Instrument only selected files (PR gcov-profile/87442).
[official-gcc.git] / gcc / ChangeLog
blobb4195db10328d2e52640ac30df8dbc9cacf47c36
1 2018-11-12  Martin Liska  <mliska@suse.cz>
3         PR gcov-profile/87442
4         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
5         options.
6         * doc/invoke.texi: Document them.
7         * tree-profile.c (parse_profile_filter): New.
8         (parse_profile_file_filtering): Likewise.
9         (release_profile_file_filtering): Likewise.
10         (include_source_file_for_profile): Likewise.
11         (tree_profiling): Filter source files based on the
12         newly added options.
14 2018-11-12  Martin Liska  <mliska@suse.cz>
16         PR target/87903
17         * doc/extend.texi: Add missing values for __builtin_cpu_is and
18         __builtin_cpu_supports for x86 target.
20 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
22         PR middle-end/21110
23         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
24         have to be Pmode.
25         (Arithmetic): Likewise for "lo_sum".
27 2018-11-12  Renlin Li  <renlin.li@arm.com>
29         PR target/87815
30         * dse.c (get_stored_val): Add check for compile-time constantness
31         of gap.
33 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
35         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
36         (ARMv8_5a): New fgroup.
37         (armv8.5-a): New arch.
38         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
39         sb and predres.
40         * config/arm/arm-tables.opt: Regenerate.
41         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
42         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
43         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
44         Add matching rules for -march=armv8.5-a and extensions.
45         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
46         Add sb and predres to all armv8-a except armv8.5-a.
48 2018-11-12  Richard Biener  <rguenther@suse.de>
50         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
51         overload for single value.
52         (value_range[_base]::set_nonnull): New.
53         (value_range[_base]::set_null): Likewise.
54         (value_range): Document bitmap copying behavior, mark
55         copy constructor and assignment operator deleted.
56         (value_range::move): New.
57         (value_range::set_and_canonicalize): Default bitmap to zero.
58         (set_value_range_to_nonnull): Remove.
59         (set_value_range_to_null): Likewise.
60         (set_value_range): Likewise.
61         (set_value_range_to_value): Likewise.
62         (extract_range_from_unary_expr): Work on value_range_base.
63         (extract_range_from_binary_expr_1): Likewise.  Rename to...
64         (extract_range_from_binary_expr): ... this.
65         * tree-vrp.c (value_range::update): Clear equiv bitmap
66         if required.
67         (value_range::move): New, move equiv bitmap.
68         (value_range_base::set_undefined): Avoid assignment.
69         (value_range::set_undefined): Likewise.
70         (value_range_base::set_varying): Likewise.
71         (value_range::set_varying): Likewise.
72         (set_value_range): Remove.
73         (value_range_base::set): New overload for value.
74         (value_range::set): Likewise.
75         (set_value_range_to_nonnull): Remove.
76         (value_range_base::set_nonnull): New.
77         (value_range::set_nonnull): Likewise.
78         (set_value_range_to_null): Remove.
79         (value_range_base::set_null): New.
80         (value_range::set_null): Likewise.
81         (range_is_null): Work on value_range_base.
82         (range_is_nonnull): Likewise.
83         (ranges_from_anti_range): Likewise.
84         (extract_range_into_wide_ints): Likewise.
85         (extract_range_from_multiplicative_op): Likewise.
86         (extract_range_from_binary_expr): Likewise.  Update for API changes.
87         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
88         handling.
89         (value_range::intersect_helper): Avoid copy and assignment.
90         (value_range::union_helper): Likewise.
91         (determine_value_range_1): Adjust.
92         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
93         Avoid assignment by using move.
94         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
95         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
96         Likewise.
97         * tree-ssanames.c (get_range_info): Likewise.
98         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
99         * vr-values.c (vr_values::get_value_range): Adjust.
100         (vr_values::update_value_range): Likewise.
101         (symbolic_range_based_on_p): Work on value_range_base.
102         (vr_values::extract_range_from_binary_expr): Use value_range_base.
103         (vr_values::extract_range_from_unary_expr): Likewise.
104         (vr_values::extract_range_from_cond_expr): Avoid assignment.
105         (vr_values::extract_range_from_comparison): Adjust.
106         (vr_values::check_for_binary_op_overflow): Use value_range_base.
107         (vr_values::extract_range_basic): Adjust.
108         (vr_values::adjust_range_with_scev): Likewise.
109         (vr_values::vrp_visit_assignment_or_call): Likewise.
110         (vr_values::get_vr_for_comparison): Change API to avoid
111         assignment and copy construction.
112         (vr_values::compare_name_with_value): Adjust accordingly.
113         (vr_values::compare_names): Likewise.
114         (vr_values::extract_range_from_phi_node): Avoid assignment and
115         bogus in-place modify of equiv bitmap.
116         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
117         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
118         for extract_range_from_unary_expr API change.
119         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
121 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
123         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
125 2018-11-12  Richard Biener  <rguenther@suse.de>
127         * tree-vrp.h (value_range_base::symbolic_p,
128         value_range_base::constant_p, value_range_base::zero_p,
129         value_range_base::singleton_p): Move from value_range.
130         (value_range::dump): Add.
131         * gimple-ssa-evrp-analyze.c
132         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
133         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
134         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
135         Use set_varying.
136         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
137         (value_range::constant_p): Likewise.
138         (value_range::singleton_p): Likewise.
139         (value_range_base::dump): Add.
140         (set_value_range_to_undefined): Remove.
141         (set_value_range_to_varying): Likewise.
142         (range_int_cst_p): Take value_range_base argument.
143         (range_int_cst_singleton_p): Likewise.
144         (value_range_constant_singleton): Likewise.
145         (vrp_set_zero_nonzero_bits): Likewise.
146         (extract_range_from_multiplicative_op): Use set_varying.
147         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
148         (extract_range_from_unary_expr): Likewise.
149         (dump_value_range_base): Change to overload of dump_value_range.
150         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
151         (vrp_prop::visit_stmt): Likewise.
152         (value_range::intersect_helper): Likewise.
153         (value_range::union_helper): Likewise.
154         (determine_value_range_1): Likewise.
156 2018-11-12  Richard Biener  <rguenther@suse.de>
158         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
159         (set_value_range_to_null): Likewise.
160         * vr-values.c (vr_values::extract_range_from_comparison):
161         Clear equiv for constant singleton ranges.
163 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
165         * config/i386/sse.md: Combine VFIXUPIMM* patterns
166         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
167         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
168         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
169         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
170         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
171         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
173 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
175         PR c/69502
176         * doc/extend.texi (Common Type Attributes): For the align type
177         attribute, copy language about decreasing alignment from the
178         corresponding variable attribute.
180 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
182         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
183         -frounding-math.
185 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
187         PR c++/43105
188         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
189         -frtti and -fno-rtti code.
191 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
193         PR c/26366
194         * doc/extend.texi (Other Builtins): Document probability associated
195         with __builtin_expect.
197 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
199         PR target/87928
200         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
201         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
202         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
203         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
205 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
207         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
209 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
211         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
213 2018-11-11  Richard Biener  <rguenther@suse.de>
215         * tree-vrp.h (class value_range_base): New base class for
216         value_range containing all but the m_equiv member.
217         (dump_value_range_base): Add.
218         (range_includes_zero_p): Work on value_range_base.
219         * tree-vrp.c (value_range_base::set): Split out base handling
220         from...
221         (value_range::set): this.
222         (value_range::set_equiv): New.
223         (value_range_base::value_range_base): New constructors.
224         (value_range_base::check): Split out base handling from...
225         (value_range::check): this.
226         (value_range::equal_p): Refactor in terms of
227         ignore_equivs_equal_p which is now member of the base.
228         (value_range_base::set_undefined): New.
229         (value_range_base::set_varying): Likewise.
230         (value_range_base::dump):Split out base handling from...
231         (value_range::dump): this.
232         (value_range_base::set_and_canonicalize): Split out base handling
233         from...
234         (value_range::set_and_canonicalize): this.
235         (value_range_base::union_): New.
236         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
237         for m_vr.
238         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
239         instead of value_range everywhere.
240         (ipcp_vr_lattice::print): Use dump_value_range_base.
241         (ipcp_vr_lattice::meet_with): Adjust.
242         (ipcp_vr_lattice::meet_with_1): Likewise.
243         (ipa_vr_operation_and_type_effects): Likewise.
244         (propagate_vr_across_jump_function): Likewise.
245         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
246         (ipa_get_value_range): Likewise.
247         (ipa_set_jfunc_vr): Likewise.
248         (ipa_compute_jump_functions_for_edge): Likewise.
250 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
252         PR middle-end/65703
253         * doc/invoke.texi (Optimize Options): Add @opindex entries
254         for the positive forms of -fno-xxx and -mno-xxx options
255         that were lacking them.
257 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
259         * combine.c (make_more_copies): Only make an intermediate copy if the
260         dest of a move is a pseudo.
262 2018-11-09  Maya Rashish  <coypu@sdf.org>
264         PR target/87221
265         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
266         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
268 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
270         PR driver/41179
271         PR middle-end/65703
272         * doc/invoke.texi (Optimize Options): Clarify default behavior
273         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
275 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
277         PR tree-optimization/87940
278         * expr.c (string_constant): Don't strip NOPS in subexpressions.
279         Fold PLUS_EXPR correctly.
281 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
283         PR target/87762
284         * config/s390/s390.c (s390_safe_relative_long_p): New function.
285         (annotate_constant_pool_refs): Skip insns which support
286         relative addressing.
287         (annotate_constant_pool_refs_1): New helper function.
288         (find_constant_pool_ref): Skip insns which support relative
289         addression.
290         (find_constant_pool_ref_1): New helper function.
291         (replace_constant_pool_ref): Skip insns which support
292         relative addressing.
293         (replace_constant_pool_ref_1): New helper function.
294         (s390_mainpool_start): Adapt to the new signature.
295         (s390_mainpool_finish): Likewise.
296         (s390_chunkify_start): Likewise.
297         (s390_chunkify_finish): Likewise.
298         (pass_s390_early_mach::execute): Likewise.
299         (s390_prologue_plus_offset): Likewise.
300         (s390_emit_prologue): Likewise.
301         (s390_emit_epilogue): Likewise.
303 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
305         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
306         but unsupported lastprivate with conditional modifier.
308 2018-11-09  Jeff Law  <law@redhat.com>
310         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
311         unused argument better.  Add gcc_unreachable to silence warning.
313 2018-11-09  Martin Sebor  <msebor@redhat.com>
315         PR middle-end/81824
316         * attribs.c (has_attribute): New helper function.
317         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
318         * attribs.h (decls_mismatched_attributes): Declare.
319         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
320         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
321         * common.opt (-Wattribute-alias): Take an argument.
322         (-Wno-attribute-alias): New option.
323         * doc/extend.texi (Common Function Attributes): Document copy.
324         (Common Variable Attributes): Same.
325         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
326         (-Wattribute-alias): Document new option argument.
328 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
330         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
331         matching criteria.  Remove unused array initializer.
333 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
334             Jinsong Ji  <jji@us.ibm.com>
336         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
337         constraints by introducing a new temporary.
338         (_mm_cvtss_si64): Likewise.
340 2018-11-09  Martin Liska  <mliska@suse.cz>
342         * common.opt: Add -fipa-stack-alignment flag.
343         * doc/invoke.texi: Document it.
344         * final.c (rest_of_clean_state): Guard stack
345         shrinking with flag.
347 2018-11-09  Martin Liska  <mliska@suse.cz>
349         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
350         to ...
351         (ipa_discover_variable_flags): ... this.
352         * common.opt: Come up with new flag -fipa-reference-addressable.
353         * doc/invoke.texi: Document it.
354         * ipa-reference.c (propagate): Call the renamed fn.
355         * ipa-visibility.c (whole_program_function_and_variable_visibility):
356         Likewise.
357         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
358         ...
359         (ipa_discover_variable_flags): ... this.  Discover
360         non-addressable variables only with the newly added flag.
361         * opts.c: Enable the newly added flag with -O1 and higher
362         optimization level.
364 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
366         * json.cc (selftest::test_writing_literals): Fix comment.
368 2018-11-09  Martin Liska  <mliska@suse.cz>
370         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
371         string to a stack buffer.
372         (aarch64_parse_cpu): Likewise.
373         (aarch64_parse_tune): Likewise.
375 2018-11-09  Richard Biener  <rguenther@suse.de>
377         PR tree-optimization/87953
378         * tree-vect-loop.c (vectorizable_reduction): For analysis
379         always pass ops[0] to vectorizable_condition.
381 2018-11-09  Stafford Horne  <shorne@gmail.com>
382             Richard Henderson  <rth@twiddle.net>
383             Joel Sherrill  <joel@rtems.org>
385         * common/config/or1k/or1k-common.c: New file.
386         * config/or1k/*: New.
387         * config.gcc (or1k*-*-*): New.
388         * configure.ac (or1k*-*-*): New test for openrisc tls.
389         * configure: Regenerated.
390         * doc/install.texi: Document OpenRISC triplets.
391         * doc/invoke.texi: Document OpenRISC arguments.
392         * doc/md.texi: Document OpenRISC.
394 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
396         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
397         (arm7tdmi-s): Delete CPU.
398         (arm710t): Add aliases for arm720t and arm740t.
399         (arm720t, arm740t): Delete CPUs.
400         (arm920t): Add aliases for arm920, arm922t and arm940t.
401         (arm920, arm922t, arm940t): Delete CPUs.
402         (arm10tdmi): Add alias for arm1020t.
403         (arm1020t): Delete CPU.
404         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
405         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
406         (arm10e): Add aliases for arm1020e and arm1022e.
407         (arm1020e, arm1022e): Delete CPU.
408         * config/arm/arm.md (generic_sched): Remove entries that are now
409         handled by aliases.
410         (generic_vfp): Likewise.
411         * config/arm/arm1020e.md: Simplify tuning selection based on alias
412         changes.
413         * config/arm/arm-tune.md: Regenerated.
414         * config/arm/arm-tables.opt: Regenerated.
416 2018-11-09  Richard Biener  <rguenther@suse.de>
418         PR tree-optimization/87621
419         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
420         op with only phi inputs.
421         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
422         (ch_base::copy_headers): Run CSE on copied loop headers.
423         (pass_ch_vect::process_loop_p): Simplify.
425 2018-11-09  Alexandre Oliva <oliva@adacore.com>
427         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
428         for non-w64 x86_64 biarch.
430 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
432         PR rtl-optimization/86438
433         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
434         of in_b for the compare if in_b is SET_DEST.
436         PR target/87793
437         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
438         non-toplevel UNSPEC.
440 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
442         * tree-vrp.c (value_range::check): Do not access internals
443         directly.
444         (value_range::singleton_p): Same.
445         (value_range::type): Same.
446         (vrp_finalize): Use value_range API.
448 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
450         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
452 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
454         * vr-values.c (vr_values::get_value_range): Use value_range API
455         instead of piecing together ranges.
456         (vr_values::update_value_range): Same.
458 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
460         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
461         of performing ad-hoc calculations.
462         * tree-ssanames.c (set_range_info): New overloaded function
463         accepting value_range &.
464         (get_range_info): Same.
465         * tree-ssanames.h (set_range_info_raw): Remove.
466         (set_range_info): New prototype.
467         (get_range_info): Same.
468         * tree-vrp.h (value_range::null_p): Rename to zero_p.
469         * tree-vrp.c (value_range::null_p): Same.
471 2018-11-09  Jan Hubicka  <jh@suse.cz>
473         * tree.c (fld_type_variant_equal_p): Test user align flag.
474         (flt_type_variant): Copy user align flag.
475         (fld_incomplete_type_of): Clear it.
477 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
479         * config/arm/neon.md (div<mode>3): New pattern.
481 2018-11-08  Andi Kleen  <ak@linux.intel.com>
483         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
484         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
485         (ix86_handle_option): Handle OPT_mptwrite.
486         * config/i386/cpuid.h (bit_PTWRITE): Add.
487         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
488         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
489         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
490         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
491         (ix86_option_override_internal): Handle PTA_PTWRITE.
492         (ix86_valid_target_attribute_inner_p): Define ptwrite.
493         (def_builtin2): Force UINT64 to be 64bit only.
494         * config/i386/i386.h (TARGET_PTWRITE): Add.
495         (TARGET_PTWRITE_P): Add.
496         (PTA_PTWRITE): Add.
497         * config/i386/i386.md: Define ptwrite.
498         * config/i386/i386.opt: Add -mptwrite.
499         * config/i386/immintrin.h (_ptwrite64): Add.
500         (_ptwrite32): Add
501         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
502         * doc/invoke.texi: Document -mptwrite.
504 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
506         PR rtl-optimization/87600
507         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
508         * lra-constraints.c (process_alt_operands): Skip illegal hard
509         register usage.  Prefer reloading non hard register operands.
511 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
513         PR other/36572
514         * doc/invoke.texi (Optimize Options): Clarify default behavior
515         for -fno-sched-interblock and -fno-sched-spec.
517 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
519         * collect2.c (linker_select):  Add USE_LLD_LD.
520         (ld_suffixes): Add ld.lld.
521         (main): Handle -fuse-ld=lld.
522         * common.opt (-fuse-ld=lld): New option.
523         * doc/invoke.texi (-fuse-ld=lld): Document.
524         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
526 2018-11-08  Paul Koning  <ni1d@arrl.net>
528         * config/pdp11/constraints.md: Add "Z" series constraints for use
529         with pre-dec and post-inc addressing.
530         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
531         (pdp11_expand_operands): Add int argument (word count).
532         (pdp11_sp_frame_offset): Delete.
533         (pdp11_cmp_length): New function.
534         (pushpop_regeq): New function.
535         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
536         Add hook.
537         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
538         frame layout.
539         (pdp11_initial_elimination_offset): Ditto.
540         (pdp11_expand_operands): Add word count argument.  Bugfixes.
541         (output_move_multiple): Change how pointer adjustment is done.
542         (pdp11_gen_int_label): Correct format.
543         (output_ascii): Ditto.
544         (pdp11_asm_output_var): Add code for DEC assembler case.
545         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
546         value.
547         (legitimate_const_double_p): Ditto.
548         (pdp11_register_move_cost): Adjust for new register classes.
549         (pdp11_regno_reg_class): Ditto.
550         (expand_block_move): Delete.
551         (pushpop_regeq): New function.
552         (pdp11_legitimate_address_p): Bugfix in check for constant
553         offset.
554         (pdp11_sp_frame_offset): Delete.
555         (pdp11_reg_save_size): New helper function for new frame layout.
556         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
557         (pdp11_expand_shift): Bugfix in check for constant shift count.
558         (pdp11_shift_length): Ditto.
559         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
560         (pdp11_cmp_length): New function.
561         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
562         some compile options.
563         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
564         (CALL_USED_REGISTERS): Ditto.
565         (ELIMINABLE_REGS): Ditto.
566         (REGISTER_NAMES): Ditto.
567         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
568         constraints.
569         (REG_CLASS_NAMES): Ditto.
570         (REG_CLASS_CONTENTS): Ditto.  Also remove
571         HARD_FRAME_POINTER_REGNUM.
572         (CPU_REG_CLASS): New macro.
573         (CLASS_MAX_NREGS): Adjust for new register classes.
574         (FUNCTION_PROFILER): Make no-op.
575         (may_call_alloca): Remove unused declaration.
576         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
577         (ASM_OUTPUT_SKIP): Fix format.
578         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
579         (HARD_FRAME_POINTER_REGNUM): Remove.
580         (return): Delete.
581         (*rts): Rename.  Remove epilogue related checks.
582         (cmpsi, cmpdi): New insn.
583         (cbranch<mode>4): Change to apply to SI and DI modes as well.
584         (mov<mode>): Change constraints to enforce that push/pop
585         destination cannot use the same register as source.
586         (*mov<mode><cc_cc>): Ditto.
587         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
588         at assembly output rather than as RTL expander.
589         (zero_extendqihi2): Bugfix in check for same registers.
590         (adddi3_nocc): Bugfix in check for constant operand.
591         (addsi3_nocc): Ditto.
592         (subdi3_nocc): Ditto.
593         (subsi3_nocc): Ditto.
594         (negdi2_nocc): Copy input to pdp11_expand_operands.
595         (negsi2_nocc): Ditto.
596         (bswap2_nocc): Ditto.
597         * config/pdp11/pdp11.opt (mlra): Fix documentation.
598         * config/pdp11/t-pdp11: Use -Os.
600 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
602         * config/arm/parsecpu.awk (/alias/): New parsing rule.
603         (/begin cpu/): Check that the cpu name hasn't been previously defined.
604         (gen_comm_data): Print out CPU alias tables.
605         (check_cpu): Match aliases when checking the CPU name.
606         * config/arm/arm-protos.h (cpu_alias): New structure.
607         (cpu_option): Add entry for aliases.
608         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
609         strongarm1100 and strongarm1110.
610         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
611         (config/arm/arm-generic.md): Remove redundant references to
612         strongarm110, strongarm1100 and strongarm1110.
613         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
614         Scan aliases for additional hints.
615         (arm_parse_cpu_option_name): Also match a cpu name against the list
616         of aliases.
617         * config/arm/arm-tables.opt: Regenerated.
618         * config/arm/arm-tune.md: Regenerated.
620 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
622         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
623         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
624         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
625         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
626         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
627         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
628         * gengtype.c (open_base_files): Add omp-general.h.
629         * gimple.c (gimple_build_omp_critical):
630         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
631         gimple_omp_taskgroup_set_clauses.
632         (gimple_build_omp_atomic_load): Add mo argument, call
633         gimple_omp_atomic_set_memory_order.
634         (gimple_build_omp_atomic_store): Likewise.
635         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
636         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
637         instead of GSS_OMP.
638         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
639         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
640         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
641         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
642         different value for GF_OMP_ATOMIC_NEED_VALUE.
643         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
644         comments.
645         (struct gimple_statement_omp_single_layout): And remove here.
646         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
647         than gimple_statement_omp_single_layout.
648         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
649         GIMPLE_OMP_TEAMS.
650         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
651         (gimple_omp_subcode): Formatting fix.
652         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
653         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
654         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
655         gimple_omp_teams_host, gimple_omp_teams_set_host,
656         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
657         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
658         gimple_omp_taskgroup_set_clauses): New inline functions.
659         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
660         (gimple_build_omp_atomic_store): Likewise.
661         (gimple_omp_atomic_seq_cst_p): Remove.
662         (gimple_omp_atomic_memory_order): New function.
663         (gimple_omp_atomic_set_seq_cst): Remove.
664         (gimple_omp_atomic_set_memory_order): New function.
665         (gimple_build_omp_taskgroup): Add clauses argument.
666         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
667         (dump_gimple_omp_task): Print taskwait with depend clauses.
668         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
669         dump_omp_atomic_memory_order.
670         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
671         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
672         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
673         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
674         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
675         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
676         ORT_UNTIED_TASKLOOP enumerators.
677         (enum gimplify_defaultmap_kind): New.
678         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
679         target_map_pointers_as_0len_arrays members, add defaultmap.
680         (new_omp_context): Initialize defaultmap member.
681         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
682         (maybe_fold_stmt): Don't fold even in host teams regions.
683         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
684         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
685         ctx->omp_firstprivatize_variable.
686         (omp_add_variable): Don't add private/firstprivate for VLAs in
687         ORT_TASKGROUP.
688         (omp_default_clause): Print "taskloop" rather than "task" if
689         ORT_*TASKLOOP.
690         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
691         Handle new defaultmap clause kinds.
692         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
693         iterator to be lastprivate or private.  Fix up diagnostics if linear
694         is used on collapse>1 simd iterator.
695         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
696         (gimplify_omp_depend): New function.
697         (gimplify_scan_omp_clauses): Add shared clause on parallel for
698         combined parallel master taskloop{, simd} if taskloop has
699         firstprivate, lastprivate or reduction clause.  Handle
700         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
701         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
702         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
703         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
704         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
705         lastprivate.
706         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
707         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
708         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
709         OMP_CLAUSE_{TASK,IN}_REDUCTION.
710         (gimplify_omp_task): Handle taskwait with depend clauses.
711         (gimplify_omp_for): Add shared clause on parallel for combined
712         parallel master taskloop{, simd} if taskloop has firstprivate,
713         lastprivate or reduction clause.  Use ORT_TASKLOOP or
714         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
715         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
716         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
717         __for_end and __for_range temporaries on OMP_PARALLEL for
718         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
719         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
720         sandwiched in between two taskloops.
721         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
722         instead of ctx->omp_firstprivatize_variable.
723         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
724         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
725         host teams, use gimplify_and_return_first etc. for body like
726         for target or target data constructs, and at the end call
727         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
728         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
729         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
730         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
731         gimple_omp_atomic_set_seq_cst calls.
732         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
733         case, handle taskgroup clauses.
734         * lto-streamer-out.c (hash_tree): Handle
735         OMP_CLAUSE_{TASK,IN}_REDUCTION.
736         * Makefile.in (GTFILES): Add omp-general.h.
737         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
738         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
739         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
740         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
741         BUILT_IN_GOMP_LOOP_DOACROSS_START,
742         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
743         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
744         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
745         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
746         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
747         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
748         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
749         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
750         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
751         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
752         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
753         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
754         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
755         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
756         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
757         * omp-expand.c (workshare_safe_to_combine_p): Return false for
758         non-worksharing loops.
759         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
760         (determine_parallel_type): Don't combine parallel with worksharing
761         which has _reductemp_ clause.
762         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
763         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
764         if there is nonmonotonic modifier or if there is no modifier and no
765         ordered clause.  For dynamic and guided schedule without monotonic
766         and nonmonotonic modifier, default to nonmonotonic.
767         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
768         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
769         task reductions.
770         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
771         there are any reduction clauses.
772         (expand_taskwait_call): New function.
773         (expand_teams_call): New function.
774         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
775         expand_teams_call for it.  Formatting fix.  Handle taskwait with
776         depend clauses.
777         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
778         of worksharing loops with task reductions.
779         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
780         expansion of worksharing loops with task reductions.
781         (expand_omp_sections): Handle expansion of sections with task
782         reductions.
783         (expand_omp_synch): For host teams call expand_omp_taskreg.
784         (omp_memory_order_to_memmodel): New function.
785         (expand_omp_atomic_load, expand_omp_atomic_store,
786         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
787         instead of gimple_omp_atomic_seq_cst_p.
788         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
789         depend clauses as a standalone directive.
790         * omp-general.c (enum omp_requires): New variable.
791         (omp_extract_for_data): Initialize have_reductemp member.  Allow
792         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
793         GT_EXPR loops depending on incr sign.  Formatting fixes.
794         * omp-general.h (struct omp_for_data): Add have_reductemp member.
795         (enum omp_requires): New enum.
796         (omp_requires_mask): Declare.
797         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
798         Fix comment typos.
799         * omp-low.c (struct omp_context): Add task_reductions and
800         task_reduction_map fields.
801         (is_host_teams_ctx): New function.
802         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
803         (use_pointer_for_field): Use is_global_var instead of
804         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
805         in outer contexts.
806         (build_outer_var_ref): Ignore taskgroup outer contexts.
807         (delete_omp_context): Release task_reductions and task_reduction_map.
808         (scan_sharing_clauses): Don't add any fields for reduction clause on
809         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
810         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
811         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
812         Handle OMP_CLAUSE_NONTEMPORAL.
813         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
814         needed.
815         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
816         clauses with task modifier.
817         (scan_omp_task): Handle taskwait with depend clauses.
818         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
819         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
820         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
821         (scan_omp_for): Fix comment formatting.
822         (scan_omp_teams): Handle host teams constructs.
823         (check_omp_nesting_restrictions): Allow teams with no outer
824         OpenMP context.  Adjust diagnostics for teams strictly nested into
825         some explicit OpenMP construct other than target.  Allow OpenMP atomics
826         inside of simd regions.
827         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
828         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
829         taskreg_nesting_level while scanning host teams construct.
830         (task_reduction_read): New function.
831         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
832         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
833         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
834         second argument create_tmp_var if it is NULL.  Don't ignore shared
835         clauses in is_host_teams_ctx contexts.  Handle
836         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
837         clauses.
838         (lower_reduction_clauses): Ignore reduction clauses with task
839         modifier.  Remove second argument create_tmp_var if it is NULL.
840         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
841         (lower_send_clauses): Ignore reduction clauses with task modifier.
842         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
843         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
844         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
845         rely that it is the last stmt in body so far.  Ignore outer taskgroup
846         contexts.
847         (omp_task_reductions_find_first, omp_task_reduction_iterate,
848         lower_omp_task_reductions): New functions.
849         (lower_omp_sections): Handle reduction clauses with taskgroup
850         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
851         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
852         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
853         modifiers.
854         (lower_omp_taskgroup): Handle taskgroup reductions.
855         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
856         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
857         (lower_depend_clauses): If there are any
858         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
859         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
860         seen, assume lowering is done already and return early.  Set kind
861         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
862         (lower_omp_taskreg): Handle reduction clauses with task modifier on
863         parallel construct.  Handle reduction clause on taskloop construct.
864         Handle taskwait with depend clauses.
865         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
866         for host teams constructs.
867         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
868         nontemporal and _reductemp_ clause entries.
869         (omp_clause_code_name): Likewise.
870         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
871         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
872         * tree-core.h (enum omp_clause_code): Add
873         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
874         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
875         (struct tree_base): Add omp_atomic_memory_order field into union.
876         Remove OMP_ATOMIC_SEQ_CST comment.
877         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
878         and OMP_CLAUSE_DEPEND_DEPOBJ.
879         (struct tree_omp_clause): Add subcode.defaultmap_kind.
880         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
881         OpenMP constructs with body and clauses operands.
882         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
883         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
884         (OMP_TASKGROUP_CLAUSES): Define.
885         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
886         OMP_CLAUSE__LOOPTEMP_.
887         (OMP_ATOMIC_SEQ_CST): Remove.
888         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
889         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
890         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
891         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
892         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
893         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
894         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
895         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
896         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
897         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
898         Define.
899         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
900         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
901         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
902         (convert_local_omp_clauses): Likewise.  Remove useless test.
903         * tree-parloops.c (create_call_for_reduction_1): Pass
904         OMP_MEMORY_ORDER_RELAXED as new argument to
905         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
906         * tree-pretty-print.c (dump_omp_iterators): New function.
907         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
908         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
909         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
910         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
911         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
912         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
913         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
914         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
915         (dump_omp_atomic_memory_order): New function.
916         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
917         taskwait with depend clauses.
918         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
919         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
920         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
921         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
922         write_ts_omp_clause_tree_pointers): Likewise.
924 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
926         PR ipa/86395
927         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
928         "-missed", "-note", and "-all" sub-options.
929         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
930         API.
931         (can_early_inline_edge_p): Likewise.
932         (want_early_inline_function_p): Likewise.
933         (want_inline_self_recursive_call_p): Likewise.
934         (recursive_inlining): Likewise.
935         (inline_small_functions): Likewise.
936         (flatten_function): Likewise.
937         (ipa_inline): Likewise.
938         (inline_always_inline_functions): Likewise.
939         (early_inline_small_functions): Likewise.
940         (early_inliner): Likewise.
941         * tree-inline.c (expand_call_inline): Likewise.
943 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
945         * pretty-print.c (pp_format): Handle %f.
946         (selftest::test_pp_format): Add test of %f.
947         * pretty-print.h (pp_double): New macro.
949 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
951         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
952         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
953         (dump_pretty_printer::decode_format): Implement "%C" for
954         cgraph_node *.
955         (selftest::test_capture_of_dump_calls): Rename "where" to
956         "stmt_loc".  Convert test_decl to a function decl and set its
957         location.  Add a symbol_table_test RAII instance and a
958         cgraph_node, using it to test "%C" and dump_symtab_node.
960 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
962         PR middle-end/87916
963         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
965 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
967         * cgraph.c: Include "selftest.h".
968         (saved_symtab): New variable.
969         (selftest::symbol_table_test::symbol_table_test): New ctor.
970         (selftest::symbol_table_test::~symbol_table_test): New dtor.
971         (selftest::test_symbol_table_test): New test.
972         (selftest::cgraph_c_tests): New.
973         * cgraph.h (saved_symtab): New decl.
974         (selftest::symbol_table_test): New class.
975         * selftest-run-tests.c (selftest::run_tests): Call
976         selftest::cgraph_c_tests.
977         * selftest.h (selftest::cgraph_c_tests): New decl.
979 2018-11-08  Richard Biener  <rguenther@suse.de>
981         * tree-data-ref.h (lambda_int): New typedef.
982         (lambda_vector_gcd): Adjust.
983         (lambda_vector_new): Likewise.
984         (lambda_matrix_new): Likewise.
985         * tree-data-ref.c  (print_lambda_vector): Adjust.
987 2018-11-08  Richard Biener  <rguenther@suse.de>
989         PR tree-optimization/87929
990         * tree-complex.c (expand_complex_comparison): Clean EH.
992 2018-11-08  Martin Liska  <mliska@suse.cz>
994         * doc/extend.texi: Reword.
995         * predict.c (expr_expected_value_1): Likewise.
997 2018-11-08  Richard Biener  <rguenther@suse.de>
999         PR tree-optimization/87913
1000         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1001         of extreme values to ordered comparisons.
1003 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1005         PR middle-end/42726
1006         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1008 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1010         * doc/invoke.texi: Remove leading dash from @opindex entries
1011         throughout the file.
1013 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1015         PR driver/80828
1016         * doc/invoke.texi (Option Summary): Add -e and --entry.
1017         (Link Options): Likewise.
1019 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1021         PR 87926
1022         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1023         --disable-checking bootstrap.
1025 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1027         * configure: Regenerated.
1029 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1031         PR c/87691
1032         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1033         to the mode of the widest field iff the widest field has mode class
1034         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1035         reference.
1037 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1039         * common.opt: Add -Wattribute-warning.
1040         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1041         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1042         call to allow user configure behavior of "warning" attribute.
1044 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1046         * target.def: Put @: after every vs., e.g., and i.e. where it is
1047         followed by whitespace.
1048         * doc/extend.texi: Ditto.
1049         * doc/fragments.texi: Ditto.
1050         * doc/gimple.texi: Ditto.
1051         * doc/implement-c.texi: Ditto.
1052         * doc/install.texi: Ditto.
1053         * doc/invoke.texi: Ditto.
1054         * doc/md.texi: Ditto.
1055         * doc/plugins.texi: Ditto.
1056         * doc/rtl.texi: Ditto.
1057         * doc/sourcebuild.texi: Ditto.
1058         * doc/tm.texi.in: Ditto.
1059         * doc/ux.texi: Ditto.
1060         * doc/tm.texi: Regenerate.
1062 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1064         * config/arm/arm-cpus.in (ares): New entry.
1065         * config/arm/arm-tables.opt: Regenerate.
1066         * config/arm/arm-tune.md: Likewise.
1067         * doc/invoke.texi (ARM Options): Document ares.
1069 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1071         * config/aarch64/aarch64-cores.def (ares): Define.
1072         * config/aarch64/aarch64-tune.md: Regenerate.
1073         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1075 2018-11-07  Jan Hubicka  <jh@suse.cz>
1077         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1078         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1079         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1080         TYPE_DECL.
1082 2018-11-07  Richard Biener  <rguenther@suse.de>
1084         PR tree-optimization/87914
1085         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1086         of nested cycles.
1087         (vectorizable_reduction): Handle shifts and rotates by dispatching
1088         to vectorizable_shift.
1089         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1090         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1091         def cases.
1092         (vectorizable_shift): Export and handle being called as
1093         vect_nested_cycle.
1094         (vect_analyze_stmt): Call vectorizable_shift after
1095         vectorizable_reduction.
1096         * tree-vectorizer.h (vectorizable_shift): Declare.
1098 2018-11-07  Jan Hubicka  <jh@suse.cz>
1100         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1101         than const decls in TREE_VALUE of enum.
1102         (dump_type_inheritance_graph): Improve duplicate dumping.
1103         (free_enum_values): New.
1104         (build_type_inheritance_graph): Use it.
1105         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1106         which are not main variants or not ODR types.
1107         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1109 2018-11-07  Richard Biener  <rguenther@suse.de>
1111         * ipa-inline.c (want_inline_small_function_p): Compute
1112         big_speedup_p lazily and last.
1114 2018-11-07  Jan Hubicka  <jh@suse.cz>
1116         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1117         building incomplete variant of complete type.
1118         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1119         variant of complete type.
1121 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1123         * config/mips/mips.c: Fix typo in documentation of
1124         mips_loongson_ext2_prefetch_cookie.
1125         (mips_option_override): fix brain twister logical.
1126         * config/mips/mips.h: Fix typo in documentation of
1127         ISA_HAS_CTZ_CTO and define pattern.
1128         * config/mips/mips.md (prefetch): Hoist EXT2 above
1129         the 2EF/EXT block.
1130         (prefetch_indexed): Hoist EXT2 above the EXT block.
1132 2018-11-07  Jan Hubicka  <jh@suse.cz>
1134         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1135         return and parameter types of function and method types.
1136         (free_lang_data_in_cgraph): Update.
1138 2018-11-07  Martin Liska  <mliska@suse.cz>
1140         PR rtl-optimization/87868
1141         * postreload-gcse.c (eliminate_partially_redundant_load): Set
1142         threshold to max_count if we would overflow.
1143         * profile-count.h: Make max_count a public constant.
1145 2018-11-07  Martin Liska  <mliska@suse.cz>
1147         * mem-stats.h: Fix GNU coding style.
1149 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1151         * config/mips/gs264e.md: New.
1152         * config/mips/mips-cpus.def: Define gs264e.
1153         * config/mips/mips-tables.opt: Regenerate.
1154         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1155         gs264e.
1156         (mips_issue_rate): Add support for gs264e.
1157         (mips_multipass_dfa_lookahead): Likewise.
1158         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1159         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1160         (MIPS_ASE_MSA_SPEC): New.
1161         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1162         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1163         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1164         * config/mips/mips.md: Include gs264e.md.
1165         (processor): Add gs264e.
1166         * config/mips/mips.opt (MSA): Use Mask instead of Var.
1167         * doc/invoke.texi: Add gs264e to supported architectures.
1169 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1171         * config/mips/gs464e.md: New.
1172         * config/mips/mips-cpus.def: Define gs464e.
1173         * config/mips/mips-tables.opt: Regenerate.
1174         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1175         gs464e.
1176         (mips_issue_rate): Add support for gs464e.
1177         (mips_multipass_dfa_lookahead): Likewise.
1178         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1179         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1180         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1181         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1182         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1183         * config/mips/mips.md: Include gs464e.md.
1184         (processor): Add gs464e.
1185         * doc/invoke.texi: Add gs464e to supported architectures.
1187 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1189         * config/mips/loongson3a.md: Rename to ...
1190         * config/mips/gs464.md: ... here.
1191         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1192         as an alias of gs464 processor.
1193         * config/mips/mips-tables.opt: Regenerate.
