
29582         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
29583         expand_vec_perm for constant permutation vectors.  Assert that
29584         the mode of variable permutation vectors is the integer equivalent
29585         of the mode that is being permuted.
29586         * optabs-query.h (selector_fits_mode_p): Declare.
29587         * optabs-query.c: Include vec-perm-indices.h.
29588         (selector_fits_mode_p): New function.
29589         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
29590         is defined, instead of checking whether the vec_perm_const_optab
29591         exists.  Use targetm.vectorize.vec_perm_const instead of
29592         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
29593         fit in the vector mode before using a variable permute.
29594         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
29595         vec_perm_indices instead of an rtx.
29596         (expand_vec_perm): Replace with...
29597         (expand_vec_perm_const): ...this new function.  Take the selector
29598         as a vec_perm_indices rather than an rtx.  Also take the mode of
29599         the selector.  Update call to shift_amt_for_vec_perm_mask.
29600         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
29601         Use vec_perm_indices::new_expanded_vector to expand the original
29602         selector into bytes.  Check whether the indices fit in the vector
29603         mode before using a variable permute.
29604         (expand_vec_perm_var): Make global.
29605         (expand_mult_highpart): Use expand_vec_perm_const.
29606         * fold-const.c: Includes vec-perm-indices.h.
29607         * tree-ssa-forwprop.c: Likewise.
29608         * tree-vect-data-refs.c: Likewise.
29609         * tree-vect-generic.c: Likewise.
29610         * tree-vect-loop.c: Likewise.
29611         * tree-vect-slp.c: Likewise.
29612         * tree-vect-stmts.c: Likewise.
29613         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
29614         Delete.
29615         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
29616         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
29617         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
29618         (aarch64_vectorize_vec_perm_const): ...this new function.
29619         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29620         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29621         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
29622         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
29623         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29624         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29625         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
29626         into...
29627         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
29628         check for NEON modes.
29629         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
29630         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
29631         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
29632         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
29633         into...
29634         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
29635         the old VEC_PERM_CONST conditions.
29636         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
29637         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
29638         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
29639         (ia64_vectorize_vec_perm_const_ok): Merge into...
29640         (ia64_vectorize_vec_perm_const): ...this new function.
29641         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
29642         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
29643         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
29644         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
29645         * config/mips/mips.c (mips_expand_vec_perm_const)
29646         (mips_vectorize_vec_perm_const_ok): Merge into...
29647         (mips_vectorize_vec_perm_const): ...this new function.
29648         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
29649         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
29650         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
29651         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
29652         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
29653         (rs6000_expand_vec_perm_const): Delete.
29654         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
29655         Delete.
29656         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29657         (altivec_expand_vec_perm_const_le): Take each operand individually.
29658         Operate on constant selectors rather than rtxes.
29659         (altivec_expand_vec_perm_const): Likewise.  Update call to
29660         altivec_expand_vec_perm_const_le.
29661         (rs6000_expand_vec_perm_const): Delete.
29662         (rs6000_vectorize_vec_perm_const_ok): Delete.
29663         (rs6000_vectorize_vec_perm_const): New function.
29664         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29665         an element count and rtx array.
29666         (rs6000_expand_extract_even): Update call accordingly.
29667         (rs6000_expand_interleave): Likewise.
29668         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
29669         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
29670         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
29671         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
29672         (rs6000_expand_vec_perm_const): Delete.
29673         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29674         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29675         (altivec_expand_vec_perm_const_le): Take each operand individually.
29676         Operate on constant selectors rather than rtxes.
29677         (altivec_expand_vec_perm_const): Likewise.  Update call to
29678         altivec_expand_vec_perm_const_le.
29679         (rs6000_expand_vec_perm_const): Delete.
29680         (rs6000_vectorize_vec_perm_const_ok): Delete.
29681         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
29682         reference to the SPE evmerge intructions.
29683         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29684         an element count and rtx array.
29685         (rs6000_expand_extract_even): Update call accordingly.
29686         (rs6000_expand_interleave): Likewise.
29687         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
29688         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
29689         new function.
29690         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29692 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29694         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
29695         vector mode and that that mode matches the mode of the data
29696         being permuted.
29697         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
29698         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
29699         directly using expand_vec_perm_1 when forcing selectors into
29700         registers.
29701         (expand_vec_perm_var): New function, split out from expand_vec_perm.
29703 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29705         * optabs-query.h (can_vec_perm_p): Delete.
29706         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
29707         * optabs-query.c (can_vec_perm_p): Split into...
29708         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
29709         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
29710         particular selector is valid.
29711         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29712         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29713         (vect_grouped_load_supported): Likewise.
29714         (vect_shift_permute_load_chain): Likewise.
29715         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29716         (vect_transform_slp_perm_load): Likewise.
29717         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29718         (vectorizable_bswap): Likewise.
29719         (vect_gen_perm_mask_checked): Likewise.
29720         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
29721         implementations of variable permutation vectors into account
29722         when deciding which selector to use.
29723         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
29724         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
29725         with a false third argument.
29726         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
29727         to test whether the constant selector is valid and can_vec_perm_var_p
29728         to test whether a variable selector is valid.
29730 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29732         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
29733         * optabs-query.c (can_vec_perm_p): Likewise.
29734         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
29735         instead of vec_perm_indices.
29736         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
29737         (vect_gen_perm_mask_checked): Likewise,
29738         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
29739         (vect_gen_perm_mask_checked): Likewise,
29741 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29743         * optabs-query.h (qimode_for_vec_perm): Declare.
29744         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
29745         (qimode_for_vec_perm): ...this new function.
29746         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
29748 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29750         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
29751         does not have a conditional at the top.
29753 2018-01-02  Richard Biener  <rguenther@suse.de>
29755         * ipa-inline.c (big_speedup_p): Fix expression.
29757 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29759         PR target/81616
29760         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
29761         for generic 4->6.
29763 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29765         PR target/81616
29766         Generic tuning.
29767         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
29768         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
29769         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
29770         cond_taken_branch_cost 3->4.
29772 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
29774         PR tree-optimization/83581
29775         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
29776         TODO_cleanup_cfg if any changes have been made.
29778         PR middle-end/83608
29779         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
29780         convert_modes if target mode has the right side, but different mode
29781         class.
29783         PR middle-end/83609
29784         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
29785         last argument when extracting from CONCAT.  If either from_real or
29786         from_imag is NULL, use expansion through memory.  If result is not
29787         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
29788         the parts directly to inner mode, if even that fails, use expansion
29789         through memory.
29791         PR middle-end/83623
29792         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29793         check for bswap in mode rather than HImode and use that in expand_unop
29794         too.
29796 Copyright (C) 2018 Free Software Foundation, Inc.
29798 Copying and distribution of this file, with or without modification,
29799 are permitted in any medium without royalty provided the copyright
29800 notice and this notice are preserved.