1194         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1195         instead of PROCESSOR_LOONGSON_3A.
1196         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1197         TUNE_LOONGSON_3A.
1198         (mips_option_override): Enable MMI and EXT for gs464.
1199         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1200         Rename TUNE_LOONGSON_3A to TUNE_GS464.
1201         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1202         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1203         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1204         TARGET_LOONGSON_3A.
1205         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1206         (processor): Add gs464;
1207         * doc/invoke.texi: Add gs464 to supported architectures.
1209 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1211         * config/mips/mips-protos.h
1212         (mips_loongson_ext2_prefetch_cookie): New prototype.
1213         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1214         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1215         TARGET_LOONGSON_EXT2 is true.
1216         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1217         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1218         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1219         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1220         TARGET_LOONGSON_EXT2.
1221         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1222         (define_insn "ctz<mode>2"): New insn pattern.
1223         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1224         (define_insn "prefetch_indexed_<mode>"): Include
1225         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1226         * config/mips/mips.opt (-mloongson-ext2): Add option.
1227         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1229 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1231         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1232         __mips_loongson_ext.
1233         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1234         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1235         -mloongson-ext.
1236         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1237         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1238         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1239         instead of TARGET_LOONGSON_3A.
1240         * config/mips/mips.opt (-mloongson-ext): Add option.
1241         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1243 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1245         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1246         * config/mips/loongson.md: Move to ...
1247         * config/mips/loongson-mmi.md: here; Adjustment.
1248         * config/mips/loongson.h: Move to ...
1249         State as deprecated. Include loongson-mmiintrin.h for back
1250         compatibility and warning.
1251         * config/mips/loongson-mmiintrin.h: ... here.
1252         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1253         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1254         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1255         (mips_option_override): Make sure MMI use hard float;
1256         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1257         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1258         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1259         TARGET_LOONGSON_VECTORS.
1260         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1261         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1262         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1263         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1264         -mloongson-mmi.
1265         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1266         TARGET_LOONGSON_VECTORS.
1267         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1268         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1269         (Loongson MMI patterns): Include loongson-mmi.md instead of
1270         loongson.md.
1271         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1272         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1274 2018-11-07  Richard Biener  <rguenther@suse.de>
1276         PR lto/87906
1277         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1278         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1280 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
1282         PR rtl-optimization/87874
1283         * lra.c (lra_substitute_pseudo): Do not create a subreg for
1284         const wide ints.
1286 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
1288         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1289         if not in indexed or indirect form.
1290         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1291         (bswap<mode>2_store): Ditto.
1293 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
1295         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1296         the UNSPEC.
1298 2018-11-06  Richard Biener  <rguenther@suse.de>
1300         PR tree-optimization/86850
1301         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1302         instead of src.m_vec.
1304 2018-11-06  Jan Hubicka  <jh@suse.cz>
1306         * tree.c (fld_simplified_type_name): Break out form ...
1307         (free_lang_data_in_type): ... here.
1308         (fld_type_variant_equal_p): Use it.
1310 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1312         * config/default-d.c: Include memmodel.h.
1314         * config/sol2-d.c: New file.
1315         * config/t-sol2 (sol2-d.o): New rule.
1316         * config.gcc <*-*-solaris2*>: Set d_target_objs,
1317         target_has_targetdm.
1319 2018-11-06  Jan Hubicka  <jh@suse.cz>
1321         * tree.c (fld_type_variant): Also copy alignment; be sure that
1322         new variant is equal.
1324 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1326         PR target/87762
1327         * config/s390/s390.md: Add relative_long attribute.
1329 2018-11-06  Jan Hubicka  <jh@suse.cz>
1331         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1332         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1333         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1334         stream TYPE_NEEDS_CONSTRUCTING.
1335         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1336         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1338 2018-11-06  Richard Biener  <rguenther@suse.de>
1340         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1341         dump-scope ...
1342         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1344 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1346         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1347         REG_CLASS_CONTENTS[GEN_REGS].
1348         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1350 2018-11-06  Jan Hubicka  <jh@suse.cz>
1352          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1354 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1356         PR sanitizer/80953
1357         * config/sol2.h (ASAN_CC1_SPEC): Define.
1358         (LD_WHOLE_ARCHIVE_OPTION): Define.
1359         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1360         (ASAN_REJECT_SPEC): Provide default.
1361         (LIBASAN_EARLY_SPEC): Define.
1362         (LIBTSAN_EARLY_SPEC): Define.
1363         (LIBLSAN_EARLY_SPEC): Define.
1364         * config/i386/sol2.h (CC1_SPEC): Redefine.
1365         (ASAN_REJECT_SPEC): Define.
1367         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1368         (TARGET_ASAN_SHADOW_OFFSET): Define.
1369         (sparc_asan_shadow_offset): New function.
1370         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1371         (ASAN_REJECT_SPEC): Define.
1373 2018-11-06  Jan Hubicka  <jh@suse.cz>
1375         * tree.c (fld_type_variant): Copy canonical type.
1376         (fld_incomplete_type_of): Check that canonical types looks sane;
1377         copy canonical type.
1378         (verify_type): Accept when incomplete type has complete canonical type.
1380 2018-11-06  Jan Hubicka  <jh@suse.cz>
1382         * tree.c (free_lang_data): Reset overwite_assembler_name,
1383         print_xnode, print_decl, print_type and print_identifier of
1384         langhooks.
1386 2018-11-06  Richard Biener  <rguenther@suse.de>
1388         PR tree-optimization/87889
1389         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1390         Do nothing if old and new arg are the same
1392 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
1394         PR target/87723
1395         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1396         attributes for operands 3 and 4.
1398 2018-11-06  Richard Biener  <rguenther@suse.de>
1400         PR middle-end/18041
1401         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1402         matching bitfield insertion.
1404 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
1406         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
1407         (reg_next_debug_use): New.
1408         (attempt_change): Propagate adjusted expression into affected
1409         debug insns.
1410         (merge_in_block): Track uses in debug insns.
1411         (pass_inc_dec::execute): Allocate and release
1412         reg_next_debug_use.
1414 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1416         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1417         (_mm512_fixupimm_round_pd): Update parameters and builtin.
1418         (_mm512_maskz_fixupimm_round_pd): Ditto.
1419         (_mm512_fixupimm_round_ps): Ditto.
1420         (_mm512_maskz_fixupimm_round_ps): Ditto.
1421         (_mm_fixupimm_round_sd): Ditto.
1422         (_mm_maskz_fixupimm_round_sd): Ditto.
1423         (_mm_fixupimm_round_ss): Ditto.
1424         (_mm_maskz_fixupimm_round_ss): Ditto.
1425         (_mm512_fixupimm_pd): Ditto.
1426         (_mm512_maskz_fixupimm_pd): Ditto.
1427         (_mm512_fixupimm_ps): Ditto.
1428         (_mm512_maskz_fixupimm_ps): Ditto.
1429         (_mm_fixupimm_sd): Ditto.
1430         (_mm_maskz_fixupimm_sd): Ditto.
1431         (_mm_fixupimm_ss): Ditto.
1432         (_mm_maskz_fixupimm_ss): Ditto.
1433         (_mm512_mask_fixupimm_round_pd): Update builtin.
1434         (_mm512_mask_fixupimm_round_ps): Ditto.
1435         (_mm_mask_fixupimm_round_sd): Ditto.
1436         (_mm_mask_fixupimm_round_ss): Ditto.
1437         (_mm512_mask_fixupimm_pd): Ditto.
1438         (_mm512_mask_fixupimm_ps): Ditto.
1439         (_mm_mask_fixupimm_sd): Ditto.
1440         (_mm_mask_fixupimm_ss): Ditto.
1441         * config/i386/avx512vlintrin.h:
1442         (_mm256_fixupimm_pd): Update parameters and builtin.
1443         (_mm256_maskz_fixupimm_pd): Ditto.
1444         (_mm256_fixupimm_ps): Ditto.
1445         (_mm256_maskz_fixupimm_ps): Ditto.
1446         (_mm_fixupimm_pd): Ditto.
1447         (_mm_maskz_fixupimm_pd): Ditto.
1448         (_mm_fixupimm_ps): Ditto.
1449         (_mm_maskz_fixupimm_ps): Ditto.
1450         (_mm256_mask_fixupimm_pd): Update builtin.
1451         (_mm256_mask_fixupimm_ps): Ditto.
1452         (_mm_mask_fixupimm_pd): Ditto.
1453         (_mm_mask_fixupimm_ps): Ditto.
1454         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1455         * config/i386/i386-builtin.def: Update builtin definitions.
1456         * config/i386/i386.c: Handle new builtin types and remove useless ones.
1457         * config/i386/sse.md: Update VFIXUPIMM* patterns.
1458         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1459         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1460         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1461         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1462         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1463         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1464         * config/i386/subst.md:
1465         (round_saeonly_sd_mask_operand4): Add new subst_attr.
1466         (round_saeonly_sd_mask_op4): Ditto.
1467         (round_saeonly_expand_operand5): Ditto.
1468         (round_saeonly_expand): Update.
1470 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
1472         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1474 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
1476         PR rtl-optimization/87871
1477         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1479 2018-11-05  Paul Koning  <ni1d@arrl.net>
1481         * doc/sourcebuild.texi (target attributes): Document new "inf"
1482         effective target keyword.
1484 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1486         * config/s390/s390.c (s390_register_move_cost): Increase costs for
1487         moves involving the CC reg.
1489 2018-11-05  Richard Biener  <rguenther@suse.de>
1491         PR tree-optimization/87873
1492         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1493         argument.
1494         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1495         * tree-vect-loop.c (vect_transform_loop): When splitting the
1496         loop exit also create forwarder PHIs for constants.
1497         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1498         Handle constant to_arg, add extra checking we match up the correct
1499         PHIs.
1501 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1503         * config/s390/s390.md: QImode and HImode for load on condition.
1505 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1507         * config/s390/predicates.md: Fix typo.
1508         * config/s390/s390.md: Allow immediates for load on condition.
1510 2018-11-05  Martin Liska  <mliska@suse.cz>
1512         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1513         * bitmap.h (struct bitmap_usage): Likewise.
1514         * ggc-common.c (SCALE): Remove.
1515         (LABEL): Likewise.
1516         (struct ggc_usage): Use SIZE_AMOUNT. And update
1517         compare method.
1518         * ggc-page.c (SCALE): Remove.
1519         (STAT_LABEL): Remove.
1520         (ggc_print_statistics): Use SIZE_AMOUNT.
1521         * gimple.h (SCALE): Remove.
1522         (LABEL): Likewise.
1523         * input.c (ONE_K): Remove.
1524         (ONE_M): Likewise.
1525         (SCALE): Likewise.
1526         (STAT_LABEL): Likewise.
1527         (FORMAT_AMOUNT): Likewise.
1528         (dump_line_table_statistics): Use SIZE_AMOUNT.
1529         * mem-stats.h (struct mem_usage): Likewise.
1530         * rtl.c (dump_rtx_statistics): Likewise.
1531         (rtx_alloc_counts): Change type to size_t.
1532         (rtx_alloc_sizes): Likewise.
1533         (rtx_count_cmp): New.
1534         (dump_rtx_statistics): Sort first based on counts.
1535         * tree.c (tree_nodes_cmp): New.
1536         (tree_codes_cmp): New.
1537         (dump_tree_statistics): Sort first based on counts.
1538         * system.h (ONE_K): New.
1539         (ONE_M): Likewise.
1540         (SIZE_SCALE): Likewise.
1541         (SIZE_LABEL): Likewise.
1542         (SIZE_AMOUNT): Likewise.
1543         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
1544         * tree-dfa.c (dump_dfa_stats): Likewise.
1545         * tree-phinodes.c (phinodes_print_statistics): Likewise.
1546         * tree-ssanames.c (ssanames_print_statistics): Likewise.
1547         * tree.c (dump_tree_statistics): Likewise.
1548         * vec.c (struct vec_usage): Likewise.
1549         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
1550         trigger a -Werror=format-overflow with
1551         --enable-gather-detailed-stats.
1553 2018-11-05  Martin Liska  <mliska@suse.cz>
1555         * mem-stats.h (mem_alloc_description::release_instance_overhead):
1556         Return T *.
1557         * vec.c (struct vec_usage): Register m_element_size.
1558         (vec_prefix::register_overhead): New arguments: elements and
1559         element_size.
1560         (vec_prefix::release_overhead): Subtract elements.
1561         * vec.h (struct vec_prefix): Change signature.
1562         (va_heap::reserve): Pass proper arguments.
1563         (va_heap::release): Likewise.
1565 2018-11-05  Martin Liska  <mliska@suse.cz>
1567         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
1568         style.
1569         * vec.c: Likewise.
1571 2018-11-05  Richard Biener  <rguenther@suse.de>
1573         * tree-scalar-evolution.h (final_value_replacement_loop): Update
1574         prototype.
1575         * tree-scalar-evolution.c (final_value_replacement_loop): Return
1576         whether anything was done.
1577         (scev_const_prop): Remove constant propagation part, fold
1578         remains into ...
1579         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
1580         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
1581         conditionally.
1583 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
1585         PR tree-optimization/87859
1586         * gimple-ssa-store-merging.c (struct merged_store_group): Add
1587         only_constants and first_nonmergeable_order members.
1588         (merged_store_group::merged_store_group): Initialize them.
1589         (merged_store_group::do_merge): Clear only_constants member if
1590         adding something other than INTEGER_CST store.
1591         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1592         stores with order >= first_nonmergeable_order.  Use
1593         merged_store->only_constants instead of always recomputing it.
1594         Set merged_store->first_nonmergeable_order if we've skipped any
1595         stores.  Attempt to merge overlapping INTEGER_CST stores that
1596         we would otherwise skip.
1598         PR sanitizer/87837
1599         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
1601 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
1603         PR target/87853
1604         * config/i386/emmintrin.h (__v16qs): New to cope with option
1605         -funsigned-char.
1606         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
1607         (_mm_cmplt_epi8): Likewise.
1608         (_mm_cmpgt_epi8): Likewise.
1610 2018-11-05  Richard Biener  <rguenther@suse.de>
1612         PR rtl-optimization/87852
1613         * fwprop.c (use_killed_between): Only consider single-defs of the
1614         use whose definition statement dominates the use.
1616 2018-11-05  Martin Liska  <mliska@suse.cz>
1618         PR web/87829
1619         * doc/invoke.texi: Remove options that are
1620         not disabled with -Os.
1622 2018-11-05  Martin Liska  <mliska@suse.cz>
1624         PR c/87811
1625         * doc/extend.texi: Update constrain about the last argument
1626         of __builtin_expect_with_probability.
1628 2018-11-05  Martin Liska  <mliska@suse.cz>
1630         PR c/87811
1631         * predict.c (expr_expected_value_1): Verify
1632         that last argument is a real constants and emit
1633         error.
1635 2018-11-05  Martin Liska  <mliska@suse.cz>
1637         PR gcov-profile/77698
1638         * ipa-profile.c (ipa_profile): Adjust hotness threshold
1639         only in LTO mode.
1641 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1643         PR tree-optimization/86572
1644         * builtins.c (c_strlen): Handle negative offsets in a safe way.
1646 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1648         PR tree-optimization/87672
1649         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
1650         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
1652 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
1654         PR middle-end/58372
1655         * cfgexpand.c (pass_expand::execute): Move the call to
1656         finish_eh_generation in front of the call to expand_stack_alignment.
1658 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1660         * common/config/i386/i386-common.c (processor_alias_table): Add
1661         znver2 entry.
1662         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
1663         (case ${target}): Add znver2.
1664         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
1665         -march=native recognize znver2 processors.
1666         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
1667         * config/i386/i386.c (m_znver2): New definition.
1668         (m_ZNVER): New definition.
1669         (m_AMD_MULTIPLE): Includes m_znver2.
1670         (processor_cost_table): Add znver2 entry.
1671         (processor_target_table): Add znver2 entry.
1672         (get_builtin_code_for_version): Set priority for
1673         PROCESSOR_ZNVER2.
1674         (processor_model): Add M_AMDFAM17H_ZNVER2.
1675         (arch_names_table): Ditto.
1676         (ix86_reassociation_width): Include znver2.
1677         * config/i386/i386.h (TARGET_znver2): New definition.
1678         (struct ix86_size_cost): Add TARGET_ZNVER2.
1679         (enum processor_type): Add PROCESSOR_ZNVER2.
1680         * config/i386/i386.md (define_attr "cpu"): Add znver2.
1681         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
1682         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
1683         (ix86_adjust_cost): Add znver2.
1684         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
1685         * gcc/doc/extend.texi: Add details about znver2.
1686         * gcc/doc/invoke.texi: Add details about znver2.
1688 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
1690         PR target/87079
1692         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
1693         pattern.
1695 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1697         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
1698         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
1700 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
1702         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
1703         shrink-wrapping when -mtrack-speculation.
1705 2018-11-02  Richard Biener  <rguenther@suse.de>
1707         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
1708         (pop_cost_one_pair): Do not free pair.
1709         (pop_best_coalesce): Likewise.
1710         (create_coalesce_list): Initialize obstack.
1711         (delete_coalesce_list): Free obstack.
1712         (find_coalesce_pair): Obstack-allocate coalesce pairs.
1713         (add_cost_one_coalesce): Likewise.
1714         (struct live_track): Remove bitmap pointer indirections.
1715         (new_live_track): Adjust.
1716         (delete_live_track): Likewise.
1717         (live_track_remove_partition): Likewise.
1718         (live_track_add_partition): Likewise.
1719         (live_track_live_p): Likewise.
1720         (live_track_process_def): Likewise.
1721         (live_track_clear_base_vars): Likewise.
1723 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1725         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
1726         * configure: Regenerate.
1727         * config.in: Regenerate.
1728         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
1729         HAVE_AS_REGISTER_PSEUDO_OP guard.
1730         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
1731         !HAVE_AS_REGISTER_PSEUDO_OP support.
1732         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
1733         guard.
1735 2018-11-02  Richard Biener  <rguenther@suse.de>
1737         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
1738         with is_type_die.
1740 2018-11-02  Richard Biener  <rguenther@suse.de>
1742         PR tree-optimization/87776
1743         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
1744         executable when iterating but running into rpo-vn-max-loop-depth
1745         and not eliding the iteration.
1747 2018-11-30  Jan Hubicka  <jh@suse.cz>
1749         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
1750         TYPE_DECL.
1752 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
1754         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
1755         prototype.
1756         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
1757         Combine with rs6000_address_for_fpconvert.
1758         (rs6000_address_for_fpconvert): Combine with
1759         rs6000_force_indexed_or_indirect_mem.
1760         (rs6000_expand_vector_init): Change function call from
1761         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1762         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
1763         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1764         (floatsi<mode>2_lfiwax_mem): Ditto.
1765         (floatunssi<mode>2_lfiwzx): Ditto.
1766         (floatunssi<mode>2_lfiwzx_mem): Ditto.
1767         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
1768         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
1769         (fix_trunc<mode>si2_stfiwx): Ditto.
1770         (fixuns_trunc<mode>si2_stfiwx): Ditto.
1771         (float_<mode>si2_hw): Ditto.
1772         (floatuns_<mode>si2_hw): Ditto.
1773         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
1774         (vsx_splat_<mode>): Ditto.
1776 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
1778         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
1779         at end of assembler input text.
1780         * configure: Regenerate.
1782 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
1784         PR tree-optimization/87826
1785         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
1786         negative or larger or equal to type's precision.
1788 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
1790         * opts.c (default_options_table): Do not enable
1791         OPT_fdelayed_branch at -Og.
1792         * doc/invoke.texi (-fdelayed-branch): Document it.
1794 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1796         * optabs-libfuncs.c (build_libfunc_function_visibility):
1797         New, split out from...
1798         (build_libfunc_function): ... here.
1799         (init_one_libfunc_visibility): New, split out from ...
1800         (init_one_libfunc): ... here.
1802         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
1803         scratch register need not be early-clobber.  Document the reason
1804         why we cannot use ST<OP>.
1806 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
1808         PR bootstrap/82856
1809         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
1810         line for second argument of AC_DEFINE_UNQUOTED.
1811         * doc/install.texi (Tools/packages necessary for modifying GCC):
1812         Update to autoconf 2.69 and automake 1.15.1.
1813         * aclocal.m4, config.in, configure: Regenerate.
1815 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
1817         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
1818         initialization.
1820 2018-10-31  Martin Liska  <mliska@suse.cz>
1822         PR driver/83193
1823         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
1824         Add new argument invalid_extension.
1825         (aarch64_get_all_extension_candidates): New function.
1826         (aarch64_rewrite_selected_cpu): Add NULL to function call.
1827         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
1828         new argument.
1829         (aarch64_get_all_extension_candidates): New function.
1830         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
1831         argument invalid_extension.
1832         (aarch64_parse_cpu): Likewise.
1833         (aarch64_print_hint_for_extensions): New function.
1834         (aarch64_validate_mcpu): Provide hint about invalid extension.
1835         (aarch64_validate_march): Likewise.
1836         (aarch64_handle_attr_arch): Pass new argument.
1837         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
1838         (aarch64_handle_attr_isa_flags): Likewise.
1840 2018-10-31  Richard Biener  <rguenther@suse.de>
1842         PR middle-end/70359
1843         PR middle-end/86270
1844         * tree-outof-ssa.c (insert_backedge_copies): Restrict
1845         copy generation to useful cases.  Place the copy before
1846         the definition of the backedge value when possible.
1848 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1850         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
1851         * config/arc/arc.c (arc_active_insn): New function.
1852         (check_store_cacheline_hazard): Likewise.
1853         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
1854         (arc_override_options): Disable delay slot scheduler for older
1855         A7.
1856         (arc_store_addr_hazard_p): New implementation, old one renamed to
1857         ...
1858         (arc_store_addr_hazard_internal_p): Renamed.
1859         (arc_reorg): Don't combine into brcc instructions which are part
1860         of hardware hazard solution.
1861         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
1862         (tune_arc700): Likewise.
1863         * config/arc/arc.opt (arc7xx): New tune value.
1864         * config/arc/arc700.md: Improve A7 scheduler.
1866 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1868         * config/arc/arc.c (arc_override_options): Remove
1869         TARGET_COMPACT_CASESI.
1870         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
1871         (CASE_VECTOR_MODE): Likewise.
1872         (CASE_VECTOR_PC_RELATIVE): Likewise.
1873         (CASE_VECTOR_SHORTEN_MODE): Likewise.
1874         (CASE_VECTOR_SHORTEN_MODE1): Delete.
1875         (ADDR_VEC_ALIGN): Update.
1876         (ASM_OUTPUT_CASE_LABEL): Undefine.
1877         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
1878         (TARGET_BI_BIH): Define.
1879         (DEFAULT_BRANCH_INDEX): Likewise.
1880         * config/arc/arc.md (casesi): Rework to accept BI/BIH
1881         instructions, remove compact_casesi use case.
1882         (casesi_compact_jump): Remove.
1883         (casesi_dispatch): New pattern.
1884         * config/arc/arc.opt: Add mbranch-index option. Deprecate
1885         compact_casesi option.
1886         * doc/invoke.texi: Document mbranch-index option.
1888 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1890         * config/arc/arc.c (arc_get_tp): Remove function.
1891         (arc_emit_call_tls_get_addr): Likewise.
1892         (arc_call_tls_get_addr): New function.
1893         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
1894         * config/arc/arc.md (tls_load_tp_soft): Remove.
1895         (tls_gd_get_addr): Likewise.
1897 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1899         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
1900         (mulsi3_600_lib): Remove pattern.
1901         (umulsi3_highpart_600_lib_le): Likewise.
1902         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
1903         (umulsidi3): Remove call to umulsidi3_600_lib.
1904         (umulsidi3_600_lib): Remove pattern.
1905         (peephole2): Remove peephole using the above deprecated patterns.
1907 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
1909         PR target/87374
1910         * config/arm/arm.c (arm_option_check_internal): Disable the combined
1911         use of -mslow-flash-data and -mword-relocations.
1912         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
1913         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
1914         flag_pic.
1915         * doc/invoke.texi (-mword-relocations): Mention conflict with
1916         -mslow-flash-data.
1917         (-mslow-flash-data): Reciprocally.
1919 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1921         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
1922         16-byte modes held in GP registers to use an even regno.
1924         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
1925         (aarch64_atomic_ldop_supported_p): Remove.
1926         (aarch64_gen_atomic_ldop): Remove.
1927         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
1928         Fully expand LSE operations here.
1929         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
1930         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
1931         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
1932         and use ATOMIC_LDOP instead; use register_operand for the input;
1933         drop the split and emit insns directly.
1934         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
1935         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
1936         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
1938         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
1939         (aarch64_gen_atomic_ldop): Don't call it.
1940         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
1941         Use aarch64_reg_or_zero.
1942         (aarch64_atomic_exchange<ALLI>): Likewise.
1943         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
1944         operand 0; use aarch64_reg_or_zero for input; merge ...
1945         (@aarch64_atomic_swp<ALLI>): ... this and remove.
1947         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
1948         (aarch64_split_compare_and_swap): Use it.
1949         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
1950         test oldval against the proper predicate.
1951         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
1952         Use nonmemory_operand for expected.
1953         (cas_short_expected_pred): New.
1954         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
1955         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
1956         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
1957         (aarch64_plushi_operand): New.
1959         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
1960         Force oldval into the rval register for TARGET_LSE; emit the compare
1961         during initial expansion so that it may be deleted if unused.
1962         (aarch64_gen_atomic_cas): Remove.
1963         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
1964         Change =&r to +r for operand 0; use match_dup for operand 2;
1965         remove is_weak and mod_f operands as unused.  Drop the split
1966         and merge with...
1967         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
1968         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
1969         (@aarch64_atomic_cas<GPI>): Similarly.
1971 2018-10-31  Richard Biener  <rguenther@suse.de>
1973         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
1974         using ABSU_EXPR.
1976 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1978         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
1979         * config/aarch64/aarch64.md: Include saphira.md
1980         * config/aarch64/saphira.md: New file for pipeline description.
1982 2018-10-30  Martin Sebor  <msebor@redhat.com>
1984         PR middle-end/87041
1985         * gimple-ssa-sprintf.c (format_directive): Use %G to include
1986         inlining context.
1987         (sprintf_dom_walker::compute_format_length):
1988         Avoid setting POSUNDER4K here.
1989         (get_destination_size): Handle null argument values.
1990         (get_user_idx_format): New function.
1991         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
1992         functions, including user-defined with attribute format printf.
1993         Use %G to include inlining context.
1994         Set POSUNDER4K here.
1996 2018-10-30  Jan Hubicka  <jh@suse.cz>
1998         * params.def (lto-partitions): Bump from 32 to 128.
2000 2018-10-30  Jan Hubicka  <jh@suse.cz>
2002         * tree.c
2003         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2004         head in file.
2005         (free_lang_data_in_type): Forward declare.
2006         (fld_type_variant_equal_p): New function.
2007         (fld_type_variant): New function
2008         (fld_incomplete_types): New hash.
2009         (fld_incomplete_type_of): New function
2010         (fld_simplfied-type): New function.
2011         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2012         (free_lang_data): Allocate and free fld_incomplete_type; update call
2013         of free_lang_data_in_decl.
2015 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2017         * gcov.c (output_lines): Remove duplicate line.
2019 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2021         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2022         if not in indexed or indirect form.
2023         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2024         (bswapdi2_store): Ditto.
2025         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2026         helper function.
2027         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2028         Prototype for helper function.
2030 2018-10-30  Martin Sebor  <msebor@redhat.com>
2032         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2033         (target, pragma GCC optimize, pragma GCC target): Ditto.
2035 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2037         * doc/extend.texi: Fix prototype and description of
2038         __builtin_expect_with_probability.
2040 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2042         * cgraph.h (clone_function_name_1): Replaced by new
2043           clone_function_name_numbered that takes name as string; for
2044           privatize_symbol_name_1 use only.
2045           (clone_function_name): Renamed to
2046           clone_function_name_numbered to be explicit about numbering.
2047           (clone_function_name): New two-argument function that does
2048           not number its output.
2049           (clone_function_name): New three-argument function that
2050           takes a number to append to its output.
2051         * cgraphclones.c (duplicate_thunk_for_node):
2052           (clone_function_name_1): Renamed.
2053           (clone_function_name_numbered): Two new functions.
2054           (clone_function_name): Improved documentation.
2055           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2056         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2057         * final.c (final_scan_insn_1): Use the new clone_function_name
2058           without numbering.
2059         * multiple_target.c (create_dispatcher_calls): Ditto.
2060           (create_target_clone): Ditto.
2061         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2062         * omp-low.c (create_omp_child_function_name): Ditto.
2063         * omp-simd-clone.c (simd_clone_create): Ditto.
2064         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2065           new clone_function_name without numbering.
2067 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2069         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2070         Assert that the allocation size is not zero.
2072 2018-10-30  Richard Biener  <rguenther@suse.de>
2074         PR tree-optimization/87800
2075         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2076         non-induction or reduction PHIs.
2078 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2080         * config/aarch64/falkor-tag-collision-avoidance.c
2081         (execute_tag_collision_avoidance): Call df_note_add_problem.
2083 2018-10-30  Martin Liska  <mliska@suse.cz>
2085         * doc/extend.texi: Fix typo in documentation
2086         of __builtin_expect_with_probability.
2088 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2090         PR c++/87721
2091         * input.c (get_substring_ranges_for_loc): Detect if
2092         linemap_resolve_location gives us a NULL map, and reject
2093         this case.
2095 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2097         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2099 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2100             Martin Sebor  <msebor@redhat.com>
2101             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2103         * doc/ux.texi (Quoting): New subsection, adapted from material at
2104         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2105         MartinSebor and ManuelLopezIbanez.
2106         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2107         translation.
2109 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2111         PR middle-end/87469
2112         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2113         max value.
2115 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2117         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2119 2018-10-29  Paul Koning  <ni1d@arrl.net>
2121         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2123 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2125         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2126         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2127         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2128         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2129         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2130         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2131         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2132         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2133         Change 'vector' to '__vector'.
2134         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2135         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2136         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2137         _mm_avg_pu16): Likewise.  And, whitespace corrections.
2139 2018-10-29  Richard Biener  <rguenther@suse.de>
2141         PR tree-optimization/87785
2142         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2143         internal defs.
2145 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2147         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2149 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2151         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2152         ports configuration macro, defaults to "gnu".
2153         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2154         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2156 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2158         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2160 2018-10-29  Richard Biener  <rguenther@suse.de>
2162         PR tree-optimization/87790
2163         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2164         (vect_make_slp_decision): Adjust.
2165         (vect_slp_analyze_bb_1): Likewise.
2166         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2167         edges.
2169 2018-10-29  Richard Biener  <rguenther@suse.de>
2171         PR tree-optimization/87785
2172         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2173         and processing.
2174         (vect_build_slp_tree): Likewise.
2175         (vect_gather_slp_loads): New function.
2176         (vect_analyze_slp_instance): Gather loads separately from the
2177         SLP tree build.
2179 2018-10-29  Martin Liska  <mliska@suse.cz>
2181         * Makefile.in: Make dependency to json.o.
2182         * doc/gcov.texi: Document new JSON format, remove
2183         old intermediate format documentation.
2184         * gcov.c (struct function_info): Come up with m_name and
2185         m_demangled_name.
2186         (function_info::function_info): Initialize it.
2187         (function_info::~function_info): Release it.
2188         (main): Rename flag_intermediate_format to flag_json_format.
2189         (print_usage): Describe --json-format.
2190         (process_args): Set flag_json_format.
2191         (output_intermediate_line): Remove.
2192         (output_intermediate_json_line): Likewise.
2193         (get_gcov_intermediate_filename): Return new extension
2194         ".gcov.json.gz".
2195         (output_intermediate_file): Implement JSON emission.
2196         (output_json_intermediate_file): Implement JSON emission.
2197         (generate_results): Use ::get_name for function name.
2198         Handle JSON output file.
2199         (read_graph_file): Use ::get_name instead of cplus_demangle.
2200         (read_count_file): Likewise.
2201         (solve_flow_graph): Likewise.
2202         (add_line_counts): Likewise.
2203         (accumulate_line_counts): Use new flag_json_format.
2204         (output_function_details): Use ::get_name instead of cplus_demangle.
2205         (output_lines): Likewise.
2206         * json.cc (test_writing_literals): Add new tests.
2207         * json.h (class literal): Add new boolean constructor.
2209 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
2211         PR rtl-optimization/87701
2212         PR rtl-optimization/87780
2213         * combine.c (make_more_copies): Rewrite.
2215 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2217         * doc/generic.texi (ABSU_EXPR): Document.
2218         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2219         (absu(absu(X)) -> absu(X)): Likewise.
2220         (absu(-X) -> absu(X)): Likewise.
2221         (absu(X)  where X is nonnegative -> X): Likewise.
2223 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2225         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2226         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2227         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2228         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2229         (s-tm-texi): Also check timestamp on d-target.def.
2230         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2231         (build/genhooks.o): Also depend on D_TARGET_DEF.
2232         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2233         variables.
2234         * config/aarch64/aarch64-d.c: New file.
2235         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2236         Define.
2237         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2238         prototype.
2239         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2240         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2241         * config/arm/arm-d.c: New file.
2242         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2243         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2244         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2245         * config/arm/t-arm (arm-d.o): New rule.
2246         * config/default-d.c: New file.
2247         * config/glibc-d.c: New file.
2248         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2249         * config/i386/i386-d.c: New file.
2250         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2251         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2252         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2253         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2254         * config/i386/t-i386 (i386-d.o): New rule.
2255         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2256         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2257         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2258         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2259         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2260         * config/mips/mips-d.c: New file.
2261         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2262         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2263         * config/mips/t-mips (mips-d.o): New rule.
2264         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2265         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2266         * config/powerpcspe/powerpcspe-d.c: New file.
2267         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2268         New prototype.
2269         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2270         Support GNU D by using 0 as the language type.
2271         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2272         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2273         * config/riscv/riscv-d.c: New file.
2274         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2275         prototype.
2276         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2277         * config/riscv/t-riscv (riscv-d.o): New rule.
2278         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2279         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2280         * config/rs6000/rs6000-d.c: New file.
2281         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2282         prototype.
2283         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2284         Support GNU D by using 0 as the language type.
2285         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2286         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2287         * config/s390/s390-d.c: New file.
2288         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2289         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2290         * config/s390/t-s390 (s390-d.o): New rule.
2291         * config/sparc/sparc-d.c: New file.
2292         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2293         prototype.
2294         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2295         * config/sparc/t-sparc (sparc-d.o): New rule.
2296         * config/t-glibc (glibc-d.o): New rule.
2297         * configure: Regenerated.
2298         * configure.ac (tm_d_file): New variable.
2299         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2300         * doc/contrib.texi (Contributors): Add self for the D frontend.
2301         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2302         * doc/install.texi (Configuration): Mention libphobos as an option for
2303         --enable-shared.  Mention d as an option for --enable-languages.
2304         (Testing): Mention check-d as a target.
2305         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2306         name suffixes.  Mention d as a -x option.
2307         * doc/sourcebuild.texi (Top Level): Mention libphobos.
2308         * doc/standards.texi (Standards): Add section on D language.
2309         * doc/tm.texi: Regenerated.
2310         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2311         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2312         * dwarf2out.c (is_dlang): New function.
2313         (gen_compile_unit_die): Use DW_LANG_D for D.
2314         (declare_in_namespace): Return module die for D, instead of adding
2315         extra declarations into the namespace.
2316         (gen_namespace_die): Generate DW_TAG_module for D.
2317         (gen_decl_die): Handle CONST_DECLSs for D.
2318         (dwarf2out_decl): Likewise.
2319         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2320         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2321         kinds of aggregates.
2322         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2323         * genhooks.c: Include d/d-target.def.
2325 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
2327         PR target/85669
2328         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2329         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2330         (STACK_DYNAMIC_OFFSET): Likewise.
2332 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
2334         PR target/80024
2335         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2336         error message.
2338 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
2340         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2341         return type and other typos.
2343 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2345         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2346         a shorter sequence with fewer branches.
2347         (emit_final_str_compare_gpr): Ditto.
2349 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2351         * config/rs6000/tmmintrin.h: New file.
2352         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2354 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2356         * config/rs6000/mmintrin.h: Enable 32bit compilation.
2357         * config/rs6000/xmmintrin.h: Likewise.
2359 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2361         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2363 2018-10-26  Richard Biener  <rguenther@suse.de>
2365         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2366         and wrapper.
2367         (vect_mark_slp_stmts_relevant): Likewise.
2368         (vect_detect_hybrid_slp_stmts): Likewise.
2369         (vect_bb_slp_scalar_cost): Likewise.
2370         (vect_remove_slp_scalar_calls): Likewise.
2372 2018-10-26  Jan Hubicka  <jh@suse.cz>
2374         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2375         (warn_types_mismatch): Fix walk of DECL_NAME.
2376         (odr_types_equivalent_p): Fix overactive assert.
2378 2018-10-26  Richard Biener  <rguenther@suse.de>
2380         PR tree-optimization/87105
2381         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2382         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2383         refcnt.
2384         (vect_create_new_slp_node): Initialize refcnt to one.
2385         (bst_traits): Move.
2386         (scalar_stmts_set_t, bst_fail): Remove.
2387         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2388         (vect_build_slp_tree): Add bst_map argument and lookup
2389         already created SLP nodes.
2390         (vect_print_slp_tree): Handle a SLP graph, print SLP node
2391         addresses.
2392         (vect_slp_rearrange_stmts): Handle a SLP graph.
2393         (vect_analyze_slp_instance): Adjust and free SLP nodes from
2394         the CSE map.  Fix indenting.
2395         (vect_schedule_slp_instance): Add short-cut.
2397 2018-10-26  Martin Liska  <mliska@suse.cz>
2399         PR testsuite/86158
2400         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2401         addr_expr and not with pointers.
2403 2018-10-26  Jan Hubicka  <jh@suse.cz>
2405         * tree.c (free_lang_data_in_type): Only check main variants.
2406         * ipa-devirt.c (warn_odr): Make static.
2407         (types_same_for_odr): Drop strict variant.
2408         (types_odr_comparable): Likewise.
2409         (odr_or_derived_type_p): Look for main variants.
2410         (odr_name_hasher::equal): Cleanup comment.
2411         (odr_subtypes_equivalent): Add warn and warned arguments; check main
2412         variants.
2413         (type_variants_equivalent_p): break out from ...
2414         (odr_types_equivalent): ... here; go for main variants where needed.
2415         (warn_odr): ... here; turn static.
2416         (warn_types_mismatch): Compare mangled names of main variants.
2417         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2418         (type_with_linkage_p): Sanity check that we look at main variant.
2419         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2420         * tree.h (types_same_for_odr): Drop strict argument.
2422 2018-10-26  Richard Biener  <rguenther@suse.de>
2424         PR tree-optimization/87746
2425         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2426         Simplify and fix WRT strided store groups with size not
2427         equal to step in element count.
2428         (vect_analyze_group_access_1): Dump the whole group.
2430 2018-10-25  Carl Love  <cel@us.ibm.com>
2432         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2433         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2434         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
2435         precicion entry for each overloaded builtin.
2436         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2437         VSCEDPUO): Rename overloaded name.
2438         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2439         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2440         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2441         define_expand for xscmpexqp instruction.
2442         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2444 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2445             Jinsong Ji <jji@us.ibm.com>
2447         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2448         function with vec_sl.
2449         (_mm_slli_epi32): Likewise.
2450         (_mm_slli_epi64): Likewise.
2451         (_mm_srai_epi16): Replace deprecated function with vec_sra.
2452         (_mm_srai_epi32): Likewise.
2453         (_mm_srli_epi16): Replace deprecated function with vec_sr.
2454         (_mm_srli_epi32): Likewise.
2455         (_mm_srli_epi64): Likewise.
2456         (_mm_sll_epi16): Replace deprecated function with vec_sl.
2457         (_mm_sll_epi32): Likewise.
2458         (_mm_sll_epi64): Likewise.
2459         (_mm_sra_epi16): Replace deprecated function with vec_sra.
2460         (_mm_sra_epi32): Likewise.
2461         (_mm_srl_epi16): Replace deprecated function with vec_sr.
2462         (_mm_srl_epi32): Likewise.
2463         (_mm_srl_epi64): Likewise.
2465 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2466             Jinsong Ji <jji@us.ibm.com>
2468         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2469         comparison operators with vec_cmp* for compatibility due to
2470         unfortunate history; clean up formatting and use types more
2471         appropriately.
2472         (_mm_sll_epi32): Likewise.
2473         (_mm_sll_epi64): Likewise.
2474         (_mm_srl_epi16): Likewise.
2475         (_mm_srl_epi32): Likewise.
2476         (_mm_srl_epi64): Likewise.
2478 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2479             Jinsong Ji <jji@us.ibm.com>
2481         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2482         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2483         __vector __bool int.  Use vec_cmpgt in preference to deprecated
2484         function vec_vcmpgtfp.
2485         (_mm_max_ps): Likewise.
2487 2018-10-25  Jeff Law  <law@redhat.com>
2489         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2490         if returning false.
2492 2018-10-25  Martin Sebor  <msebor@redhat.com>
2494         * doc/extend.texi (aligned): Expand attribute description.
2495         (Alignment): Rename section.  Discuss function arguments.
2497 2018-10-25  Jan Hubicka  <jh@suse.cz>
2499         * ipa-devirt.c (main_odr_variant): Remove.
2500         (hash_odr_name, types_same_for_odr, types_odr_comparable,
2501         odr_name_hasher::equal, odr_subtypes_equivalent_p):
2502         Drop use of main_odr_variant.
2503         (add_type_duplicate): Silence confused warnings on integer types.
2504         (get_odr_type): Always look for main variant.
2505         (register_odr_type): Simplify.
2507 2018-10-25  Richard Biener  <rguenther@suse.de>
2509         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2510         Initialize ng to silence error with release checking bootstrap.
2512 2018-10-25  Richard Biener  <rguenther@suse.de>
2514         * tree-if-conv.c: Include tree-ssa-sccvn.h.
2515         (tree_if_conversion): Run CSE on the if-converted loop body.
2517 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2519         * config/s390/constraints.md (ZL): New constraint.
2520         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2521         operands.
2522         * config/s390/s390.md (movdi_larl): Remove.
2523         (movdi_64): Add the LARL alternative.
2525 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2527         PR bootstrap/87747
2528         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
2529         (RTX_CODE_HWINT_P): New macro.
2530         (rtx_code_size): Use RTX_CODE_HWINT_P ().
2532 2018-10-25  Jan Hubicka  <jh@suse.cz>
2534         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
2535         is anonymous.
2537 2018-10-25  Richard Biener  <rguenther@suse.de>
2539         PR tree-optimization/87665
2540         PR tree-optimization/87745
2541         * tree-vectorizer.h (get_earlier_stmt): Remove.
2542         (get_later_stmt): Pick up UID from the original non-pattern stmt.
2544 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
2546         * options.texi (Deprecated): Move list to Var section.
2548 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
2549             Jinsong Ji <jji@us.ibm.com>
2551         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
2552         __vector long to __vector long long.
2553         (_mm_cvtpd_ps): Likewise.
2554         (_mm_cvttpd_epi32): Likewise.
2555         (_mm_cvtpi32_pd): Likewise.
2556         (_mm_unpackhi_epi64): Likewise.
2557         (_mm_unpacklo_epi64): Likewise.
2559 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
2561         PR rtl-optimization/87720
2562         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
2564 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
2566         * gimple-ssa-isolate-paths.c
2567         (find_implicit_erroneous_behavior): Do not change code if the
2568         pass is running for warnings only.
2569         (find_explicit_erroneous_behavior): Likewise.
2571 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
2573         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
2574         Define as rs6000_mangle_decl_assembler_name.
2575         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
2576         long double to IEEE long double, switch the names of the long
2577         double built-in functions to be <func>f128 instead of <func>l.
2579 2018-10-24  Martin Sebor  <msebor@redhat.com>
2581         * doc/extend.texi (nonnull): List no-argument form.  Reference
2582         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
2584 2018-10-24  Richard Biener  <rguenther@suse.de>
2586         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
2588 2018-10-24  Martin Liska  <mliska@suse.cz>
2590         PR tree-optimization/84436
2591         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
2592         Remove.
2593         (switch_conversion::contains_linear_function_p): New.
2594         (switch_conversion::build_one_array): Support linear
2595         transformation on input.
2596         * tree-switch-conversion.h (struct switch_conversion): Add
2597         contains_linear_function_p declaration.
2599 2018-10-24  Richard Biener  <rguenther@suse.de>
2601         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
2602         if its argument is CONSTANT_CLASS_P.
2604 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2606         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
2607         it is wrong for forward declarations.
2609 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2611         * config/s390/s390.c (s390_check_qrst_address): Add the missing
2612         SYMBOL_REF_P () check.
2614 2018-10-24  Richard Biener  <rguenther@suse.de>
2616         PR tree-optimization/87105
2617         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
2618         dump classification.
2619         (vect_analyze_data_ref_accesses): Handle duplicate loads and
2620         stores by splitting the affected group after the fact.
2621         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
2622         fail the SLP build because of size constraints.
2624 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2626         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
2627         * configure: Regenerate.
2628         * config.in: Regenerate.
2629         * varasm.c (mergeable_string_section): Use readonly_data_section
2630         if linker doesn't support SHF_MERGE with alignment > 8.
2631         (mergeable_constant_section): Likewise.
2633 2018-10-24  Richard Biener  <rguenther@suse.de>
2635         PR tree-optimization/84013
2636         * tree-ssa-structalias.c (struct msdi_data): New struct for
2637         marshalling data to walk_stmt_load_store_ops.
2638         (maybe_set_dependence_info): Refactor as callback for
2639         walk_stmt_load_store_ops.
2640         (compute_dependence_clique): Set restrict info on all stmt kinds.
2642 2018-10-24  Martin Liska  <mliska@suse.cz>
2644         * cgraph.c (cgraph_node::dump):
2645         Remove reduntant dumps and make tp_first_run dump more compact.
2647 2018-10-24  Richard Biener  <rguenther@suse.de>
2649         PR tree-optimization/87665
2650         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2651         to reflect reality.
2653 2018-10-12  Jeff Law  <law@redhat.com>
2655         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
2656         for H8/S.
2658 2018-10-23  Richard Biener  <rguenther@suse.de>
2660         * tree-vrp.c (add_assert_info): Guard dump_printf with
2661         dump_enabled_p.
2662         * gimple-ssa-evrp-analyze.c
2663         (evrp_range_analyzer::record_ranges_from_incoming_edge):
2664         Use value_range::ignore_equivs_equal_p.
2666 2018-10-23  Richard Biener  <rguenther@suse.de>
2668         PR tree-optimization/87105
2669         PR tree-optimization/87608
2670         * passes.def (pass_all_early_optimizations): Add early phi-opt
2671         after dce.
2672         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
2673         addition to debug stmts.
2674         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
2675         and abs replacement early.
2676         * tree-cfg.c (gimple_empty_block_p): Likewise.
2678 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
2680         PR target/86383
2681         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
2682         specified to configure.
2683         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
2685 2018-10-23  Richard Biener  <rguenther@suse.de>
2687         PR tree-optimization/87700
2688         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2690 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
2692         PR target/87674
2693         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
2694         second argument from __mmask16 to __mmask8.
2695         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
2696         _mm_mask_packs_epi32): Likewise.
2697         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
2698         Likewise.
2699         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
2701 2018-10-23  Richard Biener  <rguenther@suse.de>
2703         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
2705 2018-10-23  Richard Biener  <rguenther@suse.de>
2707         PR tree-optimization/86144
2708         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
2709         over simd attribute.
2711 2018-10-23  Richard Biener  <rguenther@suse.de>
2713         PR tree-optimization/87693
2714         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
2715         the case we do not find the taken edge.
2717 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
2718             Jinsong Ji  <jji@us.ibm.com>
2720         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
2721         (_mm_store_pd): Use unaligned vector type for pointer cast.
2722         (_mm_maskmoveu_si128): Likewise.
2723         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
2724         (_mm_store_ps): Use unaligned vector type for pointer cast.
2726 2018-10-22  Paul Koning  <ni1d@arrl.net>
2728         * symtab.c (symtab_node::increase_alignment): Correct max
2729         alignment check.
2731 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
2733         PR tree-optimization/87633
2734         * match.pd: Do not generate unordered integer comparisons.
2736 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2738         PR rtl-optimization/87600
2739         * combine.c: Add include of expr.h.
2740         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
2741         register to a pseudo.
2742         (make_more_copies): New function, add a copy to a new pseudo after
2743         the moves from hard registers into pseudos.
2744         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
2745         later.  Call make_more_copies.
2747 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
2749         * lra-constraints.c (process_alt_operands): New local array,
2750         matching_early_clobber.  Check matching_early_clobber before
2751         decrementing reject, and set matching_early_clobber after.
2753 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2755         PR target/87598
2756         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
2757         call output_addr_const and hope for the best.
2759 2018-10-22  Richard Biener  <rguenther@suse.de>
2761         * gimple-ssa-evrp-analyze.c
2762         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
2763         smarter about what ranges to use.
2764         * tree-vrp.c (add_assert_info): Dump here.
2765         (register_edge_assert_for_2): Instead of here at multiple but
2766         not all places.
2768         * gcc.dg/tree-ssa/evrp12.c: New testcase.
2769         * gcc.dg/predict-6.c: Adjust.
2770         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
2771         * gcc.dg/tree-ssa/vrp02.c: Likewise.
2772         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
2774 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
2775         Richard Biener  <rguenther@suse.de>
2777         * bitmap.h: Update data structure documentation, including a
2778         description of bitmap views as either linked-lists or splay trees.
2779         (struct bitmap_element_def): Update comments for splay tree bitmaps.
2780         (struct bitmap_head_def): Likewise.
2781         (bitmap_list_view, bitmap_tree_view): New prototypes.
2782         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
2783         tree_form fields.
2784         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
2785         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
2786         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
2787         released bitmap element here.
2788         (bitmap_element_free): Remove.
2789         (bitmap_elt_clear_from): Work on splay tree bitmaps.
2790         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
2791         this function similar ones such that linked-list bitmap implementation
2792         functions are grouped.
2793         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
2794         and moved for grouping.
2795         (bitmap_list_insert_element_after): Renamed from
2796         bitmap_elt_insert_after, and moved for grouping.
2797         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
2798         (bitmap_tree_link_left, bitmap_tree_link_right,
2799         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
2800         bitmap_tree_link_element, bitmap_tree_unlink_element,
2801         bitmap_tree_find_element): New functions for splay-tree bitmap
2802         implementation.
2803         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
2804         Renamed and moved, see above entries.
2805         (bitmap_tree_listify_from): New function to convert part of a splay
2806         tree bitmap to a linked-list bitmap.
2807         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
2808         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
2809         (bitmap_find_bit): Remove.
2810         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
2811         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
2812         Handle splay tree bitmaps.
2813         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
2814         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
2815         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
2816         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
2817         bitmap_intersect_compl_p, bitmap_ior_and_compl,
2818         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
2819         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
2820         corresponding changes to use linked-list specific bitmap_element
2821         manipulation functions as applicable for efficiency.
2822         (bitmap_tree_to_vec): New function.
2823         (debug_bitmap_elt_file): New function split out from ...
2824         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
2825         (bitmap_print): Likewise.
2827         PR tree-optimization/63155
2828         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
2829         SSA edge worklists.
2830         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
2831         in tree-view.
2833 2018-10-22  Martin Liska  <mliska@suse.cz>
2835         PR tree-optimization/87686
2836         Revert
2837         2018-08-29  Martin Liska  <mliska@suse.cz>
2839         * tree-switch-conversion.c (switch_conversion::expand):
2840         Strenghten assumption about gswitch statements.
2842 2018-10-22  Martin Liska  <mliska@suse.cz>
2844         * ipa-icf.c (sem_item::compare_attributes): Remove.
2845         (sem_item::compare_referenced_symbol_properties): Use
2846         attribute_list_equal instead.
2847         (sem_function::equals_wpa): Likewise.
2848         * ipa-icf.h: Remove compare_attributes.
2850 2018-10-22  Richard Biener  <rguenther@suse.de>
2852         PR middle-end/87682
2853         * mem-stats.h (mem_usage::operator==): Fix pasto.
2855 2018-10-22  Richard Biener  <rguenther@suse.de>
2857         PR tree-optimization/87640
2858         * tree-vrp.c (set_value_range_with_overflow): Decompose
2859         incomplete result.
2860         (extract_range_from_binary_expr_1): Adjust.
2862 2018-10-22  Martin Jambor  <mjambor@suse.cz>
2864         * tree-eh.h (stmt_could_throw_p): Add function parameter.
2865         (stmt_can_throw_external): Likewise.
2866         (stmt_can_throw_internal): Likewise.
2867         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
2868         (lower_eh_constructs_2): Likewise.
2869         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
2870         (stmt_can_throw_external): Likewise.
2871         (stmt_can_throw_internal): Likewise.
2872         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
2873         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
2874         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
2875         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
2876         (pass_lower_eh_dispatch::execute): Pass cfun to
2877         stmt_can_throw_external.
2878         (cleanup_empty_eh): Likewise.
2879         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
2880         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
2881         stmt_can_throw_external instead of pushing it to cfun.
2882         (symbol_table::create_edge): Likewise.
2883         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
2884         stmt_can_throw_internal.
2885         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
2886         to stmt_could_throw_p.
2887         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
2888         stmt_can_throw_internal.
2889         (pass_store_merging::execute): Likewise.
2890         * gimple-ssa-strength-reduction.c
2891         (find_candidates_dom_walker::before_dom_children): Pass cfun to
2892         stmt_could_throw_p.
2893         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
2894         stmt_can_throw_internal.
2895         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
2896         to stmt_can_throw_external.
2897         (check_stmt): Pass cfun to stmt_could_throw_p.
2898         (check_stmt): Pass cfun to stmt_can_throw_external.
2899         (pass_nothrow::execute): Likewise.
2900         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
2901         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
2902         stmt_can_throw_internal.
2903         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
2904         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
2905         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
2906         * tree-complex.c (expand_complex_libcall): Pass cfun to
2907         stmt_could_throw_p and to stmt_can_throw_internal.
2908         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
2909         * tree-inline.c (copy_edges_for_bb): Likewise.
2910         (maybe_move_debug_stmts_to_successors): Likewise.
2911         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
2912         stmt_could_throw_p.
2913         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
2914         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
2915         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
2916         stmt_can_throw_internal.
2917         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2918         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
2919         stmt_could_throw_p.
2920         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
2921         stmt_can_throw_internal.
2922         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
2923         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
2924         stmt_could_throw_p.
2925         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
2926         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
2927         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2928         (convert_mult_to_fma_1): Likewise.
2929         (convert_to_divmod): Likewise.
2930         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
2931         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
2932         * tree-ssa-propagate.c
2933         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
2934         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
2935         (maybe_optimize_range_tests): Likewise.
2936         (linearize_expr_tree): Likewise.
2937         (reassociate_bb): Likewise.
2938         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2939         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
2940         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
2941         (handle_char_store): Likewise.
2942         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
2943         stmt_can_throw_internal.
2944         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
2945         stmt_could_throw_p.
2946         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2947         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
2948         (vectorizable_simd_clone_call): Likewise.
2949         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
2950         (gimple_stringop_fixed_value): Likewise.
2952 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
2954         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
2955         literal pool references.
2956         (s390_check_qrst_address): Adapt to the new behavior of
2957         s390_loadrelative_operand_p ().
2959 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2961         PR target/72782
2962         * config/i386/sse.md (*andnot<mode>3_bcst): New.
2964 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2966         PR target/72782
2967         * config/i386/sse.md (*<code><mode>3_bcst): New.
2969 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2971         PR target/72782
2972         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
2973         V4DI, V16SI and V8DI.
2974         (*sub<mode>3<mask_name>_bcst): New.
2975         (*add<mode>3<mask_name>_bcst): Likewise.
2977 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
2978             Jinsong Ji  <jji@us.ibm.com>
2980         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
2981         __m64 with __vector unsigned long long for compatibility.
2982         (_mm_movemask_epi8): Likewise.
2983         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
2984         (_mm_cvttps_pi32): Likewise.
2985         (_mm_cvtpi32_ps): Likewise.
2986         (_mm_cvtps_pi16): Likewise.
2987         (_mm_loadh_pi): Likewise.
2988         (_mm_storeh_pi): Likewise.
2989         (_mm_movehl_ps): Likewise.
2990         (_mm_movelh_ps): Likewise.
2991         (_mm_loadl_pi): Likewise.
2992         (_mm_storel_pi): Likewise.
2993         (_mm_movemask_ps): Likewise.
2994         (_mm_shuffle_pi16): Likewise.
2996 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2998         PR target/72782
2999         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3000         __builtin_ia32_vfnmsubpd512_mask.
3001         (_mm512_mask_fnmsub_round_pd): Likewise.
3002         (_mm512_fnmsub_pd): Likewise.
3003         (_mm512_mask_fnmsub_pd): Likewise.
3004         (_mm512_maskz_fnmsub_round_pd): Use
3005         __builtin_ia32_vfnmsubpd512_maskz.
3006         (_mm512_maskz_fnmsub_pd): Likewise.
3007         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3008         (_mm512_mask_fnmsub_round_ps): Likewise.
3009         (_mm512_fnmsub_ps): Likewise.
3010         (_mm512_mask_fnmsub_ps): Likewise.
3011         (_mm512_maskz_fnmsub_round_ps): Use
3012         __builtin_ia32_vfnmsubps512_maskz.
3013         (_mm512_maskz_fnmsub_ps): Likewise.
3014         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3015         __builtin_ia32_vfnmsubpd256_mask.
3016         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3017         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3018         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3019         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3020         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3021         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3022         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3023         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3024         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3025         __builtin_ia32_vfnmsubpd.
3026         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3027         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3028         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3029         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3030         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3031         * config/i386/i386-builtin.def: Add
3032         __builtin_ia32_vfnmsubpd256_mask,
3033         __builtin_ia32_vfnmsubpd256_maskz,
3034         __builtin_ia32_vfnmsubpd128_mask,
3035         __builtin_ia32_vfnmsubpd128_maskz,
3036         __builtin_ia32_vfnmsubps256_mask,
3037         __builtin_ia32_vfnmsubps256_maskz,
3038         __builtin_ia32_vfnmsubps128_mask,
3039         __builtin_ia32_vfnmsubps128_maskz,
3040         __builtin_ia32_vfnmsubpd512_mask,
3041         __builtin_ia32_vfnmsubpd512_maskz,
3042         __builtin_ia32_vfnmsubps512_mask,
3043         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3044         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3045         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3046         __builtin_ia32_vfnmsubpd256.
3047         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3048         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3049         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3050         Likewise.
3051         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3052         Likewise.
3053         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3054         Likewise.
3055         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3057 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3059         PR target/72782
3060         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3061         __builtin_ia32_vfnmaddpd512_mask.
3062         (_mm512_mask_fnmadd_round_pd): Likewise.
3063         (_mm512_fnmadd_pd): Likewise.
3064         (_mm512_mask_fnmadd_pd): Likewise.
3065         (_mm512_maskz_fnmadd_round_pd): Use
3066         __builtin_ia32_vfnmaddpd512_maskz.
3067         (_mm512_maskz_fnmadd_pd): Likewise.
3068         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3069         (_mm512_mask_fnmadd_round_ps): Likewise.
3070         (_mm512_fnmadd_ps): Likewise.
3071         (_mm512_mask_fnmadd_ps): Likewise.
3072         (_mm512_maskz_fnmadd_round_ps): Use
3073         __builtin_ia32_vfnmaddps512_maskz.
3074         (_mm512_maskz_fnmadd_ps): Likewise.
3075         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3076         __builtin_ia32_vfnmaddpd256_mask.
3077         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3078         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3079         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3080         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3081         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3082         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3083         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3084         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3085         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3086         __builtin_ia32_vfnmaddpd.
3087         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3088         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3089         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3090         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3091         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3092         * config/i386/i386-builtin.def: Add
3093         __builtin_ia32_vfnmaddpd256_mask,
3094         __builtin_ia32_vfnmaddpd256_maskz,
3095         __builtin_ia32_vfnmaddpd128_mask,
3096         __builtin_ia32_vfnmaddpd128_maskz,
3097         __builtin_ia32_vfnmaddps256_mask,
3098         __builtin_ia32_vfnmaddps256_maskz,
3099         __builtin_ia32_vfnmaddps128_mask,
3100         __builtin_ia32_vfnmaddps128_maskz,
3101         __builtin_ia32_vfnmaddpd512_mask,
3102         __builtin_ia32_vfnmaddpd512_maskz,
3103         __builtin_ia32_vfnmaddps512_mask,
3104         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3105         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3106         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3107         __builtin_ia32_vfnmaddpd256.
3108         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3109         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3110         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3111         Likewise.
3112         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3113         Likewise.
3114         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3115         Likewise.
3116         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3118 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3120         PR target/72782
3121         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3122         __builtin_ia32_vfmsubpd512_mask.
3123         (_mm512_mask_fmsub_round_pd): Likewise.
3124         (_mm512_fmsub_pd): Likewise.
3125         (_mm512_mask_fmsub_pd): Likewise.
3126         (_mm512_maskz_fmsub_round_pd): Use
3127         __builtin_ia32_vfmsubpd512_maskz.
3128         (_mm512_maskz_fmsub_pd): Likewise.
3129         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3130         (_mm512_mask_fmsub_round_ps): Likewise.
3131         (_mm512_fmsub_ps): Likewise.
3132         (_mm512_mask_fmsub_ps): Likewise.
3133         (_mm512_maskz_fmsub_round_ps): Use
3134         __builtin_ia32_vfmsubps512_maskz.
3135         (_mm512_maskz_fmsub_ps): Likewise.
3136         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3137         __builtin_ia32_vfmsubpd256_mask.
3138         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3139         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3140         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3141         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3142         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3143         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3144         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3145         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3146         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3147         __builtin_ia32_vfmsubpd.
3148         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3149         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3150         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3151         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3152         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3153         * config/i386/i386-builtin.def: Add
3154         __builtin_ia32_vfmsubpd256_mask,
3155         __builtin_ia32_vfmsubpd256_maskz,
3156         __builtin_ia32_vfmsubpd128_mask,
3157         __builtin_ia32_vfmsubpd128_maskz,
3158         __builtin_ia32_vfmsubps256_mask,
3159         __builtin_ia32_vfmsubps256_maskz,
3160         __builtin_ia32_vfmsubps128_mask,
3161         __builtin_ia32_vfmsubps128_maskz,
3162         __builtin_ia32_vfmsubpd512_mask,
3163         __builtin_ia32_vfmsubpd512_maskz,
3164         __builtin_ia32_vfmsubps512_mask,
3165         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3166         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3167         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3168         __builtin_ia32_vfmsubpd256.
3169         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3170         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3171         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3172         Likewise.
3173         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3174         Likewise.
3175         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3176         Likewise.
3177         (fmai_vmfmsub_<mode><round_name>): Likewise.
3179 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3181         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3182         Remove plus.  Renamed to ...
3183         (*sub<mode>3<mask_name>_bcst): This.
3184         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3185         (*add<mode>3<mask_name>_bcst): This.
3187 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3189         PR target/72782
3190         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3192 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3194         PR target/87662
3195         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3196         (_mm_or_epi32): Likewise.
3197         (_mm256_xor_epi32): Likewise.
3198         (_mm_xor_epi32): Likewise.
3199         (_mm256_or_epi64): Likewise.
3200         (_mm_or_epi64): Likewise.
3201         (_mm256_xor_epi64): Likewise.
3202         (_mm_xor_epi64): Likewise.
3204 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
3206         PR target/72782
3207         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3209 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
3211         PR middle-end/87647
3212         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3214 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3216         * doc/ux.texi: Move @section directly after @node.
3218 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3220         PR middle-end/85488
3221         PR middle-end/87649
3222         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3223         depend closely nested inside of loop with ordered clause with
3224         a parameter.
3226 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3228         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3229         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3230         * doc/ux.texi: New file.
3232 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3234         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3235         be the first CR field allocated.
3237 2018-10-19  Richard Biener  <rguenther@suse.de>
3239         PR target/87657
3240         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3241         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3243 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3245         PR target/72782
3246         * config/i386/sse.md
3247         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3248         (*add<mode>3<mask_name>_bcst_2): Likewise.
3250 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3252         * config/i386/sse.md
3253         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3254         Replace nonimmediate_operand with register_operand.
3255         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3256         Likewise.
3257         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3258         Likewise.
3260 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3262         PR rtl-optimization/87596
3263         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3264         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3265         for instructions in FROM..TO range.
3267 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
3269         * cfgexpand.c (expand_one_var): Use specific wording in error message
3270         for non-local frame variables.
3271         * stor-layout.c (layout_decl): Do not issue a warning for them.
3273 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
3275         * haifa-sched.c (priority): Add force_recompute parameter.
3276         (apply_replacement): Call priority () with force_recompute = true.
3277         (restore_pattern): Likewise.
3279 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3281         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3282         HOST_BITS_PER_WIDE_INT.
3283         (test_vector_ops_duplicate): Likewise.
3285 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3287         PR target/72782
3288         * config/i386/sse.md (VF_AVX512): New.
3289         (avx512bcst): Likewise.
3290         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3291         Likewise.
3292         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3293         Likewise.
3294         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3295         Likewise.
3297 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
3299         * doc/invoke.texi (-dumpversion): Improve grammar.
3300         (-dumpfullversion): Make more consistent with -dumpversion.
3302 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
3304         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3305         Set JUMP_LABEL to the jump insn.
3306         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3307         Predict emitted jump and add label to jump insn.
3309 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
3311         PR tree-optimization/87562
3312         * input.c (get_substring_ranges_for_loc): Use
3313         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3314         getting the linemap for the endpoint.  Verify that it's either
3315         in the same linemap as the start point's spelling location, or
3316         at least in the same file.
3318 2018-10-18  Richard Biener  <rguenther@suse.de>
3320         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3321         feed width-specific load/store costs through ix86_vec_cost.
3322         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3323         (k8_cost): Likewise.
3324         (bdver_cost): Likewise.
3325         (znver1_cost): Likewise.
3326         (btver1_cost): Likewise.
3327         (btver2_cost): Likewise.
3329 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3331         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3332         to simplify subreg of vec_merge.
3334 2018-10-18  Richard Biener  <rguenther@suse.de>
3336         * config/i386/i386.c: Fix costing of vector FMA.
3338 2018-10-18  Richard Biener  <rguenther@suse.de>
3340         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3341         and argument.
3342         (ix86_builtin_vectorization_cost): For vec_construct properly
3343         cost insertion into SSE regs.
3344         (...): Adjust calls to ix86_vec_cost.
3346 2018-10-18  Richard Biener  <rguenther@suse.de>
3348         PR middle-end/87087
3349         Revert
3350         2018-02-07  Richard Biener  <rguenther@suse.de>
3352         PR tree-optimization/84204
3353         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3354         this place.
3356 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3358         PR target/87537
3359         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3360         of vec_duplicate.
3361         (test_vector_ops_duplicate): Add test for a scalar subreg of a
3362         VEC_MERGE of a VEC_DUPLICATE.
3364 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3366         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3367         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3368         * doc/standards.texi (C Language): Document C2X.
3369         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3370         (rl78_option_override): Handle "GNU C2X" language name.
3372 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3374         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3375         Document C17 as published in 2018.
3377 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3379         PR middle-end/87623
3380         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3381         bail out if both sides do not have the same storage order.
3383 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3385         * bitmap.c (bitmap_head::dump): New.
3386         * bitmap.h (bitmap_head): Add dump().
3387         * gimple-ssa-evrp-analyze.c
3388         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3389         (evrp_range_analyzer::set_ssa_range_info): Same.
3390         (evrp_range_analyzer::record_ranges_from_phis): Same.
3391         (evrp_range_analyzer::record_ranges_from_stmt): Same.
3392         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3393         * gimple-ssa-sprintf.c (get_int_range): Same.
3394         (format_integer): Same.
3395         (sprintf_dom_walker::handle_gimple_call): Same.
3396         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3397         (ipcp_vr_lattice::top_p): Same.
3398         (ipcp_vr_lattice::bottom_p): Same.
3399         (ipcp_vr_lattice::set_to_bottom): Same.
3400         (ipa_vr_operation_and_type_effects): Same.
3401         (propagate_vr_across_jump_function): Same.
3402         (ipcp_store_vr_results): Same.
3403         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3404         (ipa_print_node_jump_functions_for_edge): Same.
3405         (ipa_get_value_range): Same.
3406         (ipa_compute_jump_functions_for_edge): Same.
3407         (ipa_write_jump_function): Same.
3408         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3409         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3410         Same.
3411         * vr-values.c (set_value_range_to_nonnegative): Same.
3412         (set_value_range_to_truthvalue): Same.
3413         (vr_values::get_value_range): Same.
3414         (vr_values::set_defs_to_varying): Same.
3415         (vr_values::update_value_range): Same.
3416         (symbolic_range_based_on_p): Same.
3417         (vr_values::op_with_boolean_value_range_p): Same.
3418         (vr_values::extract_range_for_var_from_comparison_expr): Same.
3419         (vr_values::extract_range_from_ssa_name): Same.
3420         (vr_values::extract_range_from_binary_expr): Same.
3421         (vr_values::extract_range_from_unary_expr): Same.
3422         (vr_values::extract_range_from_cond_expr): Same.
3423         (vr_values::extract_range_from_comparison): Same.
3424         (vr_values::check_for_binary_op_overflow): Same.
3425         (vr_values::extract_range_basic): Same.
3426         (vr_values::extract_range_from_assignment): Same.
3427         (compare_ranges): Same.
3428         (compare_range_with_value): Same.
3429         (vr_values::adjust_range_with_scev): Same.
3430         (vrp_valueize): Same.
3431         (vrp_valueize_1): Same.
3432         (vr_values::get_vr_for_comparison): Same.
3433         (vr_values::compare_name_with_value): Same.
3434         (vr_values::compare_names): Same.
3435         (vr_values::vrp_evaluate_conditional): Same.
3436         (find_case_label_ranges): Same.
3437         (vr_values::vrp_visit_switch_stmt): Same.
3438         (vr_values::extract_range_from_phi_node): Same.
3439         (vr_values::simplify_div_or_mod_using_ranges): Same.
3440         (vr_values::simplify_bit_ops_using_ranges): Same.
3441         (test_for_singularity): Same.
3442         (range_fits_type_p): Same.
3443         (vr_values::simplify_cond_using_ranges_1): Same.
3444         (vr_values::simplify_switch_using_ranges): Same.
3445         (vr_values::simplify_float_conversion_using_ranges): Same.
3446         (vr_values::two_valued_val_range_p): Same.
3447         (vr_values::add_equivalence): Move to value_range::equiv_add.
3448         * vr-values.h (vr_values::add_equivalence): Remove.
3449         (VR_INITIALIZER): Remove.
3450         * tree-vrp.c (value_range::set): New.
3451         (value_range::equiv_add): New.
3452         (value_range::value_range): New.
3453         (value_range::deep_copy): New.
3454         (value_range::check): New.
3455         (value_range::equal_p): New.
3456         (value_range::ignore_equivs_equal_p): New.
3457         (value_range::operator==): New.
3458         (value_range::operator!=): New.
3459         (value_range::symbolic_p): New.
3460         (value_range::numeric_p): New.
3461         (value_range::set_undefined): New.
3462         (value_range::set_varying): New.
3463         (value_range::may_contain_p): New.
3464         (value_range::equiv_clear): New.
3465         (value_range::singleton_p): New.
3466         (value_range::intersect): New.
3467         (value_range::dump): New.
3468         (value_range::set_and_canonicalize): New.
3469         (set_value_range): Adjust for value_range API.
3470         (set_value_range_to_undefined): Same.
3471         (set_value_range_to_varying): Same.
3472         (set_and_canonicalize_value_range): Same.
3473         (set_value_range_to_nonnull): Same.
3474         (set_value_range_to_null): Same.
3475         (range_is_null): Same.
3476         (range_is_nonnull): Same.
3477         (range_int_cst_p): Same.
3478         (range_int_cst_singleton_p): Same.
3479         (symbolic_range_p): Same.
3480         (range_includes_zero_p): Same.
3481         (value_range_constant_singleton): Same.
3482         (vrp_set_zero_nonzero_bits): Same.
3483         (ranges_from_anti_range): Same.
3484         (extract_range_into_wide_ints): Same.
3485         (extract_range_from_multiplicative_op): Same.
3486         (set_value_range_with_overflow): Same.
3487         (extract_range_from_binary_expr_1): Same.
3488         (extract_range_from_unary_expr): Same.
3489         (dump_value_range): Same.
3490         (debug_value_range): Same.
3491         (vrp_prop::check_array_ref): Same.
3492         (vrp_prop::check_mem_ref): Same.
3493         (vrp_prop::vrp_initialize): Same.
3494         (vrp_prop::visit_stmt): Same.
3495         (intersect_ranges): Same.
3496         (vrp_prop::visit_phi): Same.
3497         (vrp_prop::vrp_finalize): Same.
3498         (determine_value_range_1): Same.
3499         (determine_value_range): Same.
3500         (vrp_intersect_ranges_1): Rename to...
3501         (vrp_intersect_1): this.
3502         (vrp_intersect_ranges): Rename to...
3503         (value_range::intersect_helper): ...this.
3504         (vrp_meet_1): Rename to...
3505         (value_range::union_helper): ...this.
3506         (vrp_meet): Rename to...
3507         (value_range::union_): ...this.
3508         (copy_value_range): Remove.
3509         * tree-vrp.h (struct value_range): Rewrite into a proper class.
3510         (value_range::vrtype): New.
3511         (value_range::type): New.
3512         (value_range::equiv): New.
3513         (value_range::min): New.
3514         (value_range::max): New.
3515         (value_range::varying_p): New.
3516         (value_range::undefined_p): New.
3517         (value_range::null_p): New.
3518         (value_range::equiv_add): New.
3519         (copy_value_range): Remove.
3521 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
3523         * Makefile.in (SELFTEST_TARGETS): New.
3524         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3525         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
3526         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
3527         c/Make-lang.in.
3528         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
3529         (selftest-c++-gdb, selftest-c++-valgrind): Move to
3530         cp/Make-lang.in.
3531         * configure: Regenerate.
3532         * configure.ac (selftest_languages): New.
3534 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3536         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
3537         overflow wraps argument.
3538         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
3539         wide_int_range_multiplicative_op.
3540         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
3541         overflow wraps argument.
3542         (wide_int_range_multiplicative_op): Same.
3543         (wide_int_range_lshift): Same.
3544         (wide_int_range_div): Same.
3545         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
3546         (wide_int_range_lshift): Same.
3547         (wide_int_range_div): Same.
3549 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3551         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
3552         use sign as argument.
3553         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
3554         wide_int_range_shift_undefined_p.
3556 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3558         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
3559         Rename to...
3560         (@despeculate_copy<ALLI_TI:mode>): ... This.
3561         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
3562         switch statement.
3564 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3566         * config.gcc: Obsolete *-*-solaris2.10*.
3567         * doc/install.texi (Specific, *-*-solaris2*): Document it.
3569 2018-10-12  Jeff Law  <law@redhat.com>
3571         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
3572         reg + sym +- const_int addressing modes.
3574 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3576         * common.opt (fdiagnostics-minimum-margin-width=): New option.
3577         * diagnostic-show-locus.c (layout::layout): Apply the minimum
3578         margin width.
3579         (layout::start_annotation_line): Only print up to 3 of the
3580         margin character, to avoid touching the left-hand side.
3581         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
3582         minimum margin width, as set by test_diagnostic_context's ctor.
3583         (selftest::test_fixit_insert_containing_newline): Likewise.
3584         (selftest::test_fixit_insert_containing_newline_2): Likewise.
3585         (selftest::test_line_numbers_multiline_range): Clear
3586         dc.min_margin_width.
3587         * diagnostic.c (diagnostic_initialize): Initialize
3588         min_margin_width.
3589         * diagnostic.h (struct diagnostic_context): Add field
3590         "min_margin_width".
3591         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
3592         * opts.c (common_handle_option): Handle
3593         OPT_fdiagnostics_minimum_margin_width_.
3594         * selftest-diagnostic.c
3595         (selftest::test_diagnostic_context::test_diagnostic_context):
3596         Initialize min_margin_width to 6.
3597         * toplev.c (general_init): Initialize global_dc->min_margin_width.
3599 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3601         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3602         Fix usage of "error_at_rich_loc" in the comment.
3604 2018-10-15  Renlin Li  <renlin.li@arm.com>
3606         PR target/87563
3607         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
3608         if-conversioned loop when it contains ifn with types not
3609         supported by backend.
3610         * internal-fn.c (expand_direct_optab_fn): Add an assert.
3611         (direct_internal_fn_supported_p): New helper function.
3612         * internal-fn.h (direct_internal_fn_supported_p): Declare.
3614 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
3616         PR target/87572
3617         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3618         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3619         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3621 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
3623         PR tree-optimization/87022
3624         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
3625         bits in dist vector rather than the first one.
3627 2018-10-15  Richard Biener  <rguenther@suse.de>
3629         PR middle-end/87610
3630         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
3631         (visit_loadstore): When a used restrict tag escaped verify that
3632         the points-to solution of "other" pointers do not include
3633         escaped.
3634         (compute_dependence_clique): If a used restrict tag escaped
3635         communicated that down to visit_loadstore.
3637 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
3639         * config/s390/s390.c (s390_expand_vec_init): Force vector element
3640         into reg if it isn't a general operand.
3642 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3644         PR target/87599
3645         * config/i386/sse.md (*vec_dupv2di): Add register source to
3646         movddup.
3648 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3650         PR target/87572
3651         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3652         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
3653         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3654         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3656 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
3658         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
3659         (notice_args_size): Set it in the current trace if no insn that can
3660         throw internally has been seen yet.
3661         (connect_traces): When connecting args_size between traces, allow the
3662         incoming values not to match if there is an insn setting it before the
3663         first insn that can throw internally; in that case, force the creation
3664         of a CFI note on this latter insn.
3666 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
3668         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
3669         the base class, because using the injected-class-name was not clearly
3670         specified until DR 176.
3672 2018-10-12  Paul Koning  <ni1d@arrl.net>
3674         * config/pdp11/pdp11.md (doloop_end): New expander.
3675         (doloop_end_insn): renamed from "doloop_end".
3676         (addqi3): New pattern.
3677         (subqi3): New pattern.
3678         * config/pdp11/predicates.md (incdec_operand): New predicate.
3680 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
3682         PR middle-end/81376
3683         * real.c (format_helper::can_represent_integral_type_p): New function
3684         * real.h (format_helper::can_represent_integral_type_p): Ditto.
3685         * match.pd: New pattern.
3687 2018-10-12  Alexandre Oliva <oliva@adacore.com>
3689         * configure.ac: Introduce --enable-large-address-aware
3690         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
3691         * doc/install.texi: Document it.
3692         * configure, config.in: Rebuilt.
3693         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
3694         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
3695         (LINK_SPEC): Insert it.
3696         * config/i386/mingw-w64.h: Likewise.
3698         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
3700 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
3702         PR rtl-optimization/87600
3703         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
3705 2018-10-12  Paul Koning  <ni1d@arrl.net>
3707         * doc/md.texi (doloop_end): Document that the pattern code may
3708         need to check operand mode.
3710 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
3712         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
3713         to zero-extend between int and floating-point registers.
3714         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
3715         ldp into floating-point registers.  Add type and arch attributes.
3716         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
3717         Use f_loads for type attribute.
3719 2018-10-11  Martin Sebor  <msebor@redhat.com>
3721         * doc/extend.texi (attribute packed): Correct typos.
3723 2018-10-11  Martin Sebor  <msebor@redhat.com>
3725         * doc/extend.texi (attribute flatten): Mention interaction with
3726         noinline.
3728 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3730         PR target/87156
3731         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3732         Set new_decl virtual flag to zero.
3734 2018-10-11  Martin Sebor  <msebor@redhat.com>
3736         PR middle-end/87593
3737         * doc/extend.texi (attribute format_arg): Discuss using multiple
3738         attributes on a single function.
3740 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
3742         PR tree-optimization/86829
3743         * match.pd (sin (atan (x))): New simplification rules.
3744         (cos (atan (x))): Likewise.
3745         * real.c (build_sinatan_real): New function.
3746         * real.h (build_sinatan_real): Prototype.
3748 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
3750         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
3751         function.
3752         (fold_mergeeo_helper): New helper function.
3753         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
3754         intrinsics.  Correct some whitespace indentation issues.
3756 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
3758         PR target/87511
3759         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
3760         Use HOST_WIDE_INT_1U for shift.
3762 2018-10-11  Doug Rupp  <rupp@adacore.com>
3763             Olivier Hainque  <hainque@adacore.com>
3765         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
3766         Pass --relax to the linker for RTPs.
3767         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
3769 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
3770             Jan Hubicka  <jh@suse.cz>
3771             Martin Jambor  <mjambor@suse.cz>
3773         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
3774         the same elements are repeated rather than printing all of them.
3775         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
3776         "repeated" elements.
3777         * read-rtl-function.c (test_loading_repeat): New function.
3778         (read_rtl_function_c_tests): Call test_loading_repeat.
3779         * rtl-tests.c (test_dumping_repeat): New function.
3780         (rtl_tests_c_tests): Call test_dumping_repeat.
3782 2018-10-11  Richard Biener  <rguenther@suse.de>
3784         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
3785         bdver?_cost): Unify to ...
3786         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
3787         * config/i386/i386.c (processor_cost_table): Adjust.
3789 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
3791         PR middle-end/87574
3792         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
3793         the thunk when expanding to GIMPLE.
3795 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3797         * varasm.c (mergeable_string_section): Don't try to move zero-length
3798         strings to the merge section.
3800 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
3802         PR target/87573
3803         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
3805 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
3807         PR target/87550
3808         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
3809         to special_args set.
3811 2018-10-10  Richard Biener  <rguenther@suse.de>
3813         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
3814         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
3815         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
3816         and recursing and pattern terminating the recursion on SSE
3817         vector width using ix86_expand_reduc.
3818         (reduc_sminmax_scal_<mode>): Split into part reducing to half
3819         width and recursing and SSE2 vector variant doing the final
3820         reduction with ix86_expand_reduc.
3821         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
3822         with terminating the recursion at AVX level, splitting that
3823         to SSE there.
3825 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
3827         * genmatch.c (error_cb): Rename to...
3828         (diagnostic_cb): ...this, converting int params to enums.
3829         (fatal_at): Update for renaming.
3830         (warning_at): Likewise.
3831         (main): Likewise.
3832         * input.c (selftest::ebcdic_execution_charset::apply):
3833         Update for renaming of...
3834         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
3835         to...
3836         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
3837         converting level and reason to enums.
3838         (class selftest::lexer_error_sink): Rename to...
3839         (class selftest::lexer_test_options): ...this, renaming field
3840         "m_errors" to "m_diagnostics".
3841         (selftest::lexer_test_options::apply): Update for renaming of...
3842         (selftest::lexer_test_options::on_error): ...this, renaming to...
3843         (selftest::lexer_test_options::on_diagnostic): ...this
3844         converting level and reason to enums.
3845         (selftest::test_lexer_string_locations_raw_string_unterminated):
3846         Update for renamings.
3847         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
3848         "reason".
3850 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
3852         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
3853         * config/rs6000/pmmintrin.h: New file.
3855 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
3857         PR tree-optimization/86659
3858         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
3860 2018-10-09  Richard Biener  <rguenther@suse.de>
3862         PR tree-optimization/63155
3863         * tree-ssa-structalias.c: Include tree-ssa.h.
3864         (get_constraint_for_ssa_var): For undefs return nothing_id.
3865         (find_func_aliases): Cleanup PHI handling.
3867 2018-10-09  Richard Biener  <rguenther@suse.de>
3869         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
3870         replacements.
3872 2018-10-09  Martin Liska  <mliska@suse.cz>
3874         * asan.c (asan_emit_stack_protection): If a stack variable
3875         is located in a same file as current function, then emit
3876         line info into variable definition string.
3878 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3880         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
3881         information.
3883 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3885         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
3886         on the thunk.
3888 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3890         PR tree-optimization/86659
3891         * gimple-match.h (struct gimple_match_op): Add reverse field.
3892         (gimple_match_op::set_op): New overloaded method.
3893         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
3894         the REF_REVERSE_STORAGE_ORDER flag on the value.
3895         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
3896         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
3897         
3898 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
3900         PR middle-end/63155
3901         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
3902         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
3904 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
3906         PR target/87517
3907         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
3908         Defined with __builtin_ia32_vfmaddsubpd512_mask.
3910 2018-10-08  Richard Biener  <rguenther@suse.de>
3912         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
3913         cost the same as AVX128 ones.
3915 2018-10-08  Paul Koning  <ni1d@arrl.net>
3917         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
3918         (expand_block_move): New function.
3919         * config/pdp11/pdp11.c (output_block_move): Remove.
3920         (expand_block_move): New function.
3921         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
3922         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
3923         (*movmemhi1): Remove.
3925 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
3927         * config/s390/2827.md: Increase latencies for some FP instructions.
3929 2018-10-08  Richard Biener  <rguenther@suse.de>
3931         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3932         Open a dump scope.
3933         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
3934         * tree-vectorizer.h (dump_stmt_cost): Adjust.
3935         (add_stmt_cost): Dump return value of the hook.
3937 2018-10-08  Richard Biener  <rguenther@suse.de>
3939         PR tree-optimization/63155
3940         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
3941         (ssa_propagation_engine::ssa_propagate): Remove redundant
3942         bitmap bit clearing.
3944 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
3946         PR rtl-optimization/86939
3947         PR rtl-optimization/87479
3948         * ira.h (non_conflicting_reg_copy_p): New prototype.
3949         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
3950         (make_hard_regno_dead): Don't add conflicts for register
3951         ignore_reg_for_conflicts.
3952         (make_object_dead): Likewise.
3953         (non_conflicting_reg_copy_p): New function.
3954         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
3955         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
3956         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
3957         (make_hard_regno_dead): Don't add conflicts for register
3958         ignore_reg_for_conflicts.  Remove special conflict handling of
3959         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
3960         check_pic_pseudo_p and update callers.
3961         (mark_pseudo_dead): Don't add conflicts for register
3962         ignore_reg_for_conflicts.
3963         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
3965 2018-10-05  Andrew Waterman  <andrew@sifive.com>
3966             Jim Wilson  <jimw@sifive.com>
3968         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
3969         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
3970         new pattern using HONOR_SNANS that emits one extra instruction.
3972 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
3974         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
3975         patterns): Merge SI and DI patterns to a GPR pattern.
3976         (unnamed define_insn and define_split for record form of that): Merge
3977         to a single define_insn_and_split pattern.
3979 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
3981         PR c++/56856
3982         * input.c (expand_location_to_spelling_point): Add param "aspect"
3983         and use rather than hardcoding LOCATION_ASPECT_CARET.
3984         (get_substring_ranges_for_loc): Handle the case of a single token
3985         within a macro expansion.
3986         * input.h (expand_location_to_spelling_point): Add "aspect" param,
3987         defaulting to LOCATION_ASPECT_CARET.
3989 2018-10-05  Paul Koning  <ni1d@arrl.net>
3991         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
3992         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
3993         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3994         (pdp11_guard_type): New function.
3996 2018-10-05  Paul Koning  <ni1d@arrl.net>
3998         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
3999         * config/pdp11/pdp11.opt (mfloat32): Remove.
4000         (mfloat64): Remove.
4001         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4002         (pdp11 -mfloat64): Remove.
4004 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4006         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4007         (*cmp<mode>_cc_i387): Ditto.
4008         (*cmpu<mode>_cc_i387): Ditto.
4009         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4010         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4011         "scratch" argument.
4012         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4013         Emit x86_sahf_1 pattern.
4014         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4015         (ix86_expand_carry_flag_compare): Ditto.
4017 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4019         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4020         to reg_or_0_operand.  Add "C" constraint.
4021         (*cmpxf_cc_i387): Ditto.
4022         (*cmp<mode>_i387): Change operand 2 predicate
4023         to nonimm_or_0_operand.  Add "C" constraint.
4024         (*cmp<mode>_cc_i387): Ditto.
4025         (*cmp<mode>_0_i387): Remove insn pattern.
4026         (*cmp<mode>_0_cc_i387): Ditto.
4028 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4030         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4031         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4032         from vector_move_operand.  Update all uses.
4034 2018-10-05  Martin Sebor  <msebor@redhat.com>
4036         PR tree-optimization/87490
4037         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4038         consistently.
4040 2018-10-05  Richard Biener  <rguenther@suse.de>
4042         PR tree-optimization/63155
4043         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4044         vertical space in dumpfiles.
4045         * tree-ssa-propagate.h
4046         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4047         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4048         (ssa_edge_worklist_back): Likewise.
4049         (curr_order): Likewise.
4050         (cfg_blocks_get): Remove abstraction.
4051         (cfg_blocks_add): Likewise.
4052         (cfg_blocks_empty_p): Likewise.
4053         (add_ssa_edge): Add to current or next worklist based on
4054         RPO index.
4055         (add_control_edge): Likewise.
4056         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4057         into ...
4058         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4059         iteration from CFG and SSA edge worklist so we process
4060         everything in RPO order, prioritizing forward progress
4061         over iteration.
4062         (ssa_prop_init): Allocate new worklists, do not dump
4063         immediate uses.
4064         (ssa_prop_fini): Free new worklists.
4066 2018-10-05  Richard Biener  <rguenther@suse.de>
4068         * tree-core.h (tree_block::abstract_flag): Remove.
4069         (tree_block::block_num): Make full 32bits.
4070         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4071         * tree.h (BLOCK_ABSTRACT): Remove.
4072         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4073         resulting from BLOCK_ABSTRACT being always false.
4074         (gen_inlined_subroutine_die): Likewise.
4075         (gen_block_die): Likewise.
4076         * tree.c (block_ultimate_origin): Likewise.
4077         * tree-pretty-print.c (dump_block_node): Remove code dealing
4078         with BLOCK_ABSTRACT.
4079         * tree-ssa-live.c (dump_scope_block): Likewise.
4080         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4081         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4083 2018-10-05   Richard Biener  <rguenther@suse.de>
4085         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4086         is asked for initialize mode to the component mode of the
4087         vector type.
4089 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4091         PR target/87522
4092         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4093         assembler for -mavx.
4094         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4096 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4098         PR target/87509
4099         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4100         RS6000_BTM_DFP.
4101         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4102         to be DImode.  When using mffscrn, force the operand to a register.
4104 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4106         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4107         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4108         X87MODEF mode iterator.
4109         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4110         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4111         X87MODEF mode iterator.
4113 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4115         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4116         -Wno-prio-ctor-dtor.
4118 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4120         * Makefile.in (OBJS): Add opt-problem.o.
4121         * dump-context.h: Include "selftest.h.
4122         (selftest::temp_dump_context): New forward decl.
4123         (class dump_context): Make friend of class
4124         selftest::temp_dump_context.
4125         (dump_context::dump_loc_immediate): New decl.
4126         (class dump_pretty_printer): Move here from dumpfile.c.
4127         (class temp_dump_context): Move to namespace selftest.
4128         (temp_dump_context::temp_dump_context): Add param
4129         "forcibly_enable_dumping".
4130         (selftest::verify_dumped_text):
4131         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4132         (selftest::verify_item):
4133         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4134         (ASSERT_IS_TREE): Likewise.
4135         (ASSERT_IS_GIMPLE): Likewise.
4136         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4137         to...
4138         (dump_context::dump_loc_immediate): ...this new function.
4139         (class dump_pretty_printer): Move to dump-context.h.
4140         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4141         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4142         (temp_dump_context::temp_dump_context): Move to "selftest"
4143         namespace.  Add param "forcibly_enable_dumping", and use it to
4144         conditionalize the use of m_pp;
4145         (selftest::verify_dumped_text): Make non-static.
4146         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4147         (selftest::verify_item): Make non-static.
4148         (ASSERT_IS_TEXT): Move to dump-context.h.
4149         (ASSERT_IS_TREE): Likewise.
4150         (ASSERT_IS_GIMPLE): Likewise.
4151         (selftest::test_capture_of_dump_calls): Pass "true" for new
4152         param of temp_dump_context.
4153         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4154         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
4155         TDF_COMPARE_DEBUG.
4156         * opt-problem.cc: New file.
4157         * opt-problem.h: New file.
4158         * optinfo-emit-json.cc
4159         (selftest::test_building_json_from_dump_calls): Pass "true" for
4160         new param of temp_dump_context.
4161         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4162         (optinfo::emit_for_opt_problem): New function.
4163         (optinfo::emit): Clarity which emit_item is used.
4164         * optinfo.h (optinfo::get_dump_location): New accessor.
4165         (optinfo::emit_for_opt_problem): New decl.
4166         (optinfo::emit): Make const.
4167         * selftest-run-tests.c (selftest::run_tests): Call
4168         selftest::opt_problem_cc_tests.
4169         * selftest.h (selftest::opt_problem_cc_tests): New decl.
4170         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4171         bool to opt_result, converting fprintf messages to
4172         opt_result::failure_at calls.  Add "stmt" param for use by the
4173         failure_at calls.
4174         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4175         (runtime_alias_check_p): Convert return type from bool to
4176         opt_result, converting dump_printf calls to
4177         opt_result::failure_at, using the statement DDR_A for their
4178         location.
4179         (find_data_references_in_stmt): Convert return type from bool to
4180         opt_result, converting "return false" to opt_result::failure_at
4181         with a new message.
4182         * tree-data-ref.h: Include "opt-problem.h".
4183         (dr_analyze_innermost): Convert return type from bool to opt_result,
4184         and add a const gimple * param.
4185         (find_data_references_in_stmt): Convert return type from bool to
4186         opt_result.
4187         (runtime_alias_check_p): Likewise.
4188         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4189         dr_analyze_innermost.
4190         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4191         Convert return type from bool to opt_result, adding a message for
4192         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4193         (vect_analyze_data_ref_dependence): Convert return type from bool
4194         to opt_result.  Change sense of return type from "false"
4195         effectively meaning "no problems" to "false" meaning a problem,
4196         so that "return false" becomes "return opt_result::success".
4197         Convert "return true" calls to opt_result::failure_at, using
4198         the location of statement A rather than vect_location.
4199         (vect_analyze_data_ref_dependences): Convert return type from bool
4200         to opt_result.
4201         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4202         calls to opt_result::failure_at, using the stmt location rather
4203         than vect_location.
4204         (vect_verify_datarefs_alignment): Convert return type from bool
4205         to opt_result.
4206         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
4207         into multiple more-tightly-scoped copies.
4208         (vect_analyze_data_refs_alignment): Convert return type from bool
4209         to opt_result.
4210         (vect_analyze_data_ref_accesses): Likewise, converting a
4211         "return false" to a "return opt_result::failure_at", adding a
4212         new message.
4213         (vect_prune_runtime_alias_test_list): Convert return type from
4214         bool to opt_result, converting dump_printf_loc to
4215         opt_result::failure_at.  Add a %G to show the pertinent statement,
4216         and use the stmt's location rather than vect_location.
4217         (vect_find_stmt_data_reference): Convert return type from
4218         bool to opt_result, converting dump_printf_loc to
4219         opt_result::failure_at, using stmt's location.
4220         (vect_analyze_data_refs):  Convert return type from bool to
4221         opt_result.  Convert "return false" to "return
4222         opt_result::failure_at", adding messages as needed.
4223         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4224         type from bool to opt_result.
4225         (vect_determine_vf_for_stmt): Likewise.
4226         (vect_determine_vectorization_factor): Likewise, converting
4227         dump_printf_loc to opt_result::failure_at, using location of phi
4228         rather than vect_location.
4229         (vect_analyze_loop_form_1): Convert return type from bool to
4230         opt_result, converting dump_printf_loc calls, retaining the use of
4231         vect_location.
4232         (vect_analyze_loop_form): Convert return type from loop_vec_info
4233         to opt_loop_vec_info.
4234         (vect_analyze_loop_operations): Convert return type from bool to
4235         opt_result, converting dump_printf_loc calls, using the location
4236         of phi/stmt rather than vect_location where available.  Convert
4237         various "return false" to "return opt_result::failure_at" with
4238         "unsupported phi" messages.
4239         (vect_get_datarefs_in_loop): Convert return type from bool to
4240         opt_result.  Add a message for the
4241         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4242         (vect_analyze_loop_2): Convert return type from bool to
4243         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4244         each "goto again;", adding new messages where needed.
4245         Add "unsupported grouped {store|load}" messages.
4246         (vect_analyze_loop): Convert return type from loop_vec_info to
4247         opt_loop_vec_info.
4248         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4249         bool to opt_result.
4250         * tree-vect-stmts.c (process_use): Likewise, converting
4251         dump_printf_loc call and using stmt location, rather than
4252         vect_location.
4253         (vect_mark_stmts_to_be_vectorized): Likeise.
4254         (vect_analyze_stmt): Likewise, adding a %G.
4255         (vect_get_vector_types_for_stmt): Convert return type from bool to
4256         opt_result, converting dump_printf_loc calls and using stmt
4257         location, rather than vect_location.
4258         (vect_get_mask_type_for_stmt): Convert return type from tree to
4259         opt_tree, converting dump_printf_loc calls and using stmt location.
4260         * tree-vectorizer.c: Include "opt-problem.h.
4261         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4262         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4263         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4264         enabled, use it to report at the top level "couldn't vectorize
4265         loop" followed by the problem.
4266         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4267         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4268         to opt_result.
4269         (vect_analyze_stmt): Likewise.
4270         (vect_get_vector_types_for_stmt): Likewise.
4271         (tree vect_get_mask_type_for_stmt): Likewise.
4272         (vect_analyze_data_ref_dependences): Likewise.
4273         (vect_enhance_data_refs_alignment): Likewise.
4274         (vect_analyze_data_refs_alignment): Likewise.
4275         (vect_verify_datarefs_alignment): Likewise.
4276         (vect_analyze_data_ref_accesses): Likewise.
4277         (vect_prune_runtime_alias_test_list): Likewise.
4278         (vect_find_stmt_data_reference): Likewise.
4279         (vect_analyze_data_refs): Likewise.
4280         (vect_analyze_loop): Convert return type from loop_vec_info to
4281         opt_loop_vec_info.
4282         (vect_analyze_loop_form): Likewise.
4283         (vect_analyze_slp): Convert return type from bool to opt_result.
4285 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4287         * doc/invoke.texi (-fopt-info): Document new "internals"
4288         sub-option.
4289         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4290         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4291         MSG_ALL_KINDS.
4292         (optinfo_verbosity_options): Add "internals".
4293         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4294         (dump_context::apply_dump_filter_p): New member function.
4295         (dump_context::dump_loc): Use apply_dump_filter_p rather than
4296         explicitly masking the dump_kind.
4297         (dump_context::begin_scope): Increment the scope depth first.  Use
4298         apply_dump_filter_p rather than explicitly masking the dump_kind.
4299         (dump_context::emit_item): Use apply_dump_filter_p rather than
4300         explicitly masking the dump_kind.
4301         (dump_dec): Likewise.
4302         (dump_hex): Likewise.
4303         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4304         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4305         (opt_info_switch_p): Update handling of default
4306         MSG_OPTIMIZED_LOCATIONS to cope with default of
4307         MSG_PRIORITY_USER_FACING.
4308         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4309         masking the dump_kind.
4310         (selftest::test_capture_of_dump_calls): Update test_dump_context
4311         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4312         than MSG_ALL.  Generalize scope test to be run at all four
4313         combinations of with/without MSG_PRIORITY_USER_FACING and
4314         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4315         for each of the two values.
4316         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4317         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
4318         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4319         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4320         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4321         with MSG_PRIORITY_*.
4322         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4323         dump messages as MSG_PRIORITY_USER_FACING.
4324         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4325         about the interaction with MSG_PRIORITY_*.
4327 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4329         * varasm.c (output_constant): Add new parameter merge_strings.
4330         Make strings properly zero terminated in merge string sections.
4331         (mergeable_string_section): Don't fail if the last char is non-zero.
4332         (assemble_variable_contents): Handle merge string sections.
4333         (assemble_variable): Likewise.
4334         (assemble_constant_contents): Likewise.
4335         (output_constant_def_contents): Likewise.
4336         (output_constructor_array_range,
4337         output_constructor_regular_field): Adjust call to output_constant.
4338         (output_object_block): Adjust call to assemble_constant_contents
4339         and assemble_variable_contents.
4341 2018-10-04  Martin Liska  <mliska@suse.cz>
4343         PR c/87483
4344         * cgraphunit.c (process_function_and_variable_attributes):
4345         Warn about a function with alias attribute and a body.
4347 2018-10-04  Martin Liska  <mliska@suse.cz>
4349         PR ipa/82625
4350         * multiple_target.c (redirect_to_specific_clone): New function.
4351         (ipa_target_clone): Use it.
4352         * tree-inline.c: Fix comment.
4354 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4356         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4357         fields.
4358         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4359         (gcc::dump_manager::register_pass): New member function, adapted
4360         from loop body in gcc::pass_manager::register_pass, adding a
4361         call to update_dfi_for_opt_info.
4362         (gcc::dump_manager::opt_info_enable_passes): Store the
4363         -fopt-info options into the new fields.  Move the loop
4364         bodies into...
4365         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4366         function.
4367         * dumpfile.h (struct opt_pass): New forward decl.
4368         (gcc::dump_manager::register_pass): New decl.
4369         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4370         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4371         "m_optinfo_flags", and "m_optinfo_filename".
4372         * passes.c (gcc::pass_manager::register_pass): Move all of the
4373         dump-handling code to gcc::dump_manager::register_pass.
4375 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4377         PR rtl-optimization/87466
4378         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4379         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4380         * doc/tm.texi: Regenerate.
4381         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4382         * lra-lives.c (process_bb_lives): Likewise.
4383         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4384         Define.
4386 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
4388         * params.c (add_params): Fix initialization.
4390 2018-10-04  Martin Liska  <mliska@suse.cz>
4392         PR gcov-profile/84107
4393         * tree-profile.c (init_ic_make_global_vars):
4394         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4395         Come up with new ic_tuple* variables.  Emit
4396         __gcov_indirect_call{,_topn} variables.
4397         (gimple_gen_ic_profiler): Access the variable
4398         and emit gimple.
4399         (gimple_gen_ic_func_profiler): Access
4400         __gcov_indirect_call.callee field.
4401         (gimple_init_gcov_profiler): Use ptr_type_node.
4402         * value-prof.c (gimple_ic): Use ptr_type_node.
4404 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4406         PR tree-optimization/85787
4407         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4408         into this function and add support for detecting multiple phis.
4409         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4411 2018-10-04  Martin Liska  <mliska@suse.cz>
4413         PR ipa/87491
4414         * ipa-inline.c (inline_to_all_callers_1):
4415         Call ultimate_alias_target for node being inlined.
4417 2018-10-03  Jeff Law  <law@redhat.com>
4419         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4420         target's wchar_t.
4421         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4422         * tree.h (get_typenode_from_name): Prototype.
4424 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4426         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4427         Change operand 2 predicate to nonimmediate_operand.
4428         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4430 2018-10-03  Martin Sebor  <msebor@redhat.com>
4431             Jeff Law  <law@redhat.com>
4433         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4434         initialize it.
4435         (get_string_length): Detect unterminated arrays.
4436         (format_string): Same.
4437         (format_directive): Warn about unterminated arrays.
4438         (handle_gimple_call): Mark statements with no_warning as needed.
4440 2018-10-03  Jim Wilson  <jimw@sifive.com>
4442         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4443         also define __riscv_abi_rve.  Delete trailing white space.
4445 2018-10-03  Paul Koning  <ni1d@arrl.net>
4447         Enable LRA register allocator for PDP11.
4448         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4449         (R): Likewise.
4450         (D): Likewise.
4451         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4452         * config/pdp11/pdp11.opt (-mlra): New option.
4453         * doc/invoke.texi (PDP-11 Options): Document -mlra.
4455 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4457         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4458         (*<absneg:code>extend<mode>xf2): Ditto.
4460 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4462         PR tree-optimization/87415
4463         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4464         precision fields.
4466 2018-10-02  Jeff Law  <law@redhat.com>
4468         * gimple-fold.c (get_range_strlen): Only set *nonstr when
4469         an unterminated string is discovered.  Bubble up range
4470         even for unterminated strings.
4471         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4472         indicates the string was not terminated via NONSTR.
4474 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4476         * tree-vrp.c (extract_range_from_unary_expr): Special case all
4477         pointer conversions.
4478         Do not do anything special for anti-ranges.
4480 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
4482         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4483         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4485 2018-10-03  Martin Liska  <mliska@suse.cz>
4487         PR gcov-profile/86109
4488         * coverage.c (coverage_begin_function): Do not
4489         mark lambdas as artificial.
4490         * tree-core.h (struct GTY): Remove tm_clone_flag
4491         and introduce new lambda_function.
4492         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4494 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4496         PR target/87474
4497         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4498         P8_VECTOR and VSX are enabled.
4500 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4502         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4503         0x3907 as CPU model number.
4505 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4507         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4508         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4509         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4510         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4511         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4512         * config/s390/s390.md: Likewise. Rename also the cpu attribute
4513         value from arch12 to z14.
4515 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4517         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4518         (isinfxf2): Ditto.
4519         (isinf<mode>2): Ditto.
4521 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4523         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4524         before emitting fxam.  Perform calculations in XFmode.
4526 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4528         * match.pd (((X /[ex] A) +- B) * A): New transformation.
4530 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4532         PR middle-end/87319
4533         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
4534         * tree.c (signed_or_unsigned_type_for): Handle complex.
4536 2018-10-02  Jeff Law  <law@redhat.com>
4538         * gimple-fold.c (get_range_strlen): Remove dead code.
4540 2018-10-02  Martin Sebor  <msebor@redhat.com>
4541             Jeff Law  <law@redhat.com>
4543         * builtins.c (unterminated_array): Add new arguments.
4544         If argument is not terminated, bubble up size and exact
4545         state to callers.
4546         (expand_builtin_strnlen): Detect, avoid expanding
4547         and diagnose unterminated arrays.
4548         (c_strlen): Fill in offset of start of unterminated strings.
4549         * builtins.h (unterminated_array): Update prototype.
4551 2018-10-02  Richard Biener  <rguenther@suse.de>
4553         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
4554         of haddv4df, first reduce to SSE width and exploit the fact
4555         that we only need element zero with the reduction result.
4556         (reduc_plus_scal_v2df): Likewise.
4558 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4560         * dojump.h (do_jump): Delete.
4561         (do_jump_1): Likewise.
4562         (split_comparison): Move around.
4563         * dojump.c (do_jump): Make static.
4564         (do_jump_1): Likewise.
4565         (jumpifnot): Move around.
4566         (jumpifnot_1): Likewise.
4567         (jumpif): Likewise.
4568         (jumpif_1): Likewise.
4569         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
4571 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4573         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
4574         insns in the delay slot and add_insn_after for the jump insn.
4576 2018-10-02  Richard Biener  <rguenther@suse.de>
4578         * tree-inline.c (expand_call_inline): Use the location of
4579         the callee declaration for the inline-entry marker.
4580         * final.c (notice_source_line): Remove special-casing of
4581         NOTE_INSN_INLINE_ENTRY.
4583 2018-10-01  Carl Love  <cel@us.ibm.com>
4585         PR 69431
4586         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
4587         (__builtin_mtfsb0): New.
4588         (__builtin_mtfsb1): New.
4589         ( __builtin_set_fpscr_rn): New.
4590         (__builtin_set_fpscr_drn): New.
4591         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
4592         (rs6000_expand_set_fpscr_rn_builtin): Add.
4593         (rs6000_expand_set_fpscr_drn_builtin): Add.
4594         (rs6000_expand_builtin): Add case statement entries for
4595         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
4596         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
4597         RS6000_BUILTIN_MFFSL.
4598         (rs6000_init_builtins): Add ftype initialization and def_builtin
4599         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
4600         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
4601         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
4602         rs6000_mffscdrn): Add define_insn.
4603         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
4604         * doc/extend.texi: Add documentation for the builtins.
4606 2018-10-01  Richard Biener  <rguenther@suse.de>
4608         PR tree-optimization/87465
4609         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
4610         causing branch miscounts.
4612 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4614         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
4615         aarch64_option_default_param):  New.
4616         (params.h): Include.
4617         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
4618         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
4619         stack-clash protection validation code.
4621 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4623         * params.c (validate_param): New.
4624         (add_params): Use it.
4625         (set_param_value): Refactor param validation into validate_param.
4626         (diagnostic.h): Include.
4627         * diagnostic.h (diagnostic_ready_p): New.
4629 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4631         * params.c (set_param_value):
4632         Add index of parameter being validated.
4633         * common/common-target.def (option_validate_param): New.
4634         * common/common-targhooks.h (default_option_validate_param): New.
4635         * common/common-targhooks.c (default_option_validate_param): New.
4636         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
4637         * doc/tm.texi: Regenerate.
4639 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4641         PR target/86486
4642         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4643         Add validation for stack-clash parameters and set defaults.
4645 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4647         PR target/86486
4648         * configure.ac: Add stack-clash-protection-guard-size.
4649         * doc/install.texi: Document it.
4650         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
4651         * params.def: Update comment for guard-size.
4652         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
4653         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
4654         * configure: Regenerate.
4656 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4658         PR target/86486
4659         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
4660         STACK_DYNAMIC_OFFSET): New.
4661         * config/aarch64/aarch64.c (aarch64_layout_frame):
4662         Update outgoing args size.
4663         (aarch64_stack_clash_protection_alloca_probe_range,
4664         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4666 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4668         PR target/86486
4669         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
4670         probe ranges.
4671         * target.def (stack_clash_protection_alloca_probe_range): New.
4672         (stack_clash_protection_final_dynamic_probe): Remove.
4673         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
4674         (default_stack_clash_protection_final_dynamic_probe): Remove.
4675         * targhooks.c: Likewise.
4676         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4677         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
4678         * doc/tm.texi: Regenerate.
4680 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4682         PR target/86486
4683         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
4684         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
4685         aarch64_clamp_to_uimm12_shift): New.
4686         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
4687         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
4689 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4691         PR target/86486
4692         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
4694 2018-10-01  Jeff Law  <law@redhat.com>
4695             Richard Sandiford <richard.sandiford@linaro.org>
4696             Tamar Christina  <tamar.christina@arm.com>
4698         PR target/86486
4699         * config/aarch64/aarch64.md
4700         (probe_stack_range): Add k (SP) constraint.
4701         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
4702         STACK_CLASH_MAX_UNROLL_PAGES): New.
4703         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
4704         stack probes for stack clash.
4705         (aarch64_allocate_and_probe_stack_space): New.
4706         (aarch64_expand_prologue): Use it.
4707         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
4708         (aarch64_sub_sp): Add emit_move_imm optional param.
4710 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
4712         PR tree-optimization/87261
4713         * match.pd: Remove trailing whitespace.
4714         Add (x & y) | ~(x | y) -> ~(x ^ y),
4715         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
4717 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4719         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
4720         constraints.
4721         (ashlsi3_insn): Update instruction constraints.
4722         (ashrsi3_insn): Likewise.
4723         (rotrsi3): Likewise.
4724         (add_shift): Likewise.
4725         * config/arc/constraints.md (Csz): New 32 bit constraint. It
4726         avoids placing in the limm field small constants which, otherwise,
4727         could end into a small instruction.
4729 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4731         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
4732         destination register is not odd-even.
4733         (umaddsidi4_split): Likewise.
4735 2018-10-01  Richard Biener  <rguenther@suse.de>
4737         * tree-inline.c (expand_call_inline): Store origin of fn
4738         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
4739         * tree.c (block_ultimate_origin): Simplify and do some
4740         checking.
4742 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
4744         * config/i386/mmx.md (EMMS): New int iterator.
4745         (emms): New int attribute.
4746         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
4747         EMMS int iterator.  Explicitly declare clobbers.
4748         (mmx_emms): Remove expander.
4749         (mmx_femms): Ditto.
4750         * config/i386/predicates.md (emms_operation): Remove predicate.
4751         (vzeroall_pattern): New predicate.
4752         (vzeroupper_pattern): Rename from vzeroupper_operation.
4753         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
4754         vzeroupper_pattern and vzeroall_pattern predicates.
4756 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
4758         PR rtl-optimization/86939
4759         * ira-lives.c (make_hard_regno_born): Rename from this...
4760         (make_hard_regno_live): ... to this.  Remove update to conflict
4761         information.  Update function comment.
4762         (make_hard_regno_dead): Add conflict information update.  Update
4763         function comment.
4764         (make_object_born): Rename from this...
4765         (make_object_live): ... to this.  Remove update to conflict information.
4766         Update function comment.
4767         (make_object_dead):  Add conflict information update.  Update function
4768         comment.
4769         (mark_pseudo_regno_live): Call make_object_live.
4770         (mark_pseudo_regno_subword_live): Likewise.
4771         (mark_hard_reg_dead): Update function comment.
4772         (mark_hard_reg_live): Call make_hard_regno_live.
4773         (process_bb_node_lives): Likewise.
4774         * lra-lives.c (make_hard_regno_born): Rename from this...
4775         (make_hard_regno_live): ... to this.  Remove update to conflict
4776         information.  Remove now uneeded check_pic_pseudo_p argument.
4777         Update function comment.
4778         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
4779         to conflict information.  Update function comment.
4780         (mark_pseudo_live): Remove update to conflict information.  Update
4781         function comment.
4782         (mark_pseudo_dead): Add conflict information update.
4783         (mark_regno_live): Call make_hard_regno_live.
4784         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
4785         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
4787 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
4789         PR target/87370
4790         * config/i386/i386.c (construct_container): Use TImode for
4791         BLKmode values in 2 integer registers.
4793 2018-09-29  Jeff Law  <law@redhat.com>
4795         * builtins.c (unterminated_array): Pass in c_strlen_data * to
4796         c_strlen rather than just a tree *.
4797         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
4798         Update recursive calls appropriately.  If caller did not provide a
4799         suitable data pointer, create a local one.  When a non-terminated
4800         string is discovered, bubble up information about the string via the
4801         c_strlen_data object.
4802         * builtins.h (c_strlen): Update prototype.
4803         (c_strlen_data): New structure.
4804         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
4805         For a type 2 call, if c_strlen indicates a non-terminated string
4806         use the length of the non-terminated string.
4807         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
4809 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
4811         PR target/87467
4812         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
4813         __m512d type for __A argument rather than __m512.
4815 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
4817         * match.pd (simple_comparison): Don't optimize if either operand is
4818         a function pointer when target needs function pointer canonicalization.
4820 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4822         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
4823         power5 .. power9 to remove indirection.
4824         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
4825         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
4826         ASM_CPU_476_SPEC): Delete.
4827         (ASM_CPU_SPEC): Adjust.
4828         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
4829         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
4831 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4833         * config.in: Delete HAVE_AS_DCI.
4834         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
4835         * config/rs6000/rs6000.h: Ditto.
4836         * configure.ac: Delete HAVE_AS_DCI.
4837         * configure: Regenerate.
4839 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4841         * config.in (HAVE_AS_LWSYNC): Delete.
4842         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4843         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
4844         do it as a .long .
4845         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4846         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
4847         as a .long .
4848         * configure.ac: Delete HAVE_AS_LWSYNC.
4849         * configure: Regenerate.
4851 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4852             Pierre-Marie de Rodat  <derodat@adacore.com>
4854         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
4855         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
4856         (cgraph_node::create_thunk): Add indirect_offset parameter.
4857         (thunk_adjust): Likewise.
4858         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
4859         and initialize the corresponding field with it.
4860         (cgraph_node::dump): Dump indirect_offset field.
4861         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
4862         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
4863         (thunk_adjust): Add indirect_offset parameter and deal with it.
4864         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
4865         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
4866         or if the thunk is external or local.  Fix formatting.  Do not chain
4867         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
4868         if any, in the GIMPLE representation.
4869         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
4870         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
4871         (input_node): Read indirect_offset field.
4872         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
4873         call to thunk_adjust.
4874         * tree-nested.c (struct nesting_info): Add thunk_p field.
4875         (create_nesting_tree): Set it.
4876         (convert_all_function_calls): Copy static chain from targets to thunks.
4877         (finalize_nesting_tree_1): Return early for thunks.
4878         (unnest_nesting_tree_1): Do not finalize thunks.
4879         (gimplify_all_functions): Do not gimplify thunks.
4881 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4883         * opt-suggestions.c (option_proposer::build_option_suggestions):
4884         Release "option_values".
4886 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4888         * coverage.c (get_coverage_counts): Convert problem-reporting dump
4889         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
4890         * dumpfile.c (kind_as_string): New function.
4891         (dump_loc): Rather than a hardcoded prefix of "note: ", use
4892         kind_as_string to vary the prefix based on dump_kind.
4893         (selftest::test_capture_of_dump_calls): Update for above.
4895 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4897         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
4898         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
4900 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4902         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
4903         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
4904         INVALID_REGNUM instead of FPSR_REG.
4905         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
4906         * config/i386/i386.md: Update comment of FP compares.
4907         (fldenv): Do not clobber FPSR_REG.
4909 2018-09-28  Richard Biener  <rguenther@suse.de>
4911         * tree.h (BLOCK_ORIGIN): New.
4912         * omp-expand.c (grid_expand_target_grid_body): Assign
4913         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
4914         * tree-inline.c (remap_block): Likewise.
4915         * auto-profile.c (get_function_decl_from_block): Simplify
4916         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
4917         * langhooks.c (lhd_print_error_function): Likewise.
4918         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
4919         Likewise.
4920         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4921         * tree.c (block_nonartificial_location): Likewise.
4922         (block_ultimate_origin): Likewise.
4923         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
4924         no longer needed LTO case.
4926 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
4927             Jan Hubicka  <jh@suse.cz>
4928             Martin Jambor  <mjambor@suse.cz>
4930         * simplify-rtx.c (simplify_merge_mask): New function.
4931         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
4932         same masks are used in op1 or op2.
4933         (test_vec_merge): New function.
4934         (test_vector_ops): Call test_vec_merge.
4936 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4938         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
4939         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
4940         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
4941         (cypress_costs): Set it.
4942         (supersparc_costs): Likewise.
4943         (hypersparc_costs): Likewise.
4944         (leon_cost): Likewise.
4945         (leon3_costs): Likewise.
4946         (sparclet_costs): Likewise.
4947         (ultrasparc_costs): Likewise.
4948         (ultrasparc_costs): Likewise.
4949         (niagara_costs): Likewise.
4950         (niagara2_costs): Likewise.
4951         (niagara3_costs): Likewise.
4952         (niagara4_costs): Likewise.
4953         (niagara7_costs): Likewise.
4954         (m8_costs): Likewise.
4955         (TARGET_CAN_FOLLOW_JUMP): Define.
4956         (pass_work_around_errata::gate): Minor tweak.
4957         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
4958         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
4959         Automaitcally clear MASK_FSMULD mask for V7 processors.
4960         (sparc_can_follow_jump): New static function.
4961         (output_ubranch): Deal with CROSSING_JUMP_P.
4962         (sparc_use_sched_lookahead): Rewrite using switch statement.
4963         (sparc_issue_rate): Reorder.
4964         (sparc_branch_cost): New function.
4966 2018-09-27  Martin Sebor  <msebor@redhat.com>
4968         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
4969         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
4970         (int_fits_type_p): Same.
4972 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
4974         * config/i386/i386.md (FPCR_REG): Remove.
4975         (UNSPEC_FLDCW): Remove.
4976         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
4977         (x86_fldcw_1): Remove insn pattern.
4978         (fnstenv): Do not clobber FPCR_REG.
4979         (fldenv): Ditto.
4980         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
4981         (CALL_USED_REGISTERS): Ditto.
4982         (REG_ALLOC_ORDER): Ditto.
4983         (REG_CLASS_CONTENTS): Ditto.
4984         (HI_REGISTER_NAMES): Ditto.
4985         (ADDITIONAL_REGISTER_NAMES): Use defines instead
4986         of numerical constants.
4987         * config/i386/i386.c (regclass_map): Remove fpsr register.
4988         (dbx_register_map): Ditto.
4989         (dbx64_register_map): Ditto.
4990         (svr4_dbx_register_map): Ditto.
4991         (print_reg): Do not handle FPCR_REG.
4993 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
4995         PR target/87149
4996         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
4997         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
4998         Delete, always treat as true.
4999         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5000         Ditto.  Simplify remaining code.
5001         * config/powerpcspe/powerpcspe.h: Ditto.
5002         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5003         Simplify remaining code.
5004         (rs6000_expand_builtin): Ditto.
5005         * config/rs6000/rs6000.h: Ditto.
5006         * configure.ac: Ditto.
5007         * configure: Regenerate.
5009 2018-09-27  Martin Liska  <mliska@suse.cz>
5011         * coverage.c (get_coverage_counts): Revert the formatting
5012         of missing profile opt info.
5014 2018-09-27  Richard Biener  <rguenther@suse.de>
5016         PR debug/37801
5017         PR debug/87440
5018         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5019         block as we do not output that.
5020         (gen_inlined_subroutine_die): Elide the originally outermost
5021         block, matching what we do for concrete instances.
5022         (decls_for_scope): Add parameter specifying whether to recurse
5023         to subblocks.
5025 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5026             Tom de Vries  <tom@codesourcery.com>
5028         PR 82089
5030         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5031         STORE_FLAG_VALUE == 1.
5033 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5035         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5036         constant definitions.
5037         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5038         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5039         ("speculation_barrier"): New expander definition.
5041 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5043         PR gcov-profile/86957
5044         * common.opt: New warning option -Wmissing-profile.
5045         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5046         * doc/invoke.texi: Document -Wmissing-profile.
5048 2018-09-26  Jim Wilson  <jimw@sifive.com>
5050         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5051         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5053 2018-09-26  Martin Sebor  <msebor@redhat.com>
5055         * tree.c (zerop): Change return type to bool.
5056         (integer_zerop, integer_onep, integer_each_onep): Same.
5057         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5058         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5059         (real_onep, real_minus_onep, chain_index): Same.
5060         (print_type_hash_statistics, type_list_equal): Same.
5061         * tree.h (zerop): Same.
5062         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5063         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5064         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5065         (real_onep, real_minus_onep, chain_index): Same.
5066         (print_type_hash_statistics, type_list_equal): Same.
5068 2018-09-26  Jim Wilson  <jimw@sifive.com>
5070         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5072 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5074         PR target/87414
5075         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5076         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5077         call.
5079 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5081         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5083 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5085         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5086         and FP_SECOND_SSE_REGS.
5087         (REG_CLASS_NAMES): Ditto.
5088         (REG_CLASS_CONTENTS): Ditto.
5089         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5090         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5091         (ix86_preferred_output_reload_class): Ditto.
5092         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5093         clobber constraint to "=&f".
5094         (fix_truncdi_i387): Ditto.
5095         (lrintxfdi2): Ditto.
5096         (fistdi2_<rounding>): Ditto.
5097         (fpremxf4_i387): Change "=u" constraint to "=f".
5098         (fprem1xf4_i387): Ditto.
5099         (sincosxf3): Ditto.
5100         (fptanxf4_i387): Ditto.
5101         (fxtractxf3_i387): Ditto.
5102         (fscalexf4_i387): Ditto.
5103         (atan2xf3): Change "u" constraint to "f".
5104         (fyl2xxf3_i387): Ditto.
5105         (fyl2xp1xf3_i387): Ditto.
5107 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5109         PR target/87439
5110         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5111         for removed I387_MASK_PM entity.
5114 2018-09-26  Jeff Law  <law@redhat.com>
5115         Revert
5116         2018-09-26  Alexey Neyman  <stilor@att.net>
5118         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5119         headers are no longer pulled in by <isl/val.h>.
5121 2018-09-26  Richard Biener  <rguenther@suse.de>
5123         PR debug/87443
5124         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5125         or concrete instance DIE to the tree.  Create abstract origin
5126         attributes also for concrete instances.
5128 2018-09-26  Alexey Neyman  <stilor@att.net>
5130         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5131         headers are no longer pulled in by <isl/val.h>.
5133 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5135         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5136         Use new helper functions.
5137         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5138         Use new helper functions.
5139         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5140         aarch_mm_needs_release): New declarations.
5141         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5142         aarch_mm_needs_release): New.
5144 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
5146         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5147         (arm32_output_mi_thunk): Deal with long calls.
5149 2018-09-26  Richard Biener  <rguenther@suse.de>
5151         PR debug/87428
5152         PR debug/87362
5153         * tree-inline.c (expand_call_inline): When the location
5154         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5155         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5156         the inserted BLOCK to make inlined_function_outer_scope_p
5157         recognize it.
5158         * dwarf2out.c (add_call_src_coords_attributes): Do not add
5159         coords for reserved locations.
5161 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5163         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5164         (*call_indirect_nonlocal_sysv<mode>): Ditto.
5165         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5166         (*sibcall_nonlocal_sysv<mode>): Ditto.
5167         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5168         (<bd>_<mode>): Ditto.
5169         (<bd>tf_<mode>): Ditto.
5171 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5173         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5174         control string as a list of templates instead of as C code.
5175         (*altivec_movti): Ditto.
5176         * config/rs6000/darwin.md (movdf_low_di): Ditto.
5178 2018-09-25  Jim Wilson  <jimw@sifive.com>
5180         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5181         when target symbol is weak.
5183 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5185         PR c/87387
5186         * builtins.c (unterminated_array): Simplify.
5187         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
5188         where pointer arithmetic is safe.
5190 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5192         PR target/86987
5193         * config/rs6000/altivec.md (altivec_vspltb): Use
5194         const_0_to_15_operand instead of u5bit_cint_operand.
5195         (*altivec_vspltb_internal): Ditto.
5196         (altivec_vspltb_direct): Ditto.
5197         (altivec_vsplth): Use const_0_to_7_operand instead of
5198         u5bit_cint_operand.
5199         (*altivec_vsplth_internal): Ditto.
5200         (altivec_vsplth_direct): Ditto.
5201         (altivec_vspltw): Use const_0_to_3_operand instead of
5202         u5bit_cint_operand.
5203         (*altivec_vspltw_internal): Ditto.
5204         (altivec_vspltw_direct): Ditto.
5205         (altivec_vspltsf): Ditto.
5206         (*altivec_vspltsf_internal): Ditto.
5207         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5208         various splats with the proper size immediate.  Reorder the various
5209         cases by ascending size of immediate, and put all such together.
5211 2018-09-25  Richard Biener  <rguenther@suse.de>
5213         PR debug/83941
5214         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5215         GC-ification.
5216         (maybe_create_die_with_external_ref): Do not create
5217         DW_TAG_imported_unit here.
5218         (add_abstract_origin_attribute): Handle external BLOCK refs.
5219         (dwarf2out_abstract_function): Simplify LTO case.
5220         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5221         rather than using maybe_create_die_with_external_ref.
5223 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5225         PR target/71278
5226         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5227         (frndintxf2_mask_pm_i387): Ditto.
5228         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5229         Enable for !flag_trapping_math.
5230         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5231         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5232         Change operand 1 predicate to nonimmediate_operand.
5233         (attr "i387_cw"): Remove mask_pm.
5234         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5235         (enum ix86_entity): Remove I387_MASK_PM.
5236         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5237         handle I387_MASK_PM.
5238         (ix86_mode_needed): Ditto.
5239         (ix86_mode_after): Ditto.
5240         (ix86_mode_entry): Ditto.
5241         (ix86_mode_exit): Ditto.
5242         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5244 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5246         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5247         to_update_switch_stmts to vNULL instead of calling create on them
5248         immediately.
5250 2018-09-25  Richard Biener  <rguenther@suse.de>
5252         PR tree-optimization/87402
5253         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5254         (visit_phi): Re-instantiate handling of supposed to be VARYING
5255         but non-VARYING backedge value.
5257 2018-09-25  Richard Biener  <rguenther@suse.de>
5259         PR debug/83941
5260         * dwarf2out.c (struct sym_off_pair): New.
5261         (external_die_map): New global.
5262         (lookup_decl_die): When in LTO create DIEs lazily from the
5263         external_die_map.
5264         (lookup_block_die): New function, create DIEs lazily in LTO.
5265         (equate_block_to_die): New function.
5266         (dwarf2out_die_ref_for_decl): During WPA get the association
5267         from the external DIE map.
5268         (dwarf2out_register_external_die): Record mapping into the
5269         external DIE map.
5270         (maybe_create_die_with_external_ref): New function split out from
5271         DIE generation part of old dwarf2out_register_external_die.
5272         (add_abstract_origin_attribute): Do not return the DIE.  When
5273         in LTO reference externals directly.
5274         (dwarf2out_abstract_function): When in LTO ignore calls for
5275         decls with external DIEs (already present abstract instances).
5276         (gen_call_site_die): Adjust.
5277         (add_high_low_attributes): Likewise.
5278         (gen_lexical_block_die): Likewise.
5279         (gen_inlined_subroutine_die): Likewie.
5280         (gen_block_die): Likewise.
5281         (dwarf2out_inline_entry): Likewise.
5282         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5283         DIEs.
5285 2018-09-25  Martin Liska  <mliska@suse.cz>
5287         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5288         integers and not by a float value.
5290 2018-09-25  Martin Liska  <mliska@suse.cz>
5292         PR fortran/87394
5293         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5294         instead of NULL.
5295         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5296         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5297         * godump.c (go_format_type): Remove extra parenthesis.
5299 2018-09-25  Martin Liska  <mliska@suse.cz>
5301         * alias.c (set_dest_equal_p): Remove unused function.
5302         * config/i386/i386.c (def_builtin_pure2): Likewise.
5303         * diagnostic-show-locus.c (class layout): Remove
5304         unused field.
5305         (layout::layout): Likewise here.
5306         * dump-context.h (class temp_dump_context): Likewise.
5307         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5308         (add_AT_loclistsptr): Likewise.
5309         (add_AT_offset): Likewise.
5310         (get_AT_hi_pc): Likewise.
5311         (is_comdat_die): Likewise.
5312         (type_is_enum): Likewise.
5313         (ceiling): Likewise.
5314         (add_AT_vms_delta): Likewise.
5315         (is_class_die): Likewise.
5316         * edit-context.c (class line_event): Remove unused field.
5317         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5318         unused function.
5319         * ipa-cp.c (ipa_get_vr_lat): Likewise.
5320         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5321         (ok_for_base_p_nonstrict): Likewise.
5322         * tree-chrec.c (is_not_constant_evolution): Likewise.
5323         (chrec_fold_poly_cst): Likewise.
5324         * tree-if-conv.c (has_pred_critical_p): Likewise.
5325         * tree-ssa-coalesce.c (print_exprs): Likewise.
5326         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5327         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5328         * tree-vrp.c (value_ranges_intersect_p): Likewise.
5329         (value_range_nonnegative_p): Likewise.
5331 2018-09-25  Martin Liska  <mliska@suse.cz>
5333         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5334         Do not handle "GNU Pascal".
5335         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5336         Likewise.
5337         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5338         from documentation. Likewise.
5339         * dbxout.c (dbxout_range_type): Likewise.
5340         * doc/cpp.texi: Likewise.
5341         * doc/extend.texi: Likewise.
5342         * doc/frontends.texi: Likewise.
5343         * doc/invoke.texi: Remove Pascal entry.
5344         * tree.def (CLEANUP_POINT_EXPR): Likewise.
5345         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5347 2018-09-25  Martin Liska  <mliska@suse.cz>
5349         PR middle-end/86078
5350         * doc/invoke.texi: Document all parameters and remove default
5351         of the parameters.
5353 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5355         PR bootstrap/87417
5356         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5357         contains HOST_WIDE_INTs when computing its size.
5359 2018-09-24  Jim Wilson  <jimw@sifive.com>
5361         PR target/87391
5362         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5363         not TARGET_RVE.
5364         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5366 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5368         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5369         access prev before checking it for NULLness in the
5370         AARCH64_FUSE_CMP_BRANCH case.
5372 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5374         PR target/82699
5375         * config/i386/i386.c (rest_of_insert_endbranch): Set
5376         endbr_queued_at_entrance to true and don't insert ENDBR if
5377         x86_function_profiler will be called.
5378         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5379         is true.
5380         * config/i386/i386.h (machine_function): Add
5381         endbr_queued_at_entrance.
5383 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5385         * genattrtab.c (mk_attr_alt): Use alternative_mask.
5386         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5387         types.
5388         (check_attr_test): Use alternative_mask.
5389         (get_attr_value): Likewise.
5390         (compute_alternative_mask): Use alternative_mask and XWINT.
5391         (make_alternative_compare): Use alternative_mask.
5392         (attr_alt_subset_p): Use XWINT.
5393         (attr_alt_subset_of_compl_p): Likewise.
5394         (attr_alt_intersection): Use alternative_mask and XWINT.
5395         (attr_alt_union): Likewise.
5396         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5397         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5398         (simplify_test_exp): Use alternative_mask and XWINT.
5399         (write_test_expr): Use alternative_mask and XWINT, adjust bit
5400         number calculation to support 64 bits.  Generate code that
5401         checks 64-bit masks.
5402         (main): Use alternative_mask.
5403         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5405 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5407         PR target/80080
5408         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5409         RETURN+USE when returning via %r14.
5411 2018-09-24  Martin Liska  <mliska@suse.cz>
5413         * gcov.c (output_lines): Print colorization legend
5414         for both flag_use_colors and flag_use_hotness_colors.
5415         Reword the help.
5417 2018-09-24  Martin Liska  <mliska@suse.cz>
5419         * coverage.c (get_coverage_counts): Use warning_at
5420         with current_function_decl location. Use %qD in warning
5421         message.
5423 2018-09-24  Martin Liska  <mliska@suse.cz>
5425         * memory-block.h (memory_block_pool::release): Annotate with
5426         valgrind that the memory is not accessible.
5428 2018-09-24  Martin Liska  <mliska@suse.cz>
5430         PR sanitizer/85774
5431         * asan.c: Make asan_handled_variables extern.
5432         * asan.h: Likewise.
5433         * cfgexpand.c (expand_stack_vars): Make sure
5434         a representative is unpoison if another
5435         variable in the partition is handled by
5436         use-after-scope sanitization.
5438 2018-09-24  Richard Biener  <rguenther@suse.de>
5440         PR tree-optimization/63155
5441         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5442         the worklist when the edge of the respective argument isn't
5443         executable.
5445 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5447         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5448         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5449         (MASK_CLASS_P): Update for rename.
5450         (MAYBE_MASK_CLASS_P): Ditto.
5451         (REG_CLASS_NAMES): Update.
5452         (REG_CLASS_CONTENT): Update.
5453         * config/i386/i386.c (regclass_map): Update for MASK_REG
5454         and ALL_MASK_REGS rename.
5455         * config/i386/constraints.md (Yk): Update for rename.
5456         (k): Ditto.
5458 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5460         * config/i386/i386.h (enum reg_class): Remove
5461         EVEX_SSE_REGS and MOD4_SSE_REGS.
5462         (REG_CLASS_NAMES): Update.
5463         (REG_CLASS_CONTENT): Update.
5464         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5465         registers as ALL_SSE_REGS.
5466         (ix86_additional_allocno_class_p): Remove.
5467         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5468         (ix86_register_priority): Lower priority of EVEX SSE registers.
5469         Use IN_RANGE macro where appropriate.
5470         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5471         AVX-5124VNNIW checks.
5472         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5473         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5474         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5475         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5476         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5477         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5478         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5479         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5480         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5481         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5482         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5483         * config/i386/constraints.md (Yh): Remove.
5485 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5487         * config/i386/i386.c (regclass_map): Declare integer REX registers
5488         as GENERAL_REGS.
5490 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
5492         * doc/service.texi (Service): Switch the fsf.org link to https.
5494 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5496         PR target/86798
5497         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5498         Define to speculation_safe_value_not_needed.
5500 2018-09-21  Florian Weimer  <fweimer@redhat.com>
5502         PR middle-end/81035
5503         * doc/extend.texi (Common Function Attributes): Mention that
5504         noreturn suppresses tail call optimization.
5506 2018-09-21  Jeff Law  <law@redhat.com>
5508         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5509         vr_values::cleanup_edges_and_switches.
5510         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5511         vr_values class.
5512         (identify_jump_threads): Remove EDGE_IGNORE handling.
5513         (execute_vrp): Move handling of to_remove_edges and
5514         to_update_switch_stmts into vr_values class member functions.
5515         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5516         (to_update_switch_stmts): Likewise.
5517         * vr-values.c: Include cfghooks.h.
5518         (vr_values::vr_values): Initialize to_remove_edges and
5519         to_update_switch_stmts.
5520         (vr_values::~vr_values): Verify to_remove_edges and
5521         to_update_switch_stmts are empty.
5522         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5523         (vr_values::cleanup_edges_and_switches): New member function.
5524         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5525         function.  Add new data members.
5527 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
5529         PR tree-optimization/87309
5530         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
5531         calls with pflags and alt_flags.
5532         (selftest::test_capture_of_dump_calls): Add test of interaction of
5533         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
5535 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5537         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
5539 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5541         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5543 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5545         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
5546         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
5547         (VXWORKS_LIBS_RTP): Minor reordering.
5549 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5551         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
5552         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5554 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5556         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
5557         (PTRDIFF_TYPE): Likewise.
5559 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5561         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
5562         triplet, similar to support for VxWorks7.
5563         * config/vxworks-dummy.h: Provide a default definition
5564         of TARGET_VXWORKS64 to 0.
5566 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5568         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
5569         * config/vxworks-dummy.h: here.
5571 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5573         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
5575 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
5576             Bo Zhou  <zbo.zhou@hisilicon.com>
5578         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5579         * config/aarch64/aarch64-tune.md: Regenerated.
5580         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
5581         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
5582         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
5584 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
5585             Julian Brown  <julian@codesourcery.com>
5587         * builtins.c (get_builtin_sync_mem): Handle address spaces.
5589 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5591         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
5592         if the call takes a static chain.
5594 2018-09-21  Martin Liska  <mliska@suse.cz>
5596         * auto-profile.c (autofdo_source_profile::read): Do not
5597         set sum_all.
5598         (read_profile): Do not add working sets.
5599         (read_autofdo_file): Remove sum_all.
5600         (afdo_callsite_hot_enough_for_early_inline): Remove const
5601         qualifier.
5602         * coverage.c (struct counts_entry): Remove gcov_summary.
5603         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
5604         do not support GCOV_TAG_PROGRAM_SUMMARY.
5605         (get_coverage_counts): Remove summary and expected
5606         arguments.
5607         * coverage.h (get_coverage_counts): Likewise.
5608         * doc/gcov-dump.texi: Remove -w option.
5609         * gcov-dump.c (dump_working_sets): Remove.
5610         (main): Do not support '-w' option.
5611         (print_usage): Likewise.
5612         (tag_summary): Likewise.
5613         * gcov-io.c (gcov_write_summary): Do not dump
5614         histogram.
5615         (gcov_read_summary): Likewise.
5616         (gcov_histo_index): Remove.
5617         (gcov_histogram_merge): Likewise.
5618         (compute_working_sets): Likewise.
5619         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
5620         it not obsolete.
5621         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
5622         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
5623         (GCOV_HISTOGRAM_SIZE): Remove.
5624         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
5625         (struct gcov_summary): Simplify rapidly just
5626         to runs and sum_max fields.
5627         (gcov_histo_index): Remove.
5628         (NUM_GCOV_WORKING_SETS): Likewise.
5629         (compute_working_sets): Likewise.
5630         * gcov-tool.c (print_overlap_usage_message): Remove
5631         trailing empty line.
5632         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
5633         (output_lines): Remove program related line.
5634         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
5635         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
5636         histogram.
5637         (input_profile_summary): Do not read it.
5638         (merge_profile_summaries): And do not merge it.
5639         (input_symtab): Do not call removed function.
5640         * modulo-sched.c (sms_schedule): Do not print sum_max.
5641         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
5642         removed when histogram method was invented.
5643         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
5644         mode.
5645         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
5646         GCOV coding style.
5647         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
5648         and dump selected value.
5649         * profile.c (add_working_set): Remove.
5650         (get_working_sets): Likewise.
5651         (find_working_set): Likewise.
5652         (get_exec_counts): Do not work with working sets.
5653         (read_profile_edge_counts): Do not inform as sum_max is removed.
5654         (compute_branch_probabilities): Likewise.
5655         (compute_value_histograms): Remove argument for call of
5656         get_coverage_counts.
5657         * profile.h: Do not make gcov_summary const.
5659 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
5661         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
5663 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5665         PR tree-optimization/86990
5666         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
5667         Check that the entire merged store group is made of constants only for
5668         overlapping stores.
5670 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
5672         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
5673         (VTABLE_VERIFICATION_SPEC): Likewise.
5674         (SANITIZER_EARLY_SPEC): Likewise.
5675         (SANITIZER_SPEC): Likewise.
5676         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
5677         * doc/invoke.texi (Link Options): Document -r.
5679 2018-09-20  Richard Biener <rguenther@suse.de>
5681         PR middle-end/87054
5682         * gimplify.c (gimplify_expr): Retain alignment of
5683         addressable lvalue in dereference.
5685 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
5687         PR bootstrap/87013
5688         * configure.ac: Check for .loc is_stmt support.
5689         * configure, config.in: Rebuilt.
5690         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
5691         if not supported.
5693 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5695         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
5696         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
5697         -misel=no.
5699 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5701         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5702         VECTOR_OTHER.
5703         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
5704         case VECTOR_OTHER.
5706 2018-09-20  Marek Polacek  <polacek@redhat.com>
5708         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
5710 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5712         PR tree-optimization/87288
5713         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
5714         into account when determining PEELING_FOR_NITERS.
5716 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5718         PR tree-optimization/86877
5719         * tree-vect-loop.c (vect_analyze_loop_2): Call
5720         vect_verify_datarefs_alignment.
5722 2018-09-19  Marek Polacek  <polacek@redhat.com>
5724         * doc/invoke.texi: Document -Wclass-conversion.
5726 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
5728         * config/pa/pa.c (pa_adjust_priority): Delete.
5729         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
5731         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
5732         (atomic_storehi): Likewise.
5733         (atomic_storesi): Likewise.
5734         (atomic_loaddi): Restore compare and swap exchange loop code.
5736 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
5738         PR rtl-optimization/86902
5739         * combine.c (try_combine): When changing the CC mode used, don't change
5740         an unrelated mode in other_insn to that new CC mode.
5742 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
5744         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
5745         with %T in place of calls to dump_generic_expr.
5746         (prune_runtime_alias_test_list): Likewise.
5747         (create_runtime_alias_checks): Likewise.
5748         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
5749         (vect_analyze_data_ref_dependence): Likewise.
5750         (vect_slp_analyze_data_ref_dependence): Likewise.
5751         (vect_record_base_alignment): Likewise.  Use %G in place of call
5752         to dump_gimple_stmt.
5753         (vect_compute_data_ref_alignment): Likewise.
5754         (verify_data_ref_alignment): Likewise.
5755         (vect_find_same_alignment_drs): Likewise.
5756         (vect_analyze_group_access_1): Likewise.
5757         (vect_analyze_data_ref_accesses): Likewise.
5758         (dependence_distance_ge_vf): Likewise.
5759         (dump_lower_bound): Likewise.
5760         (vect_prune_runtime_alias_test_list): Likewise.
5761         (vect_find_stmt_data_reference): Likewise.
5762         (vect_analyze_data_refs): Likewise.
5763         (vect_create_addr_base_for_vector_ref): Likewise.
5764         (vect_create_data_ref_ptr): Likewise.
5765         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
5766         (vect_can_advance_ivs_p): Likewise.
5767         (vect_update_ivs_after_vectorizer): Likewise.
5768         (vect_gen_prolog_loop_niters): Likewise.
5769         (vect_prepare_for_masked_peels): Likewise.
5770         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5771         (vect_determine_vectorization_factor): Likewise.
5772         (vect_is_simple_iv_evolution): Likewise.
5773         (vect_analyze_scalar_cycles_1): Likewise.
5774         (vect_analyze_loop_operations): Likewise.
5775         (report_vect_op): Likewise.
5776         (vect_is_slp_reduction): Likewise.
5777         (check_reduction_path): Likewise.
5778         (vect_is_simple_reduction): Likewise.
5779         (vect_create_epilog_for_reduction): Likewise.
5780         (vect_finalize_reduction:): Likewise.
5781         (vectorizable_induction): Likewise.
5782         (vect_transform_loop_stmt): Likewise.
5783         (vect_transform_loop): Likewise.
5784         (optimize_mask_stores): Likewise.
5785         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
5786         (vect_split_statement): Likewise.
5787         (vect_recog_over_widening_pattern): Likewise.
5788         (vect_recog_average_pattern): Likewise.
5789         (vect_determine_min_output_precision_1): Likewise.
5790         (vect_determine_precisions_from_range): Likewise.
5791         (vect_determine_precisions_from_users): Likewise.
5792         (vect_mark_pattern_stmts): Likewise.
5793         (vect_pattern_recog_1): Likewise.
5794         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5795         (vect_record_max_nunits): Likewise.
5796         (vect_build_slp_tree_1): Likewise.
5797         (vect_build_slp_tree_2): Likewise.
5798         (vect_print_slp_tree): Likewise.
5799         (vect_analyze_slp_instance): Likewise.
5800         (vect_detect_hybrid_slp_stmts): Likewise.
5801         (vect_detect_hybrid_slp_1): Likewise.
5802         (vect_slp_analyze_operations): Likewise.
5803         (vect_slp_analyze_bb_1): Likewise.
5804         (vect_transform_slp_perm_load): Likewise.
5805         (vect_schedule_slp_instance): Likewise.
5806         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5807         (vect_mark_stmts_to_be_vectorized): Likewise.
5808         (vect_init_vector_1): Likewise.
5809         (vect_get_vec_def_for_operand): Likewise.
5810         (vect_finish_stmt_generation_1): Likewise.
5811         (vect_check_load_store_mask): Likewise.
5812         (vectorizable_call): Likewise.
5813         (vectorizable_conversion): Likewise.
5814         (vectorizable_operation): Likewise.
5815         (vectorizable_load): Likewise.
5816         (vect_analyze_stmt): Likewise.
5817         (vect_is_simple_use): Likewise.
5818         (vect_get_vector_types_for_stmt): Likewise.
5819         (vect_get_mask_type_for_stmt): Likewise.
5820         * tree-vectorizer.c (increase_alignment): Likewise.
5822 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
5824         * doc/rtl.texi: Adjust vec_select description.
5825         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
5826         non-constant selectors.
5828 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
5830         * config/aarch64/aarch64-protos.h
5831         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
5832         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
5833         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
5834         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
5835         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
5836         (AARCH64_FL_PROFILE): Move index so flags are ordered.
5837         (AARCH64_ISA_RCPC8_4): New flag.
5838         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
5839         to aarch64_offset_9bit_signed_unscaled_p.
5840         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
5841         and use stlur.
5842         * config/aarch64/constraints.md (Ust): New constraint.
5843         * config/aarch64/predicates.md.
5844         (aarch64_9bit_offset_memory_operand): New predicate.
5845         (aarch64_rcpc_memory_operand): New predicate.
5847 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
5849         PR rtl-optimization/87361
5850         * rtlanal.c (nonzero_bits1): Revert accidental change.
5852 2018-09-19  Richard Biener  <rguenther@suse.de>
5854         PR tree-optimization/87349
5855         PR tree-optimization/87342
5856         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
5858 2018-09-18  Marek Polacek  <polacek@redhat.com>
5860         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
5861         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
5863 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5865         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
5867 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5869         PR rtl-optimization/86882
5870         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
5872 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
5874         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
5875         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
5877 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
5879         PR other/87353
5880         * doc/invoke.texi (Link Options): Fix formatting and grammar.
5882 2018-09-18  Richard Biener  <rguenther@suse.de>
5884         PR middle-end/63155
5885         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
5886         (compute_samebase_partition_bases): Likewise.
5887         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
5888         (gimple_can_coalesce_p): Simplify.
5890 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
5892         Handle a library implementation of ffs calling __builtin_ffs.
5893         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
5894         (mmix_init_libfuncs): New function: make __builtin_ffs expand
5895         to __ffsdi2.
5897 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5899         * diagnostic-show-locus.c (class layout_range): Add field
5900         "m_original_idx".
5901         (layout_range::layout_range): Add "original_idx" param and use it
5902         to initialize new field.
5903         (make_range): Use 0 for original_idx.
5904         (layout::layout): Pass in index to calls to
5905         maybe_add_location_range.
5906         (layout::maybe_add_location_range): Add param "original_idx" and
5907         pass it on to layout_range.
5908         (layout::print_any_labels): Pass on range->m_original_idx to
5909         get_text call.
5910         (gcc_rich_location::add_location_if_nearby): Use 0 for
5911         original_idx.
5912         * gcc-rich-location.h (text_range_label::get_text): Update for new
5913         param.
5914         (range_label_for_type_mismatch::get_text): Likewise.
5916 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5918         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
5920 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5922         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
5923         format_string_diagnostic_t.
5924         (fmtwarn_n): Likewise.
5925         * substring-locations.c
5926         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
5927         (format_warning_n_va): Convert to...
5928         (format_string_diagnostic_t::emit_warning_n_va): ...this.
5929         (format_warning_va): Convert to...
5930         (format_string_diagnostic_t::emit_warning_va): ...this.
5931         (format_warning_at_substring): Convert to...
5932         (format_string_diagnostic_t::emit_warning): ...this.
5933         (format_warning_at_substring_n): Convert to...
5934         (format_string_diagnostic_t::emit_warning_n): ...this.
5935         * substring-locations.h (class format_string_diagnostic_t): New
5936         class.
5937         (format_warning_va): Convert to
5938         format_string_diagnostic_t::emit_warning_va.
5939         (format_warning_n_va): Convert to
5940         format_string_diagnostic_t::emit_warning_n_va.
5941         (format_warning_at_substring): Convert to
5942         format_string_diagnostic_t::emit_warning.
5943         (format_warning_at_substring_n): Convert to
5944         format_string_diagnostic_t::emit_warning_n.
5946 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
5947             Bernd Schmidt <bernds_cb1@t-online.de>
5949         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
5950         SImode args.
5952 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5954         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
5955         operand 0 predicate to nonimmediate operand.
5956         (rint<mode>2_frndint): Remove insn pattern.
5957         (rint<mode>2): Change operand 1 predicate to general_operand.
5958         Extend operand 1 to XFmode and generate rintxf2 insn.
5959         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
5960         Do not use X87MODEF mode macro.
5961         (frndintxf2_<rounding>_i387): Rename from
5962         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
5963         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
5964         to XFmode and generate significandxf3 insn.
5966 2018-09-17  Richard Biener  <rguenther@suse.de>
5968         PR tree-optimization/87328
5969         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
5970         visiting unexecutable backedges when not iterating.
5971         (do_rpo_vn): Mark all edges not executable even when not
5972         iterating.
5974 2018-09-17  Martin Jambor  <mjambor@suse.cz>
5976         PR c/63886
5977         * doc/invoke.texi (Warning Options): Likewise.
5979 2018-09-17  Richard Biener  <rguenther@suse.de>
5981         PR tree-optimization/87301
5982         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
5983         clean EH info from leftover copy assignments.
5985 2018-09-17  Martin Liska  <mliska@suse.cz>
5987         PR gcov-profile/85871
5988         * gcov.c (output_intermediate_file): Fix out of bounds
5989         access.
5991 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
5993         * config/arc/arc.c: Object attributes for core4 not reflected
5994         correctly.
5995         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
5996         core3).
5998 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6000         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6002 2018-09-17  Martin Liska  <mliska@suse.cz>
6004         * doc/gcov.texi: Document new option --use-hotness-colors.
6005         * gcov.c (struct source_info): Declare new field.
6006         (source_info::source_info): Set default for maximum_count.
6007         (print_usage): Add new -q option.
6008         (process_args): Process it.
6009         (accumulate_line_info): Save src->maximum_count.
6010         (output_line_beginning): Make color line number if
6011         flag_use_hotness_colors is set.
6012         (output_line_details): Pass default argument value.
6013         (output_lines): Pass src->maximum_count.
6015 2018-09-17  Martin Liska  <mliska@suse.cz>
6017         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6018         Use processor_names table.
6019         * config/i386/i386.c (ix86_default_align): Use
6020         processor_cost_table for alignment values.
6021         (ix86_option_override_internal): Use processor_names.
6022         (ix86_function_specific_print): Likewise.
6023         * config/i386/i386.h (struct processor_costs):
6024         Add alignment values.
6025         (struct ptt): Remove and replace with const char *.
6026         * config/i386/x86-tune-costs.h (struct processor_costs):
6027         Declare default alignments for all costs.
6029 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6031         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6032         symbolics or VR_VARYING ranges for ABS_EXPR.
6033         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6034         when range will wrap.
6036 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6038         PR middle-end/86864
6039         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6040         before and after a JUMP_TABLE_DATA.
6042 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6044         PR middle-end/87188
6045         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6046         when one operand is a function pointer.  Use POINTER_TYPE_P and
6047         FUNC_OR_METHOD_TYPE_P.
6048         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6049         * fold-const.c (build_range_check): Likewise.
6050         * match.pd (simple_comparison): Likewise.
6052 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6054         PR c/82967
6055         * spellcheck.c (get_edit_distance_cutoff): New function.
6056         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6057         (selftest::test_get_edit_distance_one_way): ...this.
6058         (selftest::test_get_edit_distance_unit): Rename to...
6059         (selftest::test_get_edit_distance_both_ways): ...this.
6060         (selftest::test_edit_distances): Move tests to this new function,
6061         and test some more pairs of strings.  Update for above renaming.
6062         (selftest::get_old_cutoff): New function.
6063         (selftest::test_get_edit_distance_cutoff): New function.
6064         (selftest::assert_suggested_for): New function.
6065         (ASSERT_SUGGESTED_FOR): New macro.
6066         (selftest::assert_not_suggested_for): New function.
6067         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6068         (selftest::test_suggestions): New function.
6069         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6070         tests to selftest::test_edit_distances and call it.  Add calls to
6071         selftest::test_get_edit_distance_cutoff and
6072         selftest::test_suggestions.
6073         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6074         (best_match::consider): Replace hard-coded cutoff calculation with
6075         a call to...
6076         (best_match::get_cutoff): New declaration.
6077         (best_match::get_best_meaningful_candidate): Likewise.
6079 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6081         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6083 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6085         revert:
6086         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6088         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6089         terminated string literal.
6091 2018-09-14  Martin Sebor  <msebor@redhat.com>
6093         * builtins.c (unterminated_array): Handle ARRAY_REF.
6094         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6095         * builtins.h (unterminated_array): Declare extern.
6096         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6097         arrays.
6098         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6099         calls.
6101 2018-09-14  Martin Sebor  <msebor@redhat.com>
6102             Jeff Law  <law@redhat.com>
6104         * builtins.c (unterminated_array): New.
6105         (expand_builtin_strcpy): Adjust.
6106         (expand_builtin_strcpy_args): Detect unterminated arrays.
6107         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6108         unterminated arrays.
6109         * gimple-fold.h (get_maxval_strlen): Add argument.
6110         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6112         * gimple-fold.c (get_range_strlen): Add argument.
6113         (get_maxval_strlen): Adjust.
6114         * gimple-fold.h (get_range_strlen): Add argument.
6116 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6118         * config/i386/movdirintrin.h: Fix copyright year.
6120 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6122         * reg-stack.c: Include regs.h.
6123         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6124         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6125         FIRST_STACK_REG, not DFmode.
6126         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6127         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6128         (change stack): Default register mode to the reg_raw_mode of
6129         FIRST_STACK_REG, not DFmode.
6130         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6131         (*swapxf): Rename from swapxf.
6133 2018-09-14  Carl Love  <cel@us.ibm.com>
6135         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6136         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6138 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
6140         PR target/87224
6141         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6142         alternatives.
6144 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
6146         PR target/85628
6147         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6149 2018-09-14  Jason Merrill  <jason@redhat.com>
6151         Fix --enable-gather-detailed-mem-stats.
6152         * hash-table.c (hash_table_usage): Change from variable to function.
6153         * hash-table.h: Adjust.
6154         * Makefile.in: Add missing dependencies on hash-table.h.
6156 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6158         PR tree-optimization/87259
6159         PR lto/87283
6160         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6161         execute_cse_reciprocals_1 has tried transforming.
6163 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
6165         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6166         VR_VARYING for PLUS/MINUS_EXPR.
6168 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6170         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6171         formatting.
6173 2018-09-14  Richard Biener  <rguenther@suse.de>
6175         PR middle-end/63155
6176         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6177         bits for the merged partition.
6179 2018-09-13  Martin Sebor  <msebor@redhat.com>
6180             Bernd Edlinger  <bernd.edlinger@hotmail.de>
6182         * builtins.h (c_srlen): Add argument.
6183         * builtins.c (warn_string_no_nul): New function.
6184         (c_strlen): Add argument and use it.  Update recursive calls.
6185         Pass DECL argument to string_constant to get info on non
6186         terminated strings.  Update *NONSTR as needed.
6187         (fold_builtin_strlen): Add argument to calls to c_strlen.
6188         Warn for unterminated arrays.
6189         (warn_string_no_null): Add prototype.
6190         * expr.c (string_constant): Update arguments.  Update recursive
6191         calls appropriately.  Detect missing NUL terminator and outermost
6192         declaration its missing in.
6193         Improve checks for arrays with nonzero lower bound or elements
6194         that are not a single byte.  Simplify offset computation.
6195         Simplify checks for non-NUL terminated strings.
6196         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6197         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6199 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6201         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6202         correctly.
6203         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
6204         argument.  Fix range checks.
6205         * fold-const.h (c_getstr): Adjust protoype.
6206         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6207         string is constant but contains no NUL byte.
6209         * expr.c (string_constant): Adjust function comment.
6210         Remove bogus check for zero termination.
6212         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6214         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6215         (get_constant_size): Don't make STRING_CSTs larger than they are.
6216         (check_string_literal): New check function for STRING_CSTs.
6217         (output_constant): Use it.
6219 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6221         PR target/86812
6222         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6224 2018-09-13  Richard Biener  <rguenther@suse.de>
6226         PR tree-optimization/87263
6227         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6228         (struct unwind_state): Add max_rpo field.
6229         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6230         Compute max_rpo, the max RPO number a block can be backwards reached
6231         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6232         separating it from the iterating mode.
6234 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6236         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6237         (rfs_decision): New scheduling decision.
6239 2018-09-13  Richard Biener  <rguenther@suse.de>
6241         PR bootstrap/87134
6242         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6243         (vn_nary_op_insert_pieces_predicated): Do not write useless
6244         valid_dominated_by_p entry outside of the allocated storage.
6246 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6247             Tom de Vries  <tdevries@suse.de>
6249         PR debug/86985
6250         * dwarf2out.c (is_c): New function.
6251         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6253 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6255         PR target/85628
6256         * config/aarch64/aarch64.md (*aarch64_bfxil):
6257         Define.
6258         * config/aarch64/constraints.md (Ulc): Define.
6259         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6260         Define.
6261         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6262         New function.
6264 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6266         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6267         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6268         aarch64_layout_frame call.
6269         (aarch64_expand_epilogue): Likewise.
6270         (aarch64_initial_elimination_offset): Likewise.
6271         (aarch64_get_separate_components): Likewise.
6272         (aarch64_use_return_insn_p): Likewise.
6273         (aarch64_layout_frame): Remove unneeded check.
6275 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6277         * configure.ac: Only append
6278         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6279         gcc_config_arguments if it was never reconfigured or last reconfigure
6280         was with different arguments.
6281         * configure: Regenerated.
6283 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6284             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6286         PR middle-end/87290
6287         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6288         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6290 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6292         PR tree-optimization/87287
6293         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6294         X % C == 0 to X % (unsigned) C == 0 optimization to ...
6295         * match.pd (X % C == 0): ... here.  New optimization.
6297 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6299         PR middle-end/82853
6300         * expr.h (maybe_optimize_mod_cmp): Declare.
6301         * expr.c (mod_inv): New function.
6302         (maybe_optimize_mod_cmp): New function.
6303         (do_store_flag): Use it.
6304         * cfgexpand.c (expand_gimple_cond): Likewise.
6306 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
6307             Julian Brown  <julian@codesourcery.com>
6309         PR middle-end/86336
6310         * gimplify.c (gimplify_scan_omp_clauses): Set
6311         target_firstprivatize_array_bases in OpenACC parallel and kernels
6312         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6313         OpenACC data regions.
6315 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
6317         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6318         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6320 2018-09-12  Richard Biener  <rguenther@suse.de>
6322         PR tree-optimization/87280
6323         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6324         edge but unreachable target.
6325         (do_rpo_vn): For conservatively handling a PHI only mark
6326         the backedge executable but not the block reachable.
6328 2018-09-12  Richard Biener  <rguenther@suse.de>
6330         PR tree-optimization/87266
6331         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6332         visited blocks.
6334 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
6336         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6337         constants.
6338         ("trunc<BFP:mode><DFP_ALL:mode>2")
6339         ("trunc<DFP_ALL:mode><BFP:mode>2")
6340         ("extend<BFP:mode><DFP_ALL:mode>2")
6341         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6342         according to the target operand type.
6344 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6345             Andreas Krebbel  <krebbel@linux.ibm.com>
6347         PR tree-optimization/86844
6348         * gimple-ssa-store-merging.c
6349         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6350         there are any overlapping stores in between them, make sure they are
6351         also coalesced or we give up completely.
6353 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6355         PR middle-end/87248
6356         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6357         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
6359 2018-09-12  Tom de Vries  <tdevries@suse.de>
6361         * common.opt (gdescribe-dies): Add option.
6362         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6363         attribute for artifical and nameless decls.
6364         (dwarf2out_register_external_die): Add description attribute to
6365         external reference die.
6366         (add_desc_attribute): New functions.
6367         (gen_subprogram_die): Add description attribute to
6368         DW_TAG_call_site_parameter.
6369         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6370         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6371         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6372         -gno-describe-dies.
6373         (@item -gdescribe-dies): Add.
6375 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6377         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6378         (extract_range_from_binary_expr_1: Call
6379         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6380         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6381         depend on sign.
6383 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6385         * gimple-ssa-warn-alloca.c
6386         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6387         field for ALLOCA_BOUND_*_LARGE.
6389 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6391         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6393 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6395         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6396         for clobbers.  Remove obsolete comment.
6398 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6400         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6401         mpxchk, mpxld and mpxst types.
6402         (define_attr length_immediate): Remove all processing of mpx types.
6403         (define_attr prefix_0f): Ditto.
6404         (define_attr memory): Ditto.
6406 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6408         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6409         (log<mode>2): Change operand 1 predicate to general_operand.
6410         Extend operand 1 to XFmode and generate logxf3 insn.
6411         (log10<mode>2): Change operand 1 predicate to general_operand.
6412         Extend operand 1 to XFmode and generate log10xf3 insn.
6413         (log2<mode>2): Change operand 1 predicate to general_operand.
6414         Extend operand 1 to XFmode and generate log2xf3 insn.
6415         (fyl2xp1_extend<mode>xf3_i387): Remove.
6416         (log1p<mode>2): Change operand 1 predicate to general_operand.
6417         Extend operand 1 to XFmode and generate log1pxf3 insn.
6418         (fxtract_extend<mode>xf3_i387): Remove.
6419         (logb<mode>2): Change operand 1 predicate to general_operand.
6420         Extend operand 1 to XFmode and generate logbxf3 insn.
6421         (ilogb<mode>2): Change operand 1 predicate to general_operand.
6422         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6423         (significand<mode>2): Change operand 1 predicate to general_operand.
6424         Extend operand 1 to XFmode and generate significandxf3 insn.
6426 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6428         * gcc.c (perror_with_name, pfatal_with_name): Delete.
6429         (load_specs): Use fatal_error.
6430         (DELETE_IF_ORDINARY, process_command): Use error.
6431         (execute, run_attempt): Use fatal_error.
6433 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
6435         * diagnostic-core.h (sorry_at): New prototype.
6436         * diagnostic.c (sorry_at): New function.
6438 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
6440         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6441         by zero as VR_UNDEFINED.
6443 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6445         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6446         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6447         (<sincos>mode2): New expander.
6448         (sincos_extend<mode>xf3_i387): Remove insn pattern.
6449         (sincos -> sin, cos splitters): Remove splitter patterns.
6450         (sincos<mode>3): Change operand 2 predicate to general_operand.
6451         Extend operand 2 to XFmode and generate sincosxf3 insn.
6452         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6453         Change operand 3 predicate to const1_operand.
6454         (fptan_extend<mode>xf4_i387): Remove insn pattern.
6455         (tanxf2): Update operands in the call to fptanxf4_i387.
6456         (tan<mode>2): Change operand 1 predicate to general_operand.
6457         Extend operand 1 to XFmode and generate tanxf3 insn.
6458         (atan2xf3): Rename from *fpatanxf3_i387.
6459         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6460         (atan2xf3): Remove expander.
6461         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
6462         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6463         (atan<mode>2): Change operand 1 predicate to general_operand.
6464         Extend operand 1 to XFmode and generate atanxf3 insn.
6466 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6468         * config/i386/i386.md (x87/SSE constant load splitter): Use
6469         memory_operand instead of nonimmediate_operand for input operand
6470         predicate.
6472 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
6474         * config/i386/i386.md (float partial SSE register stall splitter): Move
6475         splitter near its instruction pattern.
6476         (float_extend partial SSE register stall splitter): Ditto.
6477         (float_truncate partial SSE register stall splitter): Ditto.
6479 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
6481         PR target/86794
6482         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6483         to speculation_safe_value_not_needed.
6485         PR target/85666
6486         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6487         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6488         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6489         leaf_function_p, instead use has_hard_reg_initial_val.
6491 2018-09-09  Nathan Sidwell  <nathan@acm.org>
6493         * gcc.h (pfatal_with_name): Don't declare here.
6494         * gcc.c (pfatal_with_name): Make static.
6496 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
6498         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6499         earlyclobber.
6501 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
6503         PR rtl-optimization/85458
6504         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6505         priority hook to reduce the priority of EXPR.
6507 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
6509         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6510         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
6511         with X87_ENABLE_FLOAT.  Remove preparation code.
6512         (*float<SWI48:mode><MODEF:mode>2): Rename from
6513         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
6514         math using "enabled" attribute.
6515         (*floatdi<MODEF:mode>2_i387): Rename from
6516         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
6517         enable for 32bit targets only.
6518         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6519         splitter.
6520         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6521         as operand 1 predicate.  Rewrite as define_insn_and_split.
6522         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6524 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6526         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
6527         to fallthru to FLOAT case.
6529 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6531         PR target/86731
6532         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
6533         around folding of vec_sl to handle out of range shift values.
6535 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6537         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
6538         Update callers to gen_fix_trunc<mode>_i387_fisttp
6539         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
6540         nonimmediate_operand.
6541         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
6542         and corresponding splitters.
6543         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
6544         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
6545         (fix_truncdi_i387_with_temp): Remove insn pattern
6546         and corresponding splitters.
6547         (fix_trunc<mode>_i387): Change operand 0 predicate to
6548         nonimmediate_operand.
6549         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
6550         and corresponding splitters.
6551         (*fistdi2_1): Remove.
6552         (fistdi2): Ditto.
6553         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
6554         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
6555         (*fist<mode>2_1): Remove.
6556         (fist<mode>2): Ditto.
6557         (fist<mode>2_with_temp): Remove insn pattern and corresponding
6558         splitters.
6559         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
6560         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
6561         (fistdi2_<rounding>): Change operand 0 predicate to
6562         nonimmediate_operand.
6563         (fistdi2_<rounding>_with_temp): Remove insn pattern
6564         and corresponding splitters.
6565         (fist<mode>2_<rounding>): Change operand 0 predicate to
6566         nonimmediate_operand.
6567         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
6568         and corresponding splitters.
6570         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
6572 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6574         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6575         the init value.
6577 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6579         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
6580         early gimple folding of vec_splat().
6581         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
6582         * gimple-fold.h: Add an extern define for tree_vec_extract().
6584 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6586         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
6587         wrappers around TREE_TYPE comparisons.
6589 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6591         PR target/80080
6592         * config/s390/predicates.md: Add nonsym_memory_operand.
6593         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
6594         contains a SYMBOL_REF, load it into an intermediate pseudo.
6595         (s390_emit_compare_and_swap): Legitimize operand.
6596         * config/s390/s390.md: Use the new nonsym_memory_operand
6597         with UNSPECV_CAS patterns.
6599 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6601         PR target/80080
6602         * config/s390/s390-passes.def: New file.
6603         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
6604         declaration.
6605         (make_pass_s390_early_mach): Add declaration.
6606         * config/s390/s390.c (make_pass_s390_early_mach):
6607         (s390_option_override): Remove dynamic registration.
6608         * config/s390/t-s390: Add s390-passes.def.
6610 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6612         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6613         Remove UNSPEC_LTREL_BASE check.
6614         (annotate_constant_pool_refs): Likewise.
6615         (find_constant_pool_ref): Likewise.
6616         (find_ltrel_base): Removed.
6617         (replace_ltrel_base): Removed.
6618         (s390_mainpool_finish): Remove replace_ltrel_base call.
6619         (s390_chunkify_start): Remove pending LTREL_BASE logic.
6620         (s390_chunkify_finish): Remove replace_ltrel_base call.
6621         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
6623 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
6625         PR target/86779
6626         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6627         to speculation_safe_value_not_needed.
6629 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
6630             Bernd Schmidt  <bernds_cb1@t-online.de>
6632         * config/nvptx/nvptx-opts.h: New file.
6633         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
6634         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
6635         (ASM_SPEC): Define.
6636         (TARGET_SM35): New macro.
6637         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
6638         correct predicate.
6639         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
6640         values.
6641         (misa=): New option.
6642         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
6644 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
6646         * config/i386/i386.md (truncdfsf2): Remove expander.
6647         (truncdfsf2_with_temp): Ditto.
6648         (truncxf<mode>2): Ditto.
6649         (*truncdfsf_fast_mixed): Remove insn pattern.
6650         (*truncdfsf_fast_i387): Ditto.
6651         (*truncdfsf_mixed): Ditto.
6652         (*truncdfsf_i387): Ditto.
6653         (*truncdfsf2_i387_1): Ditto.
6654         (*truncxfsf2_mixed): Ditto.
6655         (*truncxfdf2_mixed): Ditto.
6656         (*truncxf<mode>2_i387_noop): Ditto. Update callers
6657         to call gen_truncxf<mode>2 instead.
6658         (*truncxf<mode>2_i387): Remove.
6659         (reg->reg splitters): Remove splitter pattern.
6660         (reg->mem splitters): Ditto.
6662         (truncdfsf2): New insn pattern.
6663         (truncxf<mode>2): Ditto.
6665 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6667         * tree-ssa-math-opts.c (is_mult_by): New function.
6668         (is_square_of): Use the above.
6669         (optimize_recip_sqrt): New function.
6670         (pass_cse_reciprocals::execute): Use the above.
6672 2018-09-05  Richard Biener  <rguenther@suse.de>
6674         PR bootstrap/87134
6675         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
6676         to zero-init the emplaced vec.
6678 2018-09-05  Martin Liska  <mliska@suse.cz>
6680         PR tree-optimization/87205
6681         * tree-switch-conversion.c (pass_lower_switch::execute):
6682         Group cases for switch statements.
6684 2018-09-05  Richard Biener  <rguenther@suse.de>
6686         PR tree-optimization/87217
6687         * tree-ssa-sccvn.c (vuse_valueize): New.
6688         (vn_reference_lookup_pieces): Use it.
6689         (vn_reference_lookup): Likewise.
6691 2018-09-05  Nathan Sidwell  <nathan@acm.org>
6693         PR c++/87137
6694         * stor-layout.c (place_field): Scan forwards to check last
6695         bitfield when ms_bitfield_placement is in effect.
6697 2018-09-05  Richard Biener  <rguenther@suse.de>
6699         PR bootstrap/87225
6700         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
6701         return.
6703 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6704             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
6706         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
6707         * config.gcc (extra_objs): Build it.
6708         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
6709         Likewise.
6710         * config/aarch64/aarch64-passes.def
6711         (pass_tag_collision_avoidance): New pass.
6712         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
6713         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
6714         (aarch64_classify_address): Remove static qualifier.
6715         (aarch64_address_info, aarch64_address_type): Move to...
6716         * config/aarch64/aarch64-protos.h: ... here.
6717         (make_pass_tag_collision_avoidance): New function.
6718         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
6719         New tuning flag.
6721 2018-09-05  Martin Liska  <mliska@suse.cz>
6723         * doc/gcov.texi: Update documentation of humar
6724         readable mode.
6725         * gcov.c (format_count): Print one decimal place, it provides
6726         more fine number of situations like '1G' vs. '1.4G'.
6728 2018-09-05  Martin Liska  <mliska@suse.cz>
6730         PR target/87164
6731         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
6732         * optc-gen.awk: Allow 'Var' for Deprecated options in order
6733         to generate a MASK value.
6735 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
6737         PR debug/86593
6738         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
6739         if frame pointer isn't used.
6740         (compute_frame_pointer_to_fb_displacement): Likewise.
6742 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
6744         PR target/87198
6745         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
6746         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
6747         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
6748         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
6749         and OPTION_MASK_ISA_XSAVEC_UNSET.
6751 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
6753         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
6754         XOR operations in NAND case.
6756 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6758         * wide-int-range.cc (wide_int_range_convert): New.
6759         * wide-int-range.h (wide_int_range_convert): New.
6760         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
6761         code into wide_int_range_convert.
6762         (extract_range_into_wide_ints): Do not munge anti range constants
6763         into the entire domain.  Just return the range back.
6765 2018-09-04  Martin Liska  <mliska@suse.cz>
6767         * genmatch.c (output_line_directive): Add new argument
6768         fnargs.
6769         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
6771 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6773         * doc/invoke.texi (Option Summary): Add whitespace.
6775         * doc/invoke.texi (Option Summary): Add -Waligned-new.
6777 2018-09-04  Richard Biener  <rguenther@suse.de>
6779         PR tree-optimization/87211
6780         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
6781         backedge value we're supposed to treat as VARYING also number
6782         the PHI to VARYING in case it got a different value-number already.
6784 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6786         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
6787         (extract_range_from_binary_expr_1): Do not call
6788         vrp_can_optimize_bit_op.
6789         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
6790         static.
6791         (wide_int_range_get_mask_and_bounds): New.
6792         (wide_int_range_optimize_bit_op): New.
6793         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
6794         (wide_int_range_bit_and): Same.
6795         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
6796         (wide_int_range_optimize_bit_op): New.
6797         (wide_int_range_get_mask_and_bounds): New.
6799 2018-09-04  Richard Biener  <rguenther@suse.de>
6801         PR tree-optimization/87176
6802         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
6803         variable.  When value-numbering a virtual PHI node make sure
6804         to not value-number to the backedge value.
6806 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6808         * doc/extend.texi (Long Long, Hex Floats): Document support for
6809         long long and hex floats in more recent versions of ISO C++.
6811 2018-09-03  Richard Biener  <rguenther@suse.de>
6813         PR tree-optimization/87177
6814         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
6815         cleanup.
6817 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6819         * bb-reorder.c (edge_order): Convert to C-qsort-style
6820         tri-state comparator.
6821         (reorder_basic_blocks_simple): Change std::stable_sort to
6822         gcc_stablesort.
6824 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6826         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
6827         tri-state comparator.
6828         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
6830 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6832         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
6833         (mergesort): ... here as maximum count for using netsort.
6834         (gcc_qsort): Set nlim to 3 if stable sort is requested.
6835         (gcc_stablesort): New.
6836         * system.h (gcc_stablesort): Declare.
6838 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6840         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
6841         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
6842         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
6844 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
6846         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
6847         lxsdx and stxsdx alternatives.
6848         (*mov<mode>_hardfloat64): Ditto.
6849         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
6851 2018-09-03  Richard Biener  <rguenther@suse.de>
6853         PR tree-optimization/87200
6854         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
6855         simplify result.
6857 2018-09-03  Martin Liska  <mliska@suse.cz>
6859         PR tree-optimization/87201
6860         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6861         Fix parenthesis in an expression.
6863 2018-09-03  Richard Biener  <rguenther@suse.de>
6865         PR tree-optimization/87197
6866         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
6867         visited.  CSE the VN_INFO hashtable lookup.
6869         PR tree-optimization/87169
6870         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
6871         iterating make sure there's no extra backedges from irreducible
6872         regions feeding the header.  Mark the destination block
6873         executable.
6875 2018-09-03  Martin Liska  <mliska@suse.cz>
6877         PR driver/83193
6878         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
6879         * common/common-targhooks.c (default_get_valid_option_values):
6880         New function.
6881         * common/common-targhooks.h (default_get_valid_option_values):
6882         Likewise.
6883         * common/config/i386/i386-common.c: Move processor_target_table
6884         from i386.c.
6885         (ix86_get_valid_option_values): New function.
6886         (TARGET_GET_VALID_OPTION_VALUES): New macro.
6887         * config/i386/i386.c (struct ptt): Move to i386-common.c.
6888         (PTA_*): Move all defined masks into i386-common.c.
6889         (ix86_function_specific_restore): Use new processor_cost_table.
6890         * config/i386/i386.h (struct ptt): Moved from i386.c.
6891         (struct pta): Likewise.
6892         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
6893         * doc/tm.texi.in: Likewise.
6894         * opt-suggestions.c (option_proposer::suggest_option):
6895         Pass prefix to build_option_suggestions.
6896         (option_proposer::get_completions): Likewise.
6897         (option_proposer::build_option_suggestions): Use the new target
6898         hook.
6899         * opts.c (struct option_help_tuple): New struct.
6900         (print_filtered_help): Use the new target hook.
6902 2018-09-03  Martin Liska  <mliska@suse.cz>
6904         PR middle-end/59521
6905         * predict.c (set_even_probabilities): Add likely_edges
6906         argument and handle cases where we have precisely one
6907         likely edge.
6908         (combine_predictions_for_bb): Catch also likely_edges.
6909         (tree_predict_by_opcode): Handle gswitch statements.
6910         * tree-cfg.h (find_case_label_for_value): New declaration.
6911         (find_taken_edge_switch_expr): Likewise.
6912         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6913         Find pivot in decision tree based on probabily, not by number of
6914         nodes.
6916 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
6918         * doc/standards.texi (Standards): Update Objective-C reference.
6920 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6922         * doc/install.texi (Prerequisites): Update link for MPC.
6924 2018-09-01  Michael Matz  <matz@suse.de>
6926         PR tree-optimization/87074
6927         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
6928         PHIs for outer-loop uses.
6930 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6932         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
6933         * doc/invoke.texi (C Dialect Options): Ditto.
6935 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6937         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
6939 2018-08-31  Richard Biener  <rguenther@suse.de>
6941         PR tree-optimization/87168
6942         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
6943         (rpo_elim::eliminate_avail): When OP was not visited it must
6944         be available.
6946 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
6948         * tree-vrp.c (copy_value_range): Convert param "from" from
6949         "value_range *" to "const value_range *".
6950         (range_is_null): Likewise for param "vr".
6951         (range_int_cst_p): Likewise.
6952         (range_int_cst_singleton_p): Likewise.
6953         (symbolic_range_p): Likewise.
6954         (value_ranges_intersect_p): Likewise for both params.
6955         (value_range_nonnegative_p): Likewise for param "vr".
6956         (value_range_constant_singleton): Likewise.
6957         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
6958         (extract_range_into_wide_ints): Likewise for param "vr".
6959         (extract_range_from_multiplicative_op): Likewise for params "vr0"
6960         and "vr1".
6961         (vrp_can_optimize_bit_op): Likewise.
6962         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
6963         "vr1_".
6964         (extract_range_from_unary_expr): Likewise.
6965         (debug_value_range): Likewise for param "vr".
6966         (value_range::dump): Add "const" qualifier.
6967         (vrp_prop::check_array_ref): Convert local "vr" from
6968         "value_range *" to "const value_range *".
6969         (vrp_prop::check_mem_ref): Likewise.
6970         (vrp_prop::visit_stmt): Likewise for local "old_vr".
6971         (vrp_intersect_ranges_1): Likewise for param "vr_1".
6972         (vrp_intersect_ranges): Likewise.
6973         (simplify_stmt_for_jump_threading): Likewise for local "vr".
6974         (vrp_prop::vrp_finalize): Likewise.
6975         * tree-vrp.h (value_range::dump): Add "const" qualifier.
6976         (vrp_intersect_ranges): Add "const" qualifier to params as above.
6977         (extract_range_from_unary_expr): Likewise.
6978         (value_range_constant_singleton): Likewise.
6979         (symbolic_range_p): Likewise.
6980         (copy_value_range): Likewise.
6981         (extract_range_from_binary_expr_1): Likewise.
6982         (range_int_cst_p): Likewise.
6983         (vrp_set_zero_nonzero_bits): Likewise.
6984         (range_int_cst_singleton_p): Likewise.
6986 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
6988         * config/aarch64/arm_neon.h (vabsd_s64): New.
6989         (vnegd_s64): Likewise.
6991 2018-08-31  Martin Jambor  <mjambor@suse.cz>
6993         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
6995 2018-08-31  Martin Liska  <mliska@suse.cz>
6997         * ipa-icf.c (sem_item::add_type): Use
6998         sem_item::m_type_hash_cache.
6999         * ipa-icf.h: Move the cache from sem_item_optimizer
7000         to sem_item.
7002 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7004         * doc/extend.texi (Backwards Compatibility): Remove implicit
7005         extern C leeway of () being (...).
7007 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7009         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7011 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7013         PR target/86684
7014         PR target/87149
7015         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7017 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7019         PR middle-end/87138
7020         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7021         gen_int_mode.  Formatting fixes.
7023 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7025         * target.def (custom_function_descriptors): Improve documentation.
7026         * doc/tm.texi.in (Trampolines): Expand discussion of function
7027         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7028         beginning of the section.
7029         * doc/tm.texi: Regenerated.
7031 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7033         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7034         base class in the initializer list.  This is a workaround for
7035         building with older GCC.
7036         (class auto_bb_flag): Likewise.
7038 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7040         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7041         (altivec_vcmpequ<VI_char>_p): Remove star.
7042         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7043         vector load modes.
7044         (expand_strncmp_vec_sequence): New function.
7045         (emit_final_str_compare_vec): New function.
7046         (expand_strn_compare): Add support for vector strncmp.
7047         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7048         length specification to bytes.
7049         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7050         (vcmpnezb_p): New pattern.
7051         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7052         for option -mstring-compare-inline-limit.
7054 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7056         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7057         (PTA_SKYLAKE): Add PTA_AES.
7058         (PTA_GOLDMONT): Likewise.
7060 2018-08-29  Jan Hubicka  <jh@suse.cz>
7062         PR lto/86517
7063         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7064         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7066 2018-08-29  Jan Hubicka  <jh@suse.cz>
7068         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7069         TYPE_STUB_DECL.
7070         (hash_tree): Do not visit TYPE_STUB_DECL.
7071         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7072         stream TYPE_STUB_DECL.
7073         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7074         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7075         after free_lang_data.
7076         (type_in_anonymous_namespace_p): Likewise.
7078 2018-08-29  Jan Hubicka  <jh@suse.cz>
7080         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7081         comment that it has to be even number.
7082         (class sreal): Change m_sig type to int32_t.
7083         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7084         int64_t for temporary calculations.
7085         (sreal_verify_basics): Drop one bit from minimum and maximum.
7087 2018-08-30  Richard Biener  <rguenther@suse.de>
7089         PR tree-optimization/87147
7090         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7091         (visit_phi): When the degenerate result is from the backedge and
7092         we didn't visit its definition yet drop to VARYING.
7093         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7095 2018-08-29  Jan Hubicka  <jh@suse.cz>
7097         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7098         DECL_VINDEX.
7099         (hash_tree): Likewise.
7101 2018-08-29  Jan Hubicka  <jh@suse.cz>
7103         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7104         and TYPE_NEXT_REF_TO.
7106 2018-08-29  Jan Hubicka  <jh@suse.cz>
7108         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7109         comment that it has to be even number.
7110         (class sreal): Change m_sig type to int32_t.
7111         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7112         int64_t for temporary calculations.
7113         (sreal_verify_basics): Drop one bit from minimum and maximum.
7115 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7117         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7119 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7121         PR middle-end/86995
7122         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7123         if to_add is negative.
7125 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7127         PR middle-end/87053
7128         * builtins.c (c_strlen): Improve range checks.
7130 2018-08-29  Martin Sebor  <msebor@redhat.com>
7131             Jeff Law  <law@redhat.com>
7133         PR tree-optimization/86714
7134         PR tree-optimization/86711
7135         * builtins.c (c_strlen): Add arguments to call to string_constant.
7136         * expr.c (string_constant): Add argument.  Detect missing nul
7137         terminator and outermost declaration it's missing in.
7138         * expr.h (string_constant): Add argument.
7139         * fold-const.c (read_from_constant_string): Add arguments to call to
7140         string_constant.
7141         (c_getstr): Likewise.
7142         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7143         to string_constant.
7144         * tree-ssa-strlen.c (get_stridx): Likewise.
7146 2018-08-29  Jan Hubicka  <jh@suse.cz>
7148         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7149         Do not stream DECL_VINDEX.
7150         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7151         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7152         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7154 2018-08-29  Richard Biener  <rguenther@suse.de>
7156         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7157         virtual operands that are not default defs to honor region
7158         boundaries.
7159         (rpo_vn_valueize): Remove ineffective code here.
7161 2018-08-29  Richard Biener  <rguenther@suse.de>
7163         PR tree-optimization/87132
7164         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7165         when skipping defs reachable over backedges.
7167 2018-08-29  Richard Biener  <rguenther@suse.de>
7169         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7170         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7171         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7172         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7173         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7174         (vn_reference_lookup_3): Remove use of const_parms.
7175         (free_rpo_vn): Do not free const_parms.
7176         (do_rpo_vn): Do not call init_const_parms.
7177         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7178         SSA_NAME_POINTS_TO_READONLY_MEMORY.
7179         (call_may_clobber_ref_p_1): Likewise.
7181 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
7183         PR other/86726
7184         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7185         (-O): Ditto.
7186         (-ftree-scev-cprop): Document.
7188 2018-08-29  Jan Hubicka  <jh@suse.cz>
7190         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7191         parameters.
7192         (sreal constructor): Update.
7193         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7194         sreal:operator/): Update.
7196 2018-08-29  Martin Liska  <mliska@suse.cz>
7198         * tree-switch-conversion.c (switch_conversion::expand):
7199         Strenghten assumption about gswitch statements.
7201 2018-08-29  Richard Biener  <rguenther@suse.de>
7203         PR tree-optimization/87117
7204         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7205         re-value-number released SSA VDEFs.
7207 2018-08-29  Richard Biener  <rguenther@suse.de>
7209         PR tree-optimization/87126
7210         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7212 2018-08-28  Jim Wilson  <jimw@sifive.com>
7214         * config/riscv/pic.md: Rewrite.
7215         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7216         invalid address.
7217         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7218         (SOFTF, default_load, softload, softstore): New.
7220 2018-08-28  Jeff Law  <law@redhat.com>
7222         * fold-const.c (fold_binary_loc): Remove recently added assert.
7224 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7226         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7227         to OP parmeter of generated function.
7229 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7231         PR tree-optimization/87009
7232         * match.pd: Add boolean optimizations.
7234 2018-08-28  Martin Sebor  <msebor@redhat.com>
7236         PR middle-end/86631
7237         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7238         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7239         (pass_walloca::gate): Use it.
7240         (alloca_call_type): Same.
7241         (pass_walloca::execute): Same.
7242         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7244 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7246         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7247         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7248         >= 3005 to >= 9000.
7250 2018-08-28  Richard Biener  <rguenther@suse.de>
7252         PR tree-optimization/87124
7253         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7254         constants before looking up avail.
7256 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7258         PR middle-end/87099
7259         * calls.c (maybe_warn_nonstring_arg): Punt early if
7260         warn_stringop_overflow is zero.  Don't call get_range_strlen
7261         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7262         Swap comparison operands to have constants on rhs.  Only use
7263         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
7264         increment lenrng[0].
7266 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
7268         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7269         use of tree_to_shwi.  Remove duplicated test for the size being
7270         a whole number of bytes.
7272 2018-08-28  Richard Biener  <rguenther@suse.de>
7274         PR tree-optimization/87117
7275         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7276         Handle removed stmt without LHS (GIMPLE_NOP).
7278 2018-08-28  Richard Biener  <rguenther@suse.de>
7280         PR tree-optimization/87117
7281         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7282         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7284 2018-08-28  Richard Biener  <rguenther@suse.de>
7286         PR tree-optimization/87117
7287         * tree-ssa-pre.c (compute_avail): Do not make expressions
7288         with predicated values available.
7289         (get_expr_value_id): Assert we do not run into predicated value
7290         expressions.
7292 2018-08-28  Richard Biener  <rguenther@suse.de>
7294         PR tree-optimization/87117
7295         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7296         get virtual operands.
7297         (get_expr_operands): Handle STRING_CST like other decls.
7299 2018-08-28  Martin Liska  <mliska@suse.cz>
7301         * tree.h: Update documentation of fndecl_built_in_p
7302         functions.
7305 2018-08-27  Jeff Law  <law@redhat.com>
7306         PR tree-optimization/87110
7307         * tree-ssa-dse.c (compute_trims): Handle non-constant
7308         TYPE_SIZE_UNIT.
7310 2018-08-27  Martin Sebor  <msebor@redhat.com>
7312         PR tree-optimization/86914
7313         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7315 2018-08-27  Martin Sebor  <msebor@redhat.com>
7317         PR tree-optimization/87112
7318         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7319         the type of the bound argument.
7321 2018-08-27  Jeff Law  <law@redhat.com>
7323         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7324         type does not have a TYPE_SIZE_UNIT.
7326 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
7328         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7329         with include of backend.h.
7331 2018-08-27  Richard Biener  <rguenther@suse.de>
7333         PR tree-optimization/86927
7334         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7335         use const cond reduction code.
7337 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
7339         PR tree-optimization/85758
7340         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7342 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7344         PR c++/87091
7345         * diagnostic-show-locus.c (class layout_range): Update for
7346         conversion of show_caret_p to a tri-state.
7347         (layout_range::layout_range): Likewise.
7348         (make_range): Likewise.
7349         (layout::maybe_add_location_range): Likewise.
7350         (layout::should_print_annotation_line_p): Don't show annotation
7351         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7352         (layout::get_state_at_point): Update for conversion of
7353         show_caret_p to a tri-state.  Bail out early for
7354         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7355         underlining or source colorization.
7356         (gcc_rich_location::add_location_if_nearby): Update for conversion
7357         of show_caret_p to a tri-state.
7358         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7359         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7360         Likewise.
7361         (selftest::test_one_liner_labels): Likewise.
7362         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7363         conversion of show_caret_p to a tri-state.
7364         * pretty-print.c (text_info::set_location): Likewise.
7365         * pretty-print.h (text_info::set_location): Likewise.
7366         * substring-locations.c (format_warning_n_va): Likewise.
7367         * tree-diagnostic.c (default_tree_printer): Likewise.
7368         * tree-pretty-print.c (newline_and_indent): Likewise.
7370 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7372         PR c++/87091
7373         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7374         line above for line-insertion fix-it hints.
7375         (selftest::test_fixit_insert_containing_newline): Update the
7376         expected results, and add a test with line-numbering enabled.
7378 2018-08-27  Martin Liska  <mliska@suse.cz>
7380         PR sanitizer/86962
7381         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7382         params with DECL_HAS_VALUE_EXPR_P.
7384 2018-08-27  Martin Liska  <mliska@suse.cz>
7386         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7387         selected expansion strategy.
7389 2018-08-27  Martin Liska  <mliska@suse.cz>
7391         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7392         * builtins.c (is_builtin_fn): Likewise.
7393         * attribs.c (diag_attr_exclusions): Use new function
7394         fndecl_built_in_p and remove check for FUNCTION_DECL if
7395         possible.
7396         (builtin_mathfn_code): Likewise.
7397         (fold_builtin_expect): Likewise.
7398         (fold_call_expr): Likewise.
7399         (fold_builtin_call_array): Likewise.
7400         (fold_call_stmt): Likewise.
7401         (set_builtin_user_assembler_name): Likewise.
7402         (is_simple_builtin): Likewise.
7403         * calls.c (gimple_alloca_call_p): Likewise.
7404         (maybe_warn_nonstring_arg): Likewise.
7405         * cfgexpand.c (expand_call_stmt): Likewise.
7406         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7407         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7408         (cgraph_node::verify_node): Likewise.
7409         * cgraphclones.c (build_function_decl_skip_args): Likewise.
7410         (cgraph_node::create_clone): Likewise.
7411         * config/arm/arm.c (arm_insert_attributes): Likewise.
7412         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7413         * dse.c (scan_insn): Likewise.
7414         * expr.c (expand_expr_real_1): Likewise.
7415         * fold-const.c (operand_equal_p): Likewise.
7416         (fold_binary_loc): Likewise.
7417         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7418         * gimple-low.c (lower_stmt): Likewise.
7419         * gimple-pretty-print.c (dump_gimple_call): Likewise.
7420         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7421         Likewise.
7422         * gimple.c (gimple_build_call_from_tree): Likewise.
7423         (gimple_call_builtin_p): Likewise.
7424         (gimple_call_combined_fn): Likewise.
7425         * gimplify.c (gimplify_call_expr): Likewise.
7426         (gimple_boolify): Likewise.
7427         (gimplify_modify_expr): Likewise.
7428         (gimplify_addr_expr): Likewise.
7429         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7430         * ipa-cp.c (determine_versionability): Likewise.
7431         * ipa-fnsummary.c (compute_fn_summary): Likewise.
7432         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7433         * ipa-split.c (visit_bb): Likewise.
7434         (split_function): Likewise.
7435         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7436         * lto-cgraph.c (input_node): Likewise.
7437         * lto-streamer-out.c (write_symbol): Likewise.
7438         * omp-low.c (setjmp_or_longjmp_p): Likewise.
7439         (lower_omp_1): Likewise.
7440         * predict.c (strip_predict_hints): Likewise.
7441         * print-tree.c (print_node): Likewise.
7442         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7443         * trans-mem.c (is_tm_irrevocable): Likewise.
7444         (is_tm_load): Likewise.
7445         (is_tm_simple_load): Likewise.
7446         (is_tm_store): Likewise.
7447         (is_tm_simple_store): Likewise.
7448         (is_tm_abort): Likewise.
7449         (tm_region_init_1): Likewise.
7450         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7451         * tree-cfg.c (verify_gimple_call): Likewise.
7452         (move_stmt_r): Likewise.
7453         (stmt_can_terminate_bb_p): Likewise.
7454         * tree-eh.c (lower_eh_constructs_2): Likewise.
7455         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7456         * tree-inline.c (remap_gimple_stmt): Likewise.
7457         (copy_bb): Likewise.
7458         (estimate_num_insns): Likewise.
7459         (fold_marked_statements): Likewise.
7460         * tree-sra.c (scan_function): Likewise.
7461         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7462         (optimize_stack_restore): Likewise.
7463         (pass_fold_builtins::execute): Likewise.
7464         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7465         (mark_all_reaching_defs_necessary_1): Likewise.
7466         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7467         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7468         (pass_forwprop::execute): Likewise.
7469         * tree-ssa-loop-im.c (stmt_cost): Likewise.
7470         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7471         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7472         * tree-ssa-strlen.c (get_string_length): Likewise.
7473         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7474         (find_func_aliases_for_call): Likewise.
7475         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7476         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7477         * tree-tailcall.c (find_tail_calls): Likewise.
7478         * tree.c (need_assembler_name_p): Likewise.
7479         (free_lang_data_in_decl): Likewise.
7480         (get_call_combined_fn): Likewise.
7481         * ubsan.c (is_ubsan_builtin_p): Likewise.
7482         * varasm.c (incorporeal_function_p): Likewise.
7483         * tree.h (DECL_BUILT_IN): Remove and replace with
7484         fndecl_built_in_p.
7485         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7486         (fndecl_built_in_p): New.
7488 2018-08-27  Martin Liska  <mliska@suse.cz>
7490         PR tree-optimization/86847
7491         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7492         Dump also subtree probability.
7493         (switch_decision_tree::do_jump_if_equal): New function.
7494         (switch_decision_tree::emit_case_nodes): Handle special
7495         situations in balanced tree that can be emitted much simpler.
7496         Fix calculation of probabilities that happen in tree expansion.
7497         * tree-switch-conversion.h (struct cluster): Add
7498         is_single_value_p.
7499         (struct simple_cluster): Likewise.
7500         (struct case_tree_node): Add new function has_child.
7501         (do_jump_if_equal): New.
7503 2018-08-27  Martin Liska  <mliska@suse.cz>
7505         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7506         Add new argument to bit_test_cluster constructor.
7507         (bit_test_cluster::emit): Set bits really number of values
7508         handlel by a test.
7509         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7510         probability argument.
7511         * tree-switch-conversion.h (struct bit_test_cluster):
7512         Add m_handles_entire_switch.
7514 2018-08-27  Martin Liska  <mliska@suse.cz>
7516         PR tree-optimization/86702
7517         * tree-switch-conversion.c (jump_table_cluster::emit):
7518         Make probabilities even for values in jump table
7519         according to number of cases handled.
7520         (switch_decision_tree::compute_cases_per_edge): Pass
7521         argument to reset_out_edges_aux function.
7522         (switch_decision_tree::analyze_switch_statement): Likewise.
7523         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7524         Make it static.
7526 2018-08-27  Martin Liska  <mliska@suse.cz>
7528         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
7529         cfun argument explicitly.
7530         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
7531         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
7532         function gimple_switch_default_bb.
7533         (convert_switch_statements):
7534         (expand_builtins):
7535         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
7536         * stmt.c (label_to_block_fn): Use label_to_block and pass
7537         cfun argument explicitly and use gimple_switch_label_bb.
7538         (expand_case): Likewise.
7539         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
7540         cfun argument explicitly. Likewise.
7541         (make_edges_bb): Likewise.
7542         (make_cond_expr_edges): Likewise.
7543         (get_cases_for_edge): Likewise.
7544         (make_gimple_switch_edges): Likewise.
7545         (label_to_block_fn): Likewise.
7546         (label_to_block): Likewise.
7547         (make_goto_expr_edges): Likewise.
7548         (make_gimple_asm_edges): Likewise.
7549         (main_block_label): Likewise.
7550         (group_case_labels_stmt): Likewise.
7551         (find_taken_edge_computed_goto): Likewise.
7552         (find_taken_edge_switch_expr): Likewise.
7553         (gimple_verify_flow_info): Likewise.
7554         (gimple_redirect_edge_and_branch): Likewise.
7555         (gimple_switch_label_bb): New function.
7556         (gimple_switch_default_bb): Likewise.
7557         (gimple_switch_edge): Likewise.
7558         (gimple_switch_default_edge): Likewise.
7559         * tree-cfg.h (label_to_block_fn): Remove and replace ...
7560         (label_to_block): ... with this.
7561         (gimple_switch_label_bb): New.
7562         (gimple_switch_default_bb): Likewise.
7563         (gimple_switch_edge): Likewise.
7564         (gimple_switch_default_edge): Likewise.
7565         * tree-cfgcleanup.c (convert_single_case_switch): Use
7566         new gimple functions and pass new argument to label_to_block.
7567         (cleanup_control_flow_bb):
7568         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
7569         cfun argument explicitly.
7570         (make_eh_edges): Likewise.
7571         (redirect_eh_dispatch_edge): Likewise.
7572         (lower_resx): Likewise.
7573         (lower_eh_dispatch): Likewise.
7574         (maybe_remove_unreachable_handlers): Likewise.
7575         (unsplit_eh): Likewise.
7576         (cleanup_empty_eh): Likewise.
7577         (verify_eh_edges): Likewise.
7578         (verify_eh_dispatch_edge): Likewise.
7579         * tree-ssa-dom.c (record_edge_info): Likewise.
7580         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
7581         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
7582         (thread_through_normal_block): Likewise.
7583         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
7584         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
7585         * tree-switch-conversion.c (switch_conversion::collect): Use new
7586         gimple functions.
7587         (switch_conversion::check_final_bb): Likewise.
7588         (switch_conversion::gather_default_values): Pass new argument
7589         to label_to_block.
7590         (switch_conversion::build_constructors): Likewise.
7591         (switch_decision_tree::compute_cases_per_edge): Use new
7592         gimple_switch_edge function.
7593         (switch_decision_tree::analyze_switch_statement): Pass new argument
7594         to label_to_block.
7595         (switch_decision_tree::try_switch_expansion): Use
7596         gimple_switch_default_edge.
7597         * tree-vrp.c (find_switch_asserts): Pass new argument
7598         to label_to_block.
7599         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
7600         (vr_values::simplify_switch_using_ranges): Likewise.
7602 2018-08-27  Richard Biener  <rguenther@suse.de>
7604         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
7605         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
7607         * tree-ssa-sccvn.h (struct vn_pval): New structure.
7608         (struct vn_nary_op_s): Add unwind_to member.  Add
7609         predicated_values flag and put result into a union together
7610         with a linked list of vn_pval.
7611         (struct vn_ssa_aux): Add name member to make maintaining
7612         a map of SSA name to vn_ssa_aux possible.  Remove no longer
7613         needed info, dfsnum, low, visited, on_sccstack, use_processed
7614         and range_info_anti_range_p members.
7615         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
7616         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
7617         New functions.
7618         (vn_valueize): New global.
7619         (vn_context_bb): Likewise.
7620         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
7621         VN_INFO_PTR_INFO): Remove.
7622         * tree-ssa-sccvn.c: ... (rewrite)
7623         (pass_fre::execute): For -O2+ initialize loops and run
7624         RPO VN in optimistic mode (iterating).  For -O1 and -Og
7625         run RPO VN in non-optimistic mode.
7626         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
7627         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
7628         * doc/invoke.texi (sccvn-max-scc-size): Remove.
7629         (rpo-vn-max-loop-depth): Document.
7630         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
7631         when valuezing the VUSE signals we walked out of the region.
7632         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
7633         (phi_translate): Set VN context block to use for availability
7634         lookup.
7635         (compute_avail): Likewise.
7636         (pre_valueize): New function.
7637         (pass_pre::execute): Adjust to the RPO VN API.
7639         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
7640         (propagate_constants_for_unrolling): Remove.
7641         (tree_unroll_loops_completely): Perform value-numbering
7642         on the unrolled bodies loop parent.
7644 2018-08-27  Richard Biener  <rguenther@suse.de>
7646         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
7647         for partial antic compute.
7649 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
7651         PR rtl-optimization/87065
7652         * combine.c (simplify_if_then_else): Formatting fix.
7653         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
7654         check.
7655         (known_cond): Don't return const_true_rtx for vector modes.  Use
7656         CONST0_RTX instead of const0_rtx.  Formatting fixes.
7658 2018-08-27  Martin Liska  <mliska@suse.cz>
7660         PR gcov-profile/87069
7661         * gcov.c (process_file): Record files already processed
7662         and warn about a file being processed multiple times.
7664 2018-08-27  Martin Liska  <mliska@suse.cz>
7666         PR driver/83193
7667         * config/aarch64/aarch64.c (aarch64_override_options_internal):
7668         Set default values for x_aarch64_*_string strings.
7669         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
7670         prefix.  For -mabi do not print '=ABI' in help and use
7671         <option_value> format for -msve-vector-bits and -moverride
7672         options.
7674 2018-08-26  Jeff Law  <law@redhat.com>
7676         * config/mips/frame-header-opt.c: Include "backend.h" rather than
7677         "cfg.h"
7679 2018-08-26  Marek Polacek  <polacek@redhat.com>
7681         PR c++/87029, Implement -Wredundant-move.
7682         * doc/invoke.texi: Document -Wredundant-move.
7684 2018-08-25  Martin Sebor  <msebor@redhat.com>
7686         PR tree-optimization/87059
7687         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
7688         to the same type as the other.
7689         * fold-const.c (fold_binary_loc): Assert expectation.
7691 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
7693         * config/darwin.c (machopic_legitimize_pic_address): Clean up
7694         extraneous parentheses, dead code section and formatting.
7696 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
7698         PR c++/87091
7699         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
7700         wide enough for jumps in the line-numbering to be visible.
7701         (layout::print_gap_in_line_numbering): New member function.
7702         (layout::calculate_line_spans): When using line numbering, merge
7703         line spans that are only 1 line apart.
7704         (diagnostic_show_locus): When printing line numbers, show gaps in
7705         line numbering directly, rather than printing headers.
7706         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
7707         line-numbering with multiple line spans.
7708         (selftest::test_fixit_insert_containing_newline_2): Add test of
7709         line-numbering, in which the spans are close enough to be merged.
7711 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7713         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
7714         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
7715         * tree-vrp.c (range_is_nonnull): Remove.
7716         (range_includes_zero_p): Accept value_range instead of min/max.
7717         (extract_range_from_binary_expr_1): Do not early bail on
7718         POINTER_PLUS_EXPR.
7719         Use range_includes_zero_p instead of range_is_nonnull.
7720         (extract_range_from_unary_expr): Use range_includes_zero_p instead
7721         of range_is_nonnull.
7722         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
7723         special case VR_ANTI_RANGE.
7724         (vrp_finalize): Same.
7725         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
7726         instead of min/max.
7727         (range_is_nonnull): Remove.
7728         * vr-values.c (vrp_stmt_computes_nonzero): Use
7729         range_includes_zero_p instead of range_is_nonnull.
7730         (extract_range_basic): Pass value_range to range_includes_zero_p
7731         instead of range_is_nonnull.
7733 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
7735         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
7736         * emit-rtl.h (rtl_data): Remove return_bnd.
7737         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
7738         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
7739         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
7740         (POINTER_BOUNDS_MODE): Remove definition.
7741         (make_pointer_bounds_mode): Remove.
7742         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
7743         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
7744         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
7745         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
7746         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
7747         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
7748         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
7750         * config/i386/i386-modes.def (BND32, BND64): Remove.
7751         * config/i386/i386.c (dbx_register_map): Remove bound registers.
7752         (dbx64_register_map): Ditto.
7753         (svr4_dbx_register_map): Ditto.
7754         (indirect_thunk_bnd_needed): Remove.
7755         (indirect_thunks_bnd_used): Ditto.
7756         (indirect_return_bnd_needed): Ditto.
7757         (indirect_return_via_cx_bnd): Ditto.
7758         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
7759         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
7760         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
7761         (output_indirect_thunk_function): Remove handling of
7762         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
7763         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
7764         (ix86_save_reg): Remove handling of crtl->return_bnd.
7765         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
7766         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
7767         and UNSPEC_BNDLX_ADDR.
7768         (ix86_output_indirect_branch_via_reg): Remove handling of
7769         indirect_thunk_prefix_bnd.
7770         (ix86_output_indirect_branch_via_push): Ditto.
7771         (ix86_output_function_return): Ditto.
7772         (ix86_output_indirect_function_return): Ditto.
7773         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
7774         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
7775         (CALL_USED_REGISTERS): Ditto.
7776         (REG_ALLOC_ORDER): Update for removal of bound registers.
7777         (HI_REGISTER_NAMES): Ditto.
7778         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
7779         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
7780         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
7781         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
7782         (FIRST_PSEUDO_REG): Update.
7783         (BND): Remove mode iterator.
7784         * config/i386/predicates.md (bnd_mem_operator): Remove.
7786 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7788         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
7789         vectors.
7791 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7793         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
7794         the case in which the permute needs only a single element and
7795         repeats for every vector of the result.  Extend that case to
7796         handle variable-length vectors.
7797         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7799 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
7801         PR debug/79342
7802         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
7803         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
7805 2018-08-24  Richard Biener  <rguenther@suse.de>
7807         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
7808         bb_flags_allocated members.
7809         (auto_flag): New RAII class for allocating flags.
7810         (auto_edge_flag): New RAII class for allocating edge flags.
7811         (auto_bb_flag): New RAII class for allocating bb flags.
7812         * cfgloop.c (verify_loop_structure): Allocate temporary edge
7813         flag dynamically.
7814         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
7815         in favor of temporarily allocated BB flag.
7816         * hsa-brig.c: Re-order includes.
7817         * hsa-dump.c: Likewise.
7818         * hsa-regalloc.c: Likewise.
7819         * print-rtl.c: Likewise.
7820         * profile-count.c: Likewise.
7822 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
7824         PR target/86989
7825         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
7826         the TOC register.
7828 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7830         PR 87073/bootstrap
7831         * wide-int-range.cc (wide_int_range_div): Do not ignore result
7832         from wide_int_range_multiplicative_op.
7834 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7836         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
7837         "permutaion".
7839 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
7841         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
7842         to 'expanded'.
7844 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
7846         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
7847         full GENERIC expression used for replacement.
7849 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
7851         * tree-vrp.c (abs_extent_range): Remove.
7852         (extract_range_into_wide_ints): Pass wide ints by reference.
7853         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
7854         Pass wide ints by reference in all calls to
7855         extract_range_into_wide_ints.
7856         * wide-int-range.cc (wide_int_range_div): New.
7857         * wide-int-range.h (wide_int_range_div): New.
7858         (wide_int_range_includes_zero_p): New.
7859         (wide_int_range_zero_p): New.
7861 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
7863         * config/aarch64/aarch64.md (arches): New enum.
7864         (arch): New enum attr.
7865         (arch_enabled): New attr.
7866         (enabled): Now uses arch_enabled only.
7867         (simd, sve, fp16): Removed attribute.
7868         (fp): Attr now defined in terms of 'arch'.
7869         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
7870         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
7871         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
7872         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
7873         attributes into 'arch'.
7874         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
7875         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
7876         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
7877         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
7878         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
7879         'simd' attribute into 'arch'.
7880         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
7881         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
7882         Convert use of 'fp' attribute to 'arch'.
7883         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
7884         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
7885         into 'arch'.
7886         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
7887         (different modes) Merge 'fp' and 'simd' into 'arch'.
7888         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
7889         'simd' into 'arch'.
7891 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
7893         PR rtl-optimization/87026
7894         * expmed.c (canonicalize_comparison): If we can no longer create
7895         pseudoregisters, don't.
7897 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
7899         PR target/86951
7900         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
7901         prototype.
7902         * config/arm/arm.c (speculation_barrier_libfunc): New static
7903         variable.
7904         (arm_init_libfuncs): Initialize it.
7905         (arm_emit_speculation_barrier): New function.
7906         * config/arm/arm.md (speculation_barrier): Call
7907         arm_emit_speculation_barrier for architectures that do not have
7908         DSB or ISB.
7909         (speculation_barrier_insn): Only match on Armv7 or later.
7911 2018-08-23  Richard Biener  <rguenther@suse.de>
7913         PR middle-end/87024
7914         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
7915         calls.
7917 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7919         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
7920         of single-vector TBLs.
7921         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
7922         one input is given.
7924 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7926         PR target/85910
7927         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
7928         aarch64_evpc_tbl guard.
7930 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7932         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
7933         behaviour.
7935 2018-08-22  Martin Sebor  <msebor@redhat.com>
7937         PR middle-end/87052
7938         * tree-pretty-print.c (pretty_print_string): Add argument.
7939         (dump_generic_node): Call to pretty_print_string with string size.
7941 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
7943         PR rtl-optimization/86771
7944         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
7945         of two SETs into those two SETs, one to be placed at i2, if that SETs
7946         destination is modified between i2 and i3.
7948 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7950         PR tree-optimization/86725
7951         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
7952         function.
7953         (vect_analyze_scalar_cycles_1): Check it.
7955 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7957         PR tree-optimization/86725
7958         * tree-vect-loop.c (vect_is_simple_reduction): When treating
7959         an outer loop phi as a double reduction, make sure that the
7960         single user of the phi result is an inner loop phi.
7962 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7964         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
7965         grouped stores with gaps to a strided group.
7967 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7969         * tree-vect-stmts.c (get_group_load_store_type)
7970         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
7971         first statement in a group.
7973 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7975         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
7976         the sequence used in gcc/gcc.c.
7978 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7980         PR other/704
7981         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
7982         building it.
7984 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7986         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
7987         Darwin10-specific unwinder-shim.
7988         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
7989         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
7990         New to cater for Darwin10 Rosetta.
7992 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7994         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
7995         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
7997 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7999         PR bootstrap/81033
8000         PR target/81733
8001         PR target/52795
8002         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8003         (dwarf2out_switch_text_section): Generate a local label for the
8004         second function sub-section and apply it as the second FDE start
8005         label.
8006         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8007         second sub-section start.
8009 2018-08-22  Richard Biener  <rguenther@suse.de>
8011         PR tree-optimization/86988
8012         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8014 2018-08-22  Richard Biener  <rguenther@suse.de>
8016         PR tree-optimization/86945
8017         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8019 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8021         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8022         a comment about how uses of r2 for .sdata2 come about.
8024 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8026         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8028 2018-08-21  Marek Polacek  <polacek@redhat.com>
8030         PR c++/86981, Implement -Wpessimizing-move.
8031         * doc/invoke.texi: Document -Wpessimizing-move.
8033 2018-08-21  Jan Hubicka  <jh@suse.cz>
8035         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8036         * tree.h (is_redundant_typedef): Remove.
8037         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8039 2018-08-21  Jan Hubicka  <jh@suse.cz>
8041         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8042         when possible.
8044 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8046         * expmed.c (extract_low_bits): Reject invalid subregs early.
8048 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8050         PR middle-end/86121
8051         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8052         behaviour.
8054 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8056         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8057         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8058         * config/vxworks.c: Likewise.
8060 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8062         * config/vxworks.c: Set targetm.have_ctors_dtors
8063         if HAVE_INITFINI_ARRAY_SUPPORT.
8064         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8065         if HAVE_INITFINI_ARRAY_SUPPORT.
8067 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8069         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8070         default search path for VxWorks < 7.
8072 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8074         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8075         (get_format_string): Refer to c_getstr.
8077 2018-08-21  Tom de Vries  <tdevries@suse.de>
8079         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8080         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8081         (debuginfo_early_stop): Declare.
8082         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8083         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8084         (debuginfo_early_stop): New function.
8085         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8086         and debuginfo_early_stop.
8087         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8088         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8089         (general_init): Call debuginfo_early_init.
8090         (finalize): Call debuginfo_fini.
8091         (do_compile): Call debuginfo_init.
8092         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8093         -fdump-early-debug.
8094         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8096 2018-08-21  Tom de Vries  <tdevries@suse.de>
8098         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8099         flag_dump_noaddr and flag_dump_unnumbered.
8101 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8103         * wide-int-range.cc (wide_int_range_abs): New.
8104         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8105         * wide-int-range.h (wide_int_range_abs): New.
8106         (wide_int_range_min_max): New.
8107         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8108         case to call wide_int_range_abs.
8109         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8110         (extract_range_from_abs_expr): Delete.
8112 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8114         PR target/87033
8115         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8116         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8117         for indexed loads.
8119 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8120             Jeff Law <law@redhat.com>
8122         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8123         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8125 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8127         PR other/84889
8128         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8129         (decl_attributes): Likewise.
8130         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8131         instance.
8132         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8133         * diagnostic-core.h (class auto_diagnostic_group): New class.
8134         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8135         (diagnostic_report_diagnostic): Handle the first diagnostics within
8136         a group.
8137         (emit_diagnostic): Add auto_diagnostic_group instance.
8138         (inform): Likewise.
8139         (inform_n): Likewise.
8140         (warning): Likewise.
8141         (warning_at): Likewise.
8142         (warning_n): Likewise.
8143         (pedwarn): Likewise.
8144         (permerror): Likewise.
8145         (error): Likewise.
8146         (error_n): Likewise.
8147         (error_at): Likewise.
8148         (sorry): Likewise.
8149         (fatal_error): Likewise.
8150         (internal_error): Likewise.
8151         (internal_error_no_backtrace): Likewise.
8152         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8153         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8154         * diagnostic.h (struct diagnostic_context): Add fields
8155         "diagnostic_group_nesting_depth",
8156         "diagnostic_group_emission_count", "begin_group_cb",
8157         "end_group_cb".
8158         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8159         Add auto_diagnostic_group instance(s).
8160         (find_explicit_erroneous_behavior): Likewise.
8161         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8162         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8163         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8164         (gimplify_va_arg_expr): Likewise.
8165         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8166         (HSA_SORRY_AT): Likewise.
8167         * ipa-devirt.c (compare_virtual_tables): Likewise.
8168         (warn_odr): Likewise.
8169         * multiple_target.c (expand_target_clones): Likewise.
8170         * opts-common.c (cmdline_handle_error): Likewise.
8171         * reginfo.c (globalize_reg): Likewise.
8172         * substring-locations.c (format_warning_n_va): Likewise.
8173         * tree-inline.c (expand_call_inline): Likewise.
8174         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8175         * tree-ssa-loop-niter.c
8176         (do_warn_aggressive_loop_optimizations): Likewise.
8177         * tree-ssa-uninit.c (warn_uninit): Likewise.
8178         * tree.c (warn_deprecated_use): Likewise.
8180 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
8182         PR target/87014
8183         * config/i386/i386.md (eh_return): Always update EH return
8184         address in word_mode.
8186 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
8188         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8189         TARGET_SPLIT_COMPLEX_ARG is defined.
8191 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8193         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8195 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8197         PR target/86984
8198         * expr.c (expand_assignment): Assert that bitpos is positive.
8199         (store_field): Likewise
8200         (expand_expr_real_1): Make sure that bitpos is positive.
8201         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8202         integer overflow.
8204 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8206         * Makefile.in (CPP_ID_DATA_H): Delete.
8207         (CPP_INTERNAL_H): Don't add it.
8208         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8209         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8211 2018-08-20  Richard Biener  <rguenther@suse.de>
8213         PR tree-optimization/78655
8214         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8215         pointer + offset nonnull if either operand is nonnull work.
8217 2018-08-20  Tom de Vries  <tdevries@suse.de>
8219         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8220         unless the referenced die describes the added property using
8221         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8222         Otherwise, add a DW_AT_location to the referenced die.
8224 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8226         PR target/86994
8227         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8228         register_operand when calling ix86_set_reg_reg_cost.
8229         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8230         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8231         predicate and to 1 otherwise.
8233 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8235         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8236         emit a diagnostic that it is not supported and reset the option.
8237         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8238         supported and consume the option.  (ASM_FINAL_SPEC): New.
8240 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8242         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8243         a sentence.
8245 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8247         C-SKY port: Documentation
8249         * doc/extend.texi (C-SKY Function Attributes): New section.
8250         * doc/invoke.texi (Option Summary): Add C-SKY options.
8251         (C-SKY Options): New section.
8252         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8254 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8255             Huibin Wang  <huibin_wang@c-sky.com>
8256             Sandra Loosemore  <sandra@codesourcery.com>
8257             Chung-Lin Tang  <cltang@codesourcery.com>
8259         C-SKY port: Backend implementation
8261         * config/csky/*: New.
8262         * common/config/csky/*: New.
8264 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8265             Huibin Wang  <huibin_wang@c-sky.com>
8266             Sandra Loosemore  <sandra@codesourcery.com>
8267             Chung-Lin Tang  <cltang@codesourcery.com>
8268             Andrew Jenner  <andrew@codesourcery.com>
8270         C-SKY port: Configury
8272         * config.gcc (csky-*-*): New.
8273         * configure.ac: Add csky to targets for dwarf2 debug_line support.
8274         * configure: Regenerated.
8276 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
8278         * dump-context.h: Include "dumpfile.h".
8279         (dump_context::dump_printf_va): Convert final param from va_list
8280         to va_list *.  Convert from ATTRIBUTE_PRINTF to
8281         ATTRIBUTE_GCC_DUMP_PRINTF.
8282         (dump_context::dump_printf_loc_va): Likewise.
8283         * dumpfile.c: Include "stringpool.h".
8284         (make_item_for_dump_printf_va): Delete.
8285         (make_item_for_dump_printf): Delete.
8286         (class dump_pretty_printer): New class.
8287         (dump_pretty_printer::dump_pretty_printer): New ctor.
8288         (dump_pretty_printer::emit_items): New member function.
8289         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8290         function.
8291         (dump_pretty_printer::emit_item): New member function.
8292         (dump_pretty_printer::stash_item): New member function.
8293         (dump_pretty_printer::format_decoder_cb): New member function.
8294         (dump_pretty_printer::decode_format): New member function.
8295         (dump_context::dump_printf_va): Reimplement in terms of
8296         dump_pretty_printer.
8297         (dump_context::dump_printf_loc_va): Convert final param from va_list
8298         to va_list *.
8299         (dump_context::begin_scope): Reimplement call to
8300         make_item_for_dump_printf.
8301         (dump_printf): Update for change to dump_printf_va.
8302         (dump_printf_loc): Likewise.
8303         (selftest::test_capture_of_dump_calls): Convert "stmt" from
8304         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
8305         with %T, %E, and %G.
8306         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8307         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8308         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8309         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8310         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8311         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8312         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8313         within a dump_printf_loc call to "%wu".
8314         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8315         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
8316         missing space after "=".
8317         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8318         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8319         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8320         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8321         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
8322         duplicate "vectorized" from message.
8324 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
8326         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8327         polyNxK_t element's TYPE_STRING_FLAG.
8329 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8331         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8332         (they were unnamed before).  Fix comments.
8334 2018-08-17  Nathan Sidwell  <nathan@acm.org>
8336         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8338 2018-08-17  Richard Biener  <rguenther@suse.de>
8340         PR tree-optimization/86841
8341         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8343 2018-08-17  Martin Liska  <mliska@suse.cz>
8345         * common.opt: Remove Warn, Init and Report for options with
8346         Ignore/Deprecated flag. Warning is done automatically for
8347         Deprecated flags.
8348         * config/i386/i386.opt: Likewise.
8349         * config/ia64/ia64.opt: Likewise.
8350         * config/rs6000/rs6000.opt: Likewise.
8351         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8352         Remove usage of flag_check_pointer_bounds.
8353         * lto-wrapper.c (merge_and_complain): Do not handle
8354         OPT_fcheck_pointer_bounds.
8355         (append_compiler_options): Likewise.
8356         * opt-functions.awk: Do not handle Deprecated.
8357         * optc-gen.awk: Check that Var, Report and Init are not
8358         used for an option with Ignore/Deprecated flag.
8359         * opts-common.c (decode_cmdline_option): Do not report
8360         CL_ERR_DEPRECATED.
8361         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8362         options.
8363         * opts.h (struct cl_option): Remove cl_deprecated flag.
8364         (CL_ERR_DEPRECATED): Remove error enum value.
8366 2018-08-17  Richard Biener  <rguenther@suse.de>
8368         PR middle-end/86505
8369         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8370         across a va-arg-pack using call adjust its return value accordingly.
8372 2018-08-16  Martin Sebor  <msebor@redhat.com>
8374         PR tree-optimization/86853
8375         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8376         (struct fmtresult): Add member and initialize it in ctors.
8377         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8378         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8379         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8380         (parse_directive): Handle %C same as %c.
8381         (sprintf_dom_walker::compute_format_length): Adjust.
8382         (is_call_safe): Adjust.
8384 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8386         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
8387         for determining how to count the elements.
8388         * builtins.h (c_strlen): Adjust prototype.
8389         * expr.c (string_constant): Add new parameter mem_size.
8390         Set *mem_size appropriately.
8391         * expr.h (string_constant): Adjust protoype.
8392         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8393         * gimple-fold.h (get_range_strlen): Adjust prototype.
8394         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8395         (format_string): Call get_string_length with eltsize.
8397 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8399         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8400         to emit the span, rather than setting it as the prefix.
8402 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8404         * diagnostic-show-locus.c (layout::start_annotation_line): Add
8405         "margin_char" parameter, defaulting to space.  Use it in place
8406         of pp_space for the initial part of the margin.
8407         (layout::print_leading_fixits): Use '+' when filling the margin
8408         of line-insertion fix-it hints.
8410 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8412         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8413         Delete.
8415 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8417         * config/rs6000/altivec.md: Don't set length attribute to the default
8418         value.
8419         * config/rs6000/darwin.md: Ditto.
8420         * config/rs6000/dfp.md: Ditto.
8421         * config/rs6000/htm.md: Ditto.
8422         * config/rs6000/rs6000.md: Ditto.
8423         * config/rs6000/sync.md: Ditto.
8424         * config/rs6000/vsx.md: Ditto.
8426 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8428         * config/rs6000/altivec.md: Don't set length attribute to the default
8429         value, for branch instructions.
8430         * config/rs6000/darwin.md: Ditto.
8431         * config/rs6000/rs6000.md: Ditto.
8433 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8435         * config/rs6000/rs6000.md (length): Always define as const_int 4.
8436         (unnamed conditional branch define_insn): Set length to 4 or 8
8437         depending on offset.
8438         (<bd>_<mode>): Similar, for alternative 0.
8439         (<bd>tf_<mode>): Ditto.
8441 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8443         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8445 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
8447         * doc/rtl.texi: Replace old RTX class names with new names.
8450 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
8452         * expmed.h (canonicalize_comparison): New declaration.
8453         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8454         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8455         * optabs.c (prepare_cmp_insn): Likewise.
8456         * rtl.h (unsigned_condition_p): New function which checks if a
8457         comparison operator is unsigned.
8459 2018-08-16  Nathan Sidwell  <nathan@acm.org>
8461         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8462         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8464 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8466         PR target/84711
8467         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8468         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8469         (mov<mov>): ..this and enable unconditionally.
8471 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8473         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8475 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
8477         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8478         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8479         "Common" with "Target".
8481 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8483         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8484         * doc/invoke.texi (mmitigate-rop): Remove.
8485         * config/i386/i386.c: Do not include "regrename.h".
8486         (ix86_rop_should_change_byte_p, reg_encoded_number)
8487         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8488         Remove.
8489         (ix86_reorg): Remove call to ix86_mitigate_rop.
8490         * config/i386/i386.md (attr "modrm_class"): Remove.
8491         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8492         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8493         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8495 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
8497         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8498         allow folding of mergeh() and mergel() for the float and double types.
8499         (fold_mergehl_helper): Rework to handle building a permute tree
8500         for float vectors.
8502 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8504         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8505         for TARGET_SSE.
8507 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
8509         * common.opt (fdiagnostics-show-labels): New option.
8510         * diagnostic-show-locus.c (class layout_range): Add field
8511         "m_label".
8512         (class layout): Add field "m_show_labels_p".
8513         (layout_range::layout_range): Add param "label" and use it to
8514         initialize m_label.
8515         (make_range): Pass in NULL for new "label" param of layout_range's
8516         ctor.
8517         (layout::layout): Initialize m_show_labels_p.
8518         (layout::maybe_add_location_range): Pass in loc_range->m_label
8519         when constructing layout_range instances.
8520         (struct line_label): New struct.
8521         (layout::print_any_labels): New member function.
8522         (layout::print_line): Call it if label-printing is enabled.
8523         (selftest::test_one_liner_labels): New test.
8524         (selftest::test_diagnostic_show_locus_one_liner): Call it.
8525         * diagnostic.c (diagnostic_initialize): Initialize
8526         context->show_labels_p.
8527         * diagnostic.h (struct diagnostic_context): Add field
8528         "show_labels_p".
8529         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8530         -fno-diagnostics-show-labels.
8531         * dwarf2out.c (gen_producer_string): Add
8532         OPT_fdiagnostics_show_labels to the ignored options.
8533         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
8534         param.
8535         (gcc_rich_location::maybe_add_expr): Likewise.
8536         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
8537         label" param, defaulting to NULL.
8538         (gcc_rich_location::add_expr): Add "label" param.
8539         (gcc_rich_location::maybe_add_expr): Likewise.
8540         (class text_range_label): New class.
8541         (class range_label_for_type_mismatch): New class.
8542         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
8543         of format_warning_va.
8544         (fmtwarn_n): Likewise for new params of format_warning_n_va.
8545         * lto-wrapper.c (merge_and_complain): Add
8546         OPT_fdiagnostics_show_labels to the "pick one setting" options.
8547         (append_compiler_options): Likewise to the dropped options.
8548         (append_diag_options): Likewise to the passed-on options.
8549         * opts.c (common_handle_option): Handle the new option.
8550         * selftest-diagnostic.c
8551         (test_diagnostic_context::test_diagnostic_context): Enable
8552         show_labels_p.
8553         * substring-locations.c: Include "gcc-rich-location.h".
8554         (format_warning_n_va): Add "fmt_label" and "param_label" params
8555         and use them as appropriate.
8556         (format_warning_va): Add "fmt_label" and "param_label" params,
8557         passing them on to format_warning_n_va.
8558         (format_warning_at_substring): Likewise.
8559         (format_warning_at_substring_n): Likewise.
8560         * substring-locations.h (format_warning_va): Add "fmt_label" and
8561         "param_label" params.
8562         (format_warning_n_va): Likewise.
8563         (format_warning_at_substring): Likewise.
8564         (format_warning_at_substring_n): Likewise.
8565         * toplev.c (general_init): Initialize global_dc->show_labels_p.
8567 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
8569         PR testsuite/86519
8570         * builtins.c (expand_builtin_memcmp): Do not expand the call
8571         when overflow is detected.
8573 2018-08-15  Martin Sebor  <msebor@redhat.com>
8575         PR tree-optimization/71625
8576         * config/aarch64/aarch64-builtins.c
8577         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
8579 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
8581         * config/s390/s390.c (s390_reorg): Remove loop.
8583 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8585         * config/darwin.c
8586          (darwin_function_switched_text_sections): Delete.
8587         * gcc/config/darwin.h
8588          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
8590 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8592         PR target/81685
8593         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
8594         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
8595         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
8597 2018-08-15  Martin Liska  <mliska@suse.cz>
8599         PR tree-optimization/86925
8600         * predict.c (expr_expected_value_1): When taking
8601         later predictor, assign also probability.
8602         Use fold_build2_initializer_loc in order to fold
8603         the expression in -frounding-math.
8605 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
8607         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
8608         patterns.
8609         (expand_vec_perm_1): Try the new method.
8611 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
8613         PR target/86547
8614         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
8615         Check whether lra_live_max_point is 0 before dividing.
8617 2018-08-14  Martin Sebor  <msebor@redhat.com>
8619         PR tree-optimization/86650
8620         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
8621         (vrp_prop::check_mem_ref): Same.
8623 2018-08-13  Liu Hao <lh_mouse@126.com>
8625         * pretty-print.c (eat_esc_sequence): Swap the foreground and
8626         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
8627         and clear it thereafter, as it only works for DBCS.
8629 2018-08-13  Liu Hao <lh_mouse@126.com>
8631         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
8632         handle returned by _get_osf_handle().
8634 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8636         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
8637         for folding vec_perm.
8639 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8641         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
8642         Add support for gimple-folding of vec_pack() and vec_unpack()
8643         intrinsics.
8645 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
8647         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
8648         vec_xst variants to the list.
8649         (rs6000_gimple_fold_builtin): Add support for folding unaligned
8650         vector loads and stores.
8652 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
8654         * config.gcc (rs6000-ibm-aix4.x): Delete.
8655         (rs6000-ibm-aix5.1): Delete.
8656         (rs6000-ibm-aix5.2): Delete.
8657         (rs6000-ibm-aix5.3): Delete.
8658         * config/rs6000/aix43.h: Delete.
8659         * config/rs6000/aix51.h: Delete.
8660         * config/rs6000/aix52.h: Delete.
8661         * config/rs6000/t-aix43: Delete.
8663 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
8665         * config/s390/s390.c (s390_decompose_constant_pool_ref):
8666         New function.
8667         (s390_decompose_address): Factor out constant pool ref
8668         decomposition.
8670 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8672         * config/nds32/nds32-predicates.c
8673         (nds32_can_use_bclr_p): Change return type as bool.
8674         (nds32_can_use_bset_p): Ditto.
8675         (nds32_can_use_btgl_p): Ditto.
8676         (nds32_can_use_bitci_p): Ditto.
8677         * config/nds32/nds32-protos.h
8678         (nds32_can_use_bclr_p): Change declaration.
8679         (nds32_can_use_bset_p): Ditto.
8680         (nds32_can_use_btgl_p): Ditto.
8681         (nds32_can_use_bitci_p): Ditto.
8683 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8685         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
8686         Support -msched-prolog-epilog option.
8687         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
8689 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8691         * common/config/nds32/nds32-common.c
8692         (nds32_option_optimization_table): Enalbe -malways-align.
8694 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8696         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
8697         extra_headers.
8698         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
8699         OPT_misr_secure_ case.
8700         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
8701         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
8702         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
8703         secure attribute.
8704         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
8705         (nds32_isr_info): New field security_level.
8706         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
8707         * config/nds32/nds32.md (return_internal): Consider critical attribute.
8708         * config/nds32/nds32.opt (misr-secure): New option.
8709         * config/nds32/nds32_init.inc: New file.
8710         * config/nds32/nds32_isr.h: New file.
8712 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
8714         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
8715         Update comment for atomic instructions.
8716         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
8717         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
8718         Remove.
8719         (atomic_loaddi): Revise fence expansion to only emit fence prior to
8720         load for __ATOMIC_SEQ_CST model.
8721         (atomic_loaddi_1): Remove float register target.
8722         (atomic_storedi): Handle CONST_INT values.
8723         (atomic_storedi_1): Remove float register source.  Add special case
8724         for zero value.
8725         (memory_barrier): New expander and insn.
8727 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
8729         PR tree-optimization/86835
8730         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
8731         new_stmt after def_gsi, make sure to insert new_square_stmt after
8732         that stmt, not 2 stmts before it.
8734 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
8736         PR target/82418
8737         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
8738         instead of SWI48.
8740 2018-08-10  Martin Liska  <mliska@suse.cz>
8742         PR target/83610
8743         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
8744         function type.
8745         * builtins.c (expand_builtin_expect_with_probability):
8746         New function.
8747         (expand_builtin_expect_with_probability): New function.
8748         (build_builtin_expect_predicate): Add new argumnet probability
8749         for BUILT_IN_EXPECT_WITH_PROBABILITY.
8750         (fold_builtin_expect):
8751         (fold_builtin_2):
8752         (fold_builtin_3):
8753         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
8754         * builtins.h (fold_builtin_expect): Set new argument.
8755         * doc/extend.texi: Document __builtin_expect_with_probability.
8756         * doc/invoke.texi: Likewise.
8757         * gimple-fold.c (gimple_fold_call): Pass new argument.
8758         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
8759         also BUILT_IN_EXPECT_WITH_PROBABILITY.
8760         * predict.c (get_predictor_value): New function.
8761         (expr_expected_value): Add new argument probability. Assume
8762         that predictor and probability are always non-null.
8763         (expr_expected_value_1): Likewise.  For __builtin_expect and
8764         __builtin_expect_with_probability set probability.  Handle
8765         combination in binary expressions.
8766         (tree_predict_by_opcode): Simplify code by simply calling
8767         get_predictor_value.
8768         (pass_strip_predict_hints::execute): Add handling of
8769         BUILT_IN_EXPECT_WITH_PROBABILITY.
8770         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
8771         new predictor.
8772         * tree.h (DECL_BUILT_IN_P): New function.
8774 2018-08-10  Martin Liska  <mliska@suse.cz>
8776         PR tree-optimization/85799
8777         * passes.def: Add argument for pass_strip_predict_hints.
8778         * predict.c (class pass_strip_predict_hints): Add new argument
8779         early_p.
8780         (strip_predictor_early): New function.
8781         (pass_strip_predict_hints::execute): Call the function to
8782         strip predictors.
8783         (strip_predict_hints): New function.
8784         * predict.def: Fix comment.
8786 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
8788         * Makefile.in: Clarify which tm.texi to copy over to assert the
8789         right to grant a GFDL license for all.
8791 2018-08-09  Jeff Law  <law@redhat.com>
8793         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
8794         unused variable.
8796 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
8798         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
8799         prototype.
8801 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8803         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
8804         reductions for variable-length vectors.
8806 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
8808         PR other/84889
8809         * common.opt (fdiagnostics-show-line-numbers): New option.
8810         * diagnostic-show-locus.c (class layout): Add fields
8811         "m_show_line_numbers_p" and "m_linenum_width";
8812         (num_digits): New function.
8813         (test_num_digits): New function.
8814         (layout::layout): Initialize new fields.  Update m_x_offset
8815         logic to handle any left margin.
8816         (layout::print_source_line): Print line number when requested.
8817         (layout::start_annotation_line): New member function.
8818         (layout::print_annotation_line): Call it.
8819         (layout::print_leading_fixits): Likewise.
8820         (layout::print_trailing_fixits): Likewise.  Update calls to
8821         move_to_column for new parameter.
8822         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
8823         it to potentially call start_annotation_line.
8824         (layout::show_ruler): Call start_annotation_line.
8825         (selftest::test_line_numbers_multiline_range): New selftest.
8826         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
8827         and selftest::test_line_numbers_multiline_range.
8828         * diagnostic.c (diagnostic_initialize): Initialize
8829         show_line_numbers_p.
8830         * diagnostic.h (struct diagnostic_context): Add field
8831         "show_line_numbers_p".
8832         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8833         -fno-diagnostics-show-line-numbers.
8834         * dwarf2out.c (gen_producer_string): Add
8835         OPT_fdiagnostics_show_line_numbers to the ignored options.
8836         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
8837         one setting" options.
8838         (append_compiler_options): Likewise to the dropped options.
8839         (append_diag_options): Likewise to the passed-on options.
8840         * opts.c (common_handle_option): Handle the new option.
8841         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
8843 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8845         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
8846         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
8847         third argument of type "const signed char" to descriptions of
8848         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
8849         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
8850         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
8851         __builtin_bcdsub_ov functions.
8853 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8855         PR tree-optimization/86858
8856         * tree-vect-loop.c (vect_is_simple_reduction): Restore
8857         flow_bb_inside_loop_p calls.
8859 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8861         PR tree-optimization/86871
8862         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
8863         instead of gimple_assign_lhs.
8865 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
8867         PR target/86887
8868         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
8869         register constraint to operand 0.
8870         (add<mode>3_carryinC): Likewise.
8871         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
8873 2018-08-09  Martin Liska  <mliska@suse.cz>
8875         PR c/86895
8876         * common.opt: Remove extra line.
8878 2018-08-09  Martin Liska  <mliska@suse.cz>
8880         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
8881         at the end of a line, make first letter capital and end up
8882         a sentence with a dot.
8883         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
8884         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
8885         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
8886         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
8887         (PARAM_MAX_ISL_OPERATIONS): Likewise.
8888         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
8889         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
8890         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
8891         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
8892         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
8893         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
8894         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
8895         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
8896         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
8897         (PARAM_TREE_REASSOC_WIDTH): Likewise.
8898         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
8899         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
8900         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
8902 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
8904         PR target/84332
8905         * config/s390/s390.c (s390_option_override_internal): Reduce the
8906         stack-clash-protection-probe-interval param if it would be too big
8907         for z900.
8909 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
8911         PR target/46179
8912         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
8913         * config/m68k/m68k.c (handle_move_double): Don't call
8914         m68k_final_prescan_insn.
8915         (m68k_adjust_decorated_operand): Renamed from
8916         m68k_final_prescan_insn, remove first and third operand and
8917         simplify.
8918         (print_operand): Call it.
8919         (print_operand_address): Call it.
8921 2018-08-08  Nathan Sidwell  <nathan@acm.org>
8923         * diagnostic.c (diagnostic_report_current_module): Use
8924         linemap_included_from & linemap_included_from_linemap.
8926 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
8928         * config/aarch64/aarch64-cores.def: Add phecda core.
8929         * config/aarch64/aarch64-tune.md: Regenerate.
8930         * doc/invoke.texi: Add phecda core.
8932 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8934         PR target/85295
8935         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
8936         definitions.
8937         * config/s390/s390.md ("movti"): Add more alternatives for
8938         constant to GPR copies.
8940 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8942         * config/s390/s390.c: Fix whitespace damage throughout the file.
8943         * config/s390/s390.h: Likewise.
8944         * config/s390/tpf.h: Likewise.
8946 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8948         * config/s390/s390.c (s390_loadrelative_operand_p):
8949         Remove TARGET_CPU_ZARCH usages.
8950         (s390_rtx_costs): Likewise.
8951         (s390_legitimate_constant_p): Likewise.
8952         (s390_cannot_force_const_mem): Likewise.
8953         (legitimate_reload_constant_p): Likewise.
8954         (s390_preferred_reload_class): Likewise.
8955         (legitimize_pic_address): Likewise.
8956         (legitimize_tls_address): Likewise.
8957         (s390_split_branches): Removed.
8958         (s390_add_execute): Removed.
8959         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
8960         (s390_mainpool_start): Likewise.
8961         (s390_mainpool_finish): Likewise.
8962         (s390_mainpool_cancel): Removed.
8963         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
8964         (s390_chunkify_cancel): Likewise.
8965         (s390_return_addr_rtx): Likewise.
8966         (s390_register_info): Remove split_branches_pending_p uages.
8967         (s390_optimize_register_info): Likewise.
8968         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
8969         split_branches_pending_p usages.
8970         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
8971         (s390_load_got): Likewise.
8972         (s390_expand_split_stack_prologue): Likewise.
8973         (output_asm_nops): Likewise.
8974         (s390_function_profiler): Likewise.
8975         (s390_emit_call): Likewise.
8976         (s390_conditional_register_usage): Likewise.
8977         (s390_optimize_prologue): Likewise.
8978         (s390_reorg): Remove TARGET_CPU_ZARCH and
8979         split_branches_pending_p usages.
8980         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
8981         usages.
8982         (s390_output_indirect_thunk_function): Likewise.
8983         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
8984         (TARGET_CPU_ZARCH_P): Removed.
8985         (struct machine_function): Remove split_branches_pending_p.
8986         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
8988 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8990         * common/config/s390/s390-common.c (processor_flags_table):
8991         Remove flags.
8992         * config.gcc: Remove with_arch/with_tune support.
8993         * config/s390/2064.md: Remove cpu attribute comparisons.
8994         * config/s390/driver-native.c (s390_host_detect_local_cpu):
8995         Remove MTN.
8996         * config/s390/linux.h (ASM_SPEC):
8997         Remove -march support.
8998         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
8999         Use a table to get an arch level.
9000         * config/s390/s390-opts.h (enum processor_type):
9001         Remove enum values.
9002         * config/s390/s390.c
9003         (processor_table): Remove entries, add arch_level values.
9004         (s390_issue_rate): Remove cases.
9005         (s390_option_override): Adjust
9006         s390_option_override_internal() call.
9007         (s390_option_override_internal): Remove deprecation warning.
9008         (s390_valid_target_attribute_tree): Adjust
9009         s390_option_override_internal() call.
9010         * config/s390/s390.h (struct s390_processor):
9011         Share with s390-c.c, add arch_level field.
9012         * config/s390/s390.md:
9013         Remove occurrences in cpu attribute.
9014         * config/s390/s390.opt: Remove -march/-mtune support.
9015         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9016         * doc/invoke.texi: Remove deprecation warning.
9018 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9020         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9021         global.
9022         (qdf24xx_tunings): Set vector cost structure to
9023         qdf24xx_vector_cost.
9025         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9026         <register_sextend>: Set to 3.
9028 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9030         PR target/86838
9031         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9032         * config/aarch64/aarch64-simd.md
9033         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9034         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9035         (aarch64_frecpx<mode>): ...this new pattern.
9036         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9037         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9039 2018-08-07  Martin Liska  <mliska@suse.cz>
9041         PR middle-end/83023
9042         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9043         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9044         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9045         * doc/extend.texi: Document that malloc attribute adds
9046         hit to compiler.
9048 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9050         PR target/86785
9051         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9052         Define to speculation_safe_value_not_needed.
9054 2018-08-06  Jeff Law  <law@redhat.com>
9056         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9057         the vr_values instance to cprop_into_stmt.
9058         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9059         (cprop_operand): Also query EVRP to determine if OP is a constant.
9061 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9063         * diagnostic.c (diagnostic_report_current_module): Reroll
9064         included-at loop.  Translate text.
9066 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9068         * function-tests.c (selftest::test_expansion_to_rtl): Call
9069         free_after_compilation.
9071 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9073         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9075 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9077         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9078         loops with memory block operations from getting unrolled.
9080 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9082         PR target/86807
9083         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9084         Define to speculation_safe_value_not_needed.
9086 2018-08-06  Jeff Law  <law@redhat.com>
9088         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9089         assert.
9091 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9093         PR target/86662
9094         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9095         with all enabled __intN types.
9097         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9099 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9101         * alias.c (record_set): Check for clobber high.
9102         * cfgexpand.c (expand_gimple_stmt): Likewise.
9103         * combine-stack-adj.c (single_set_for_csa): Likewise.
9104         * combine.c (find_single_use_1): Likewise.
9105         (set_nonzero_bits_and_sign_copies): Likewise.
9106         (get_combine_src_dest): Likewise.
9107         (is_parallel_of_n_reg_sets): Likewise.
9108         (try_combine): Likewise.
9109         (record_dead_and_set_regs_1): Likewise.
9110         (reg_dead_at_p_1): Likewise.
9111         (reg_dead_at_p): Likewise.
9112         * dce.c (deletable_insn_p): Likewise.
9113         (mark_nonreg_stores_1): Likewise.
9114         (mark_nonreg_stores_2): Likewise.
9115         * df-scan.c (df_find_hard_reg_defs): Likewise.
9116         (df_uses_record): Likewise.
9117         (df_get_call_refs): Likewise.
9118         * dwarf2out.c (mem_loc_descriptor): Likewise.
9119         * haifa-sched.c (haifa_classify_rtx): Likewise.
9120         * ira-build.c (create_insn_allocnos): Likewise.
9121         * ira-costs.c (scan_one_insn): Likewise.
9122         * ira.c (equiv_init_movable_p): Likewise.
9123         (rtx_moveable_p): Likewise.
9124         (interesting_dest_for_shprep): Likewise.
9125         * jump.c (mark_jump_label_1): Likewise.
9126         * postreload-gcse.c (record_opr_changes): Likewise.
9127         * postreload.c (reload_cse_simplify): Likewise.
9128         (struct reg_use): Add source expr.
9129         (reload_combine): Check for clobber high.
9130         (reload_combine_note_use): Likewise.
9131         (reload_cse_move2add): Likewise.
9132         (move2add_note_store): Likewise.
9133         * print-rtl.c (print_pattern): Likewise.
9134         * recog.c (decode_asm_operands): Likewise.
9135         (store_data_bypass_p): Likewise.
9136         (if_test_bypass_p): Likewise.
9137         * regcprop.c (kill_clobbered_value): Likewise.
9138         (kill_set_value): Likewise.
9139         * reginfo.c (reg_scan_mark_refs): Likewise.
9140         * reload1.c (maybe_fix_stack_asms): Likewise.
9141         (eliminate_regs_1): Likewise.
9142         (elimination_effects): Likewise.
9143         (mark_not_eliminable): Likewise.
9144         (scan_paradoxical_subregs): Likewise.
9145         (forget_old_reloads_1): Likewise.
9146         * reorg.c (find_end_label): Likewise.
9147         (try_merge_delay_insns): Likewise.
9148         (redundant_insn): Likewise.
9149         (own_thread_p): Likewise.
9150         (fill_simple_delay_slots): Likewise.
9151         (fill_slots_from_thread): Likewise.
9152         (dbr_schedule): Likewise.
9153         * resource.c (update_live_status): Likewise.
9154         (mark_referenced_resources): Likewise.
9155         (mark_set_resources): Likewise.
9156         * rtl.c (copy_rtx): Likewise.
9157         * rtlanal.c (reg_referenced_p): Likewise.
9158         (single_set_2): Likewise.
9159         (noop_move_p): Likewise.
9160         (note_stores): Likewise.
9161         * sched-deps.c (sched_analyze_reg): Likewise.
9162         (sched_analyze_insn): Likewise.
9164 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9166         * cse.c (invalidate_reg): New function extracted from...
9167         (invalidate): ...here.
9168         (canonicalize_insn): Check for clobber high.
9169         (invalidate_from_clobbers): invalidate clobber highs.
9170         (invalidate_from_sets_and_clobbers): Likewise.
9171         (count_reg_usage): Check for clobber high.
9172         (insn_live_p): Likewise.
9173         * cselib.c (cselib_expand_value_rtx_1):Likewise.
9174         (cselib_invalidate_regno): Check for clobber in setter.
9175         (cselib_invalidate_rtx): Pass through setter.
9176         (cselib_invalidate_rtx_note_stores):
9177         (cselib_process_insn): Check for clobber high.
9178         * cselib.h (cselib_invalidate_rtx): Add operand.
9180 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9182         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9183         (mark_not_eliminable): Likewise.
9184         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9185         * lra-lives.c (process_bb_lives): Check for clobber high.
9186         * lra.c (new_insn_reg): Remember clobber highs.
9187         (collect_non_operand_hard_regs): Check for clobber high.
9188         (lra_set_insn_recog_data): Likewise.
9189         (add_regs_to_insn_regno_info): Likewise.
9190         (lra_update_insn_regno_info): Likewise.
9192 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9194         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9195         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9197 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9199         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9200         (copy_insn_1): Likewise.
9201         (gen_hard_reg_clobber_high): New gen function.
9202         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9203         * genemit.c (gen_exp): Likewise.
9204         (gen_emit_seq): Pass through info.
9205         (gen_insn): Check for CLOBBER_HIGH.
9206         (gen_expand): Pass through info.
9207         (gen_split): Likewise.
9208         (output_add_clobbers): Likewise.
9209         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9210         (remove_clobbers): Likewise.
9211         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9213 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9215         * doc/rtl.texi (clobber_high): Add.
9216         (parallel): Add in clobber high
9217         * rtl.c (rtl_check_failed_code3): Add function.
9218         * rtl.def (CLOBBER_HIGH): Add expression.
9219         * rtl.h (RTL_CHECKC3): Add macro.
9220         (rtl_check_failed_code3): Add declaration.
9221         (XC3EXP): Add macro.
9223 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9225         PR target/86386
9226         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9227         cfun->machine->max_used_stack_alignment if needed.
9229 2018-08-04  Martin Sebor  <msebor@redhat.com>
9231         PR tree-optimization/86571
9232         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9233         NaN output to 4.
9235 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9237         PR target/86799
9238         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9239         Define.
9241 2018-08-03  Jeff Law  <law@redhat.com>
9243         PR target/86795
9244         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9245         Define to speculation_safe_value_not_needed.
9247 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9249         * doc/gcov.texi (-x): Remove duplicate "to".
9250         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9251         (-Wif-not-aligned): Remove duplicate "is".
9252         (-flto): Remove duplicate "the".
9253         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9254         duplicate "v5.00.b".
9255         (MSP430 Options): Remove duplicate "and" from the description
9256         of "-mgprel-sec=regexp".
9257         (x86 Options): Remove duplicate copies of "vmldLog102" and
9258         vmlsLog104 from description of "-mveclibabi=type".
9260 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9262         * internal-fn.h (first_commutative_argument): Declare.
9263         * internal-fn.c (first_commutative_argument): New function.
9264         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9265         restrictions for pattern statements.  Use first_commutative_argument
9266         to look for commutative operands in calls to internal functions.
9268 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
9270         * Makefile.in (wide-int-range.o): New.
9271         * tree-vrp.c: Move all the wide_int_* functions to...
9272         * wide-int-range.cc: ...here.
9273         * tree-vrp.h: Move all the wide_int_* prototypes to...
9274         * wide-int-range.h: ...here.
9276 2018-08-03  Tom de Vries  <tdevries@suse.de>
9278         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9279         UI_NONE.
9280         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9281         * except.c (output_function_exception_table): Do early exit if
9282         targetm_common.except_unwind_info (&global_options) == UI_NONE.
9284 2018-08-03  Martin Liska  <mliska@suse.cz>
9286         * predict.c (dump_prediction): Change to 2 digits
9287         in fraction part.
9289 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9291         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9292         neon_dup_q to...
9293         (falkor_am_1_gtov_gtov): ... a new insn reservation.
9295 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
9297         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9298         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9299         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9300         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9301         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9303 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9305         * diagnostic-show-locus.c (diagnostic_show_locus): Use
9306         pp_take_prefix when saving the existing prefix.
9307         * diagnostic.c (diagnostic_append_note): Likewise.
9308         * langhooks.c (lhd_print_error_function): Likewise.
9309         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9310         param's type.  Free the existing prefix.
9311         (pp_take_prefix): New function.
9312         (pretty_printer::pretty_printer): Drop the prefix parameter.
9313         Rename the length parameter to match the comment.
9314         (pretty_printer::~pretty_printer): Free the prefix.
9315         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9316         parameter.
9317         (struct pretty_printer): Drop the "const" from "prefix" field's
9318         type and clarify memory management.
9319         (pp_set_prefix): Drop the "const" from the 2nd param.
9320         (pp_take_prefix): New decl.
9322 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9324         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9325         for word_mode_ok here instead of passing as argument.
9326         (expand_block_compare): Change select_block_compare_mode() call.
9327         (expand_strncmp_gpr_sequence): New function.
9328         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9330 2018-08-02  Jeff Law  <law@redhat.com>
9332         PR target/86790
9333         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9334         Define to speculation_safe_value_not_needed.
9336         PR target/86784
9337         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9338         Define to speculation_safe_value_not_needed.
9340 2018-08-02  Tom de Vries  <tdevries@suse.de>
9342         PR target/86660
9343         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9344         function.  Return UI_TARGET unconditionally.
9345         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9346         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9348 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9350         * genemit.c (print_overload_test): Fix typo.
9352 2018-08-02  Richard Biener  <rguenther@suse.de>
9354         PR tree-optimization/86816
9355         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9356         which checks for value availability before querying it.
9357         (gvn_uses_equal): Use it.
9358         (same_succ_hash): Likewise.
9359         (gimple_equal_p): Likewise.
9361 2018-08-02  Nick Clifton  <nickc@redhat.com>
9363         PR target/86813
9364         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9365         Define to speculation_safe_value_not_needed.
9367         PR target/86810
9368         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9369         Define to speculation_safe_value_not_needed.
9371         PR target/86810
9372         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9373         Define to speculation_safe_value_not_needed.
9375         PR target/86803
9376         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9377         Define to speculation_safe_value_not_needed.
9379         PR target/86797
9380         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9381         Define to speculation_safe_value_not_needed.
9383         PR target/86791
9384         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9385         Define to speculation_safe_value_not_needed.
9387         PR target/86789
9388         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9389         Define to speculation_safe_value_not_needed.
9391         PR target/86787
9392         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9393         Define to speculation_safe_value_not_needed.
9395         PR target/86782
9396         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9397         speculation_safe_value_not_needed.
9399         PR target/86781
9400         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9401         to speculation_safe_value_not_needed.
9403 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9405         * doc/md.texi: Expand the documentation of instruction names
9406         to mention port-local uses.  Document '@' in pattern names.
9407         * read-md.h (overloaded_instance, overloaded_name): New structs.
9408         (mapping): Declare.
9409         (md_reader::handle_overloaded_name): New member function.
9410         (md_reader::get_overloads): Likewise.
9411         (md_reader::m_first_overload): New member variable.
9412         (md_reader::m_next_overload_ptr): Likewise.
9413         (md_reader::m_overloads_htab): Likewise.
9414         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9415         m_next_overload_ptr and m_overloads_htab.
9416         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9417         (get_mode_token, get_code_token, get_int_token): New functions.
9418         (map_attr_string): Add an optional argument that passes back
9419         the associated iterator.
9420         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9421         (md_reader::handle_overloaded_name, add_overload_instance): New
9422         functions.
9423         (apply_iterators): Handle '@' names.  Report an error if '@'
9424         is used without iterators.
9425         (initialize_iterators): Initialize the new iterator_group fields.
9426         * genopinit.c (handle_overloaded_code_for)
9427         (handle_overloaded_gen): New functions.
9428         (main): Use them to print declarations of maybe_code_for_* and
9429         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9430         * genemit.c (print_overload_arguments, print_overload_test)
9431         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9432         (main): Use it to print definitions of maybe_code_for_* and
9433         maybe_gen_* functions.
9434         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9435         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9436         instead of explicit mode checks.
9437         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9438         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9439         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9440         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9441         (aarch64_expand_compare_and_swap): Likewise
9442         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9443         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9444         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9445         (aarch64_constant_pool_reload_icode): Delete.
9446         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9447         instead of aarch64_constant_pool_reload_icode.  Use
9448         code_for_aarch64_reload_mov instead of explicit mode checks.
9449         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9450         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9451         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9452         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9453         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9454         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9455         (aarch64_atomic_load_op_code): Delete.
9456         (aarch64_emit_atomic_load_op): Likewise.
9457         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9458         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
9459         instead of aarch64_emit_atomic_load_op.
9460         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9461         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9462         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9463         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9464         character before the pattern name.
9465         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9466         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9467         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9468         (aarch64_frecps<mode>): Likewise.
9469         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9470         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9471         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9472         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9473         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9475 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9477         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9478         Allow HFmode constants if TARGET_FP_F16INST.
9480 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
9482         PR target/86014
9483         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9484         No longer check last store for clobber of address register.
9486 2018-08-02  Martin Liska  <mliska@suse.cz>
9488         PR gcov-profile/86817
9489         * gcov.c (process_all_functions): New function.
9490         (main): Call it.
9491         (process_file): Move functions processing to
9492         process_all_functions.
9494 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9496         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9497         "const" to the "gimple *" and "rtx_insn *" parameters.
9498         * dumpfile.h (dump_user_location_t::dump_user_location_t):
9499         Likewise.
9500         (dump_location_t::dump_location_t): Likewise.
9502 2018-08-01  Martin Sebor  <msebor@redhat.com>
9504         PR tree-optimization/86650
9505         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9506         rather than a "gcall *".  Directly pass the data of interest
9507         to percent_K_format, rather than building a temporary CALL_EXPR
9508         to hold it.
9509         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9510         (gimple_fold_builtin_strncat): Adjust.
9511         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9512         gcall* argument with gimple*.
9513         * gimple-ssa-warn-restrict.c (check_call): Same.
9514         (wrestrict_dom_walker::before_dom_children): Same.
9515         (builtin_access::builtin_access): Same.
9516         (check_bounds_or_overlap): Same
9517         (maybe_diag_overlap): Same.
9518         (maybe_diag_offset_bounds): Same.
9519         * tree-diagnostic.c (default_tree_printer): Move usage of
9520         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9521         to this callsite.
9522         * tree-pretty-print.c (percent_K_format): Add argument.
9523         * tree-pretty-print.h: Add argument.
9524         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9525         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
9526         (maybe_diag_stxncpy_trunc): Same.
9527         (handle_builtin_stxncpy): Same.
9528         (handle_builtin_strcat): Same.
9530 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9532         * match.pd: Optimise pointer range checks.
9534 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9536         PR tree-optimization/86758
9537         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
9538         to remove pattern statements.
9540 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9542         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
9543         result of dfs_enumerate_from when constructing stmt_vec_infos,
9544         instead of additionally calling get_loop_body.
9546 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9548         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
9549         parameter.
9550         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
9551         When creating an iv, assert that the step is not known to be zero.
9552         (vect_setup_realignment): Update call accordingly.
9553         * tree-vect-stmts.c (vectorizable_store): Likewise.
9554         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
9556 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9558         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
9559         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
9560         (vectorizable_reduction): Likewise.
9561         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9562         (vect_detect_hybrid_slp_stmts): Likewise.
9563         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9565 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
9567         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
9568         (wide_int_set_zero_nonzero_bits): ...this.
9569         (zero_nonzero_bits_from_vr): Rename to...
9570         (vrp_set_zero_nonzero_bits): ...this.
9571         (extract_range_from_multiplicative_op_1): Abstract wide int
9572         code...
9573         (wide_int_range_multiplicative_op): ...here.
9574         (extract_range_from_binary_expr_1): Extract wide int binary
9575         operations into their own functions.
9576         (wide_int_range_lshift): New.
9577         (wide_int_range_can_optimize_bit_op): New.
9578         (wide_int_range_shift_undefined_p): New.
9579         (wide_int_range_bit_xor): New.
9580         (wide_int_range_bit_ior): New.
9581         (wide_int_range_bit_and): New.
9582         (wide_int_range_trunc_mod): New.
9583         (extract_range_into_wide_ints): New.
9584         (vrp_shift_undefined_p): New.
9585         (extract_range_from_multiplicative_op): New.
9586         (vrp_can_optimize_bit_op): New.
9587         * tree-vrp.h (value_range::dump): New.
9588         (wide_int_range_multiplicative_op): New.
9589         (wide_int_range_lshift):New.
9590         (wide_int_range_shift_undefined_p): New.
9591         (wide_int_range_bit_xor): New.
9592         (wide_int_range_bit_ior): New.
9593         (wide_int_range_bit_and): New.
9594         (wide_int_range_trunc_mod): New.
9595         (zero_nonzero_bits_from_bounds): Rename to...
9596         (wide_int_set_zero_nonzero_bits): ...this.
9597         (zero_nonzero_bits_from_vr): Rename to...
9598         (vrp_set_zero_nonzero_bits): ...this.
9599         (range_easy_mask_min_max): Rename to...
9600         (wide_int_range_can_optimize_bit_op): this.
9601         * vr-values.c (simplify_bit_ops_using_ranges): Rename
9602         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
9604 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9606         * tree-vectorizer.h (vect_orig_stmt): New function.
9607         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
9608         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
9609         (vect_create_epilog_for_reduction): Likewise.
9610         (vectorizable_live_operation): Likewise.
9611         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9612         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
9613         * tree-vect-stmts.c (vectorizable_call): Likewise.
9614         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
9616 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9618         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
9619         argument.
9620         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
9621         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
9622         (vect_transform_loop): Likewise.
9623         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9625 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9627         * tree-vectorizer.h (vect_schedule_slp): Return void.
9628         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9629         (vect_schedule_slp): Likewise.
9631 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9633         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
9634         argument.
9635         (vect_transform_loop): Update calls accordingly.  Schedule SLP
9636         instances before the main loop, if any exist.
9638 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9640         PR tree-optimization/86749
9641         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
9642         If the lhs is used in a COND_EXPR, check that it is being used
9643         as the "then" or "else" value.
9645 2018-08-01  Tom de Vries  <tdevries@suse.de>
9647         PR target/86800
9648         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9649         speculation_safe_value_not_needed.
9651 2018-08-01  Richard Biener  <rguenther@suse.de>
9653         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
9654         as base and offset.
9656 2018-08-01  Martin Liska  <mliska@suse.cz>
9658         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
9659         format how successful transformation is dumped.
9660         (gimple_mod_pow2_value_transform): Likewise.
9661         (gimple_mod_subtract_transform): Likewise.
9662         (gimple_stringops_transform): Likewise.
9664 2018-08-01  Martin Liska  <mliska@suse.cz>
9666         PR value-prof/35543
9667         * value-prof.c (interesting_stringop_to_profile_p):
9668         Simplify the code and add BUILT_IN_MEMMOVE.
9669         (gimple_stringops_transform): Likewise.
9671 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
9673         * config/aarch64/aarch64-simd.md
9674         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
9675         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
9676         use GPI iterator instead of SI mode.
9678 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9680         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
9681         rs6000_speculation_barrier.
9682         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
9683         new barrier pattern name.
9685 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9687         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
9688         (speculation_barrier): New insn.
9690 2018-08-01  Richard Biener  <rguenther@suse.de>
9692         PR bootstrap/86724
9693         * graphite.h: Include isl/id.h and isl/space.h to allow build
9694         with ISL 0.20.
9696 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
9698         PR target/86651
9699         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9700         mode for COFF targets.
9701         * defaults.h (TARGET_COFF): Define.
9702         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
9703         TARGET_COFF): Define.
9704         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
9705         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
9706         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
9708 2018-07-31  Alexandre Oliva <oliva@adacore.com>
9710         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
9711         * gimple-streamer-out.c (output_bb): Save it.
9712         * lto-streamer-in.c (input_struct_function_base): Restore
9713         instance discriminator if available.  Create map on demand.
9714         * lto-streamer-out.c (output_struct_function_base): Save it if
9715         available.
9716         * final.c (decl_to_instance_map): Document LTO strategy.
9718 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
9719             Olivier Hainque  <hainque@adacore.com>
9721         * debug.h (decl_to_instance_map_t): New type.
9722         (decl_to_instance_map): Declare.
9723         (maybe_create_decl_to_instance_map): New inline function.
9724         * final.c (bb_discriminator, last_bb_discriminator): New statics,
9725         to track basic block discriminators.
9726         (final_start_function_1): Initialize them.
9727         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
9728         bb_discriminator.
9729         (decl_to_instance_map): New variable.
9730         (map_decl_to_instance, maybe_set_discriminator): New functions.
9731         (notice_source_line): Set discriminator.
9733 2018-07-31  Ian Lance Taylor  <iant@golang.org>
9735         * targhooks.c (default_have_speculation_safe_value): Add
9736         ATTRIBUTE_UNUSED.
9738 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9740         * dump-context.h: Include "pretty-print.h".
9741         (dump_context::refresh_dumps_are_enabled): New decl.
9742         (dump_context::emit_item): New decl.
9743         (class dump_context): Add fields "m_test_pp" and
9744         "m_test_pp_flags".
9745         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9746         (temp_dump_context::get_dumped_text): New decl.
9747         (class temp_dump_context): Add field "m_pp".
9748         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
9749         (dump_context::refresh_dumps_are_enabled): ...and add a test for
9750         m_test_pp.
9751         (set_dump_file): Update for above change.
9752         (set_alt_dump_file): Likewise.
9753         (dump_loc): New overload, taking a pretty_printer *.
9754         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
9755         to any test pretty-printer.
9756         (make_item_for_dump_gimple_stmt): New function, adapted from
9757         optinfo::add_gimple_stmt.
9758         (dump_context::dump_gimple_stmt): Call it, and use the result,
9759         eliminating the direct usage of dump_file and alt_dump_file in
9760         favor of indirectly using them via emit_item.
9761         (make_item_for_dump_gimple_expr): New function, adapted from
9762         optinfo::add_gimple_expr.
9763         (dump_context::dump_gimple_expr): Call it, and use the result,
9764         eliminating the direct usage of dump_file and alt_dump_file in
9765         favor of indirectly using them via emit_item.
9766         (make_item_for_dump_generic_expr): New function, adapted from
9767         optinfo::add_tree.
9768         (dump_context::dump_generic_expr): Call it, and use the result,
9769         eliminating the direct usage of dump_file and alt_dump_file in
9770         favor of indirectly using them via emit_item.
9771         (make_item_for_dump_printf_va): New function, adapted from
9772         optinfo::add_printf_va.
9773         (make_item_for_dump_printf): New function.
9774         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
9775         and use the result, eliminating the direct usage of dump_file and
9776         alt_dump_file in favor of indirectly using them via emit_item.
9777         (make_item_for_dump_dec): New function.
9778         (dump_context::dump_dec): Call it, and use the result,
9779         eliminating the direct usage of dump_file and alt_dump_file in
9780         favor of indirectly using them via emit_item.
9781         (make_item_for_dump_symtab_node): New function, adapted from
9782         optinfo::add_symtab_node.
9783         (dump_context::dump_symtab_node): Call it, and use the result,
9784         eliminating the direct usage of dump_file and alt_dump_file in
9785         favor of indirectly using them via emit_item.
9786         (dump_context::begin_scope): Reimplement, avoiding direct usage
9787         of dump_file and alt_dump_file in favor of indirectly using them
9788         via emit_item.
9789         (dump_context::emit_item): New member function.
9790         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9791         Set up test pretty-printer on the underlying context.  Call
9792         refresh_dumps_are_enabled.
9793         (temp_dump_context::~temp_dump_context): Call
9794         refresh_dumps_are_enabled.
9795         (temp_dump_context::get_dumped_text): New member function.
9796         (selftest::verify_dumped_text): New function.
9797         (ASSERT_DUMPED_TEXT_EQ): New macro.
9798         (selftest::test_capture_of_dump_calls): Run all tests twice, with
9799         and then without optinfo enabled.  Add uses of
9800         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
9801         * dumpfile.h: Update comment for the dump_* API.
9802         * optinfo-emit-json.cc
9803         (selftest::test_building_json_from_dump_calls): Update for new
9804         param for temp_dump_context ctor.
9805         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
9806         and "m_owned" field.
9807         (optinfo_item::~optinfo_item): Likewise.
9808         (optinfo::add_item): New member function.
9809         (optinfo::emit): Update comment.
9810         (optinfo::add_string): Delete.
9811         (optinfo::add_printf): Delete.
9812         (optinfo::add_printf_va): Delete.
9813         (optinfo::add_gimple_stmt): Delete.
9814         (optinfo::add_gimple_expr): Delete.
9815         (optinfo::add_tree): Delete.
9816         (optinfo::add_symtab_node): Delete.
9817         (optinfo::add_dec): Delete.
9818         * optinfo.h (class dump_context): New forward decl.
9819         (optinfo::add_item): New decl.
9820         (optinfo::add_string): Delete.
9821         (optinfo::add_printf): Delete.
9822         (optinfo::add_printf_va): Delete.
9823         (optinfo::add_gimple_stmt): Delete.
9824         (optinfo::add_gimple_expr): Delete.
9825         (optinfo::add_tree): Delete.
9826         (optinfo::add_symtab_node): Delete.
9827         (optinfo::add_dec): Delete.
9828         (optinfo::add_poly_int): Delete.
9829         (optinfo_item::optinfo_item): Remove "owned" param.
9830         (class optinfo_item): Remove field "m_owned".
9832 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9834         PR middle-end/86705
9835         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
9836         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
9837         requested variable alignment.
9838         (expand_one_ssa_partition): Likewise.
9839         (expand_one_var): Likewise.
9841 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9843         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9844         to speculation_safe_value_not_needed.
9846 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9848         * targhooks.h (speculation_safe_value_not_needed): New prototype.
9849         * targhooks.c (speculation_safe_value_not_needed): New function.
9850         * target.def (have_speculation_safe_value): Update documentation.
9851         * doc/tm.texi: Regenerated.
9853 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9855         * config/aarch64/iterators.md (ALLI_TI): New iterator.
9856         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
9857         expand.
9858         (despeculate_copy<ALLI:mode>_insn): New insn.
9859         (despeculate_copyti_insn): New insn.
9860         (despeculate_simple<ALLI:mode>): New insn
9861         (despeculate_simpleti): New insn.
9862         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
9863         function.
9864         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
9865         aarch64_speculation_safe_value.
9866         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
9868 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9870         * config/aarch64/aarch64-speculation.cc: New file.
9871         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
9872         before pass_reorder_blocks.
9873         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
9874         prototype.
9875         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
9876         X14 and X15 when tracking speculation.
9877         * config/aarch64/aarch64.md (register name constants): Add
9878         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
9879         (unspec): Add UNSPEC_SPECULATION_TRACKER.
9880         (speculation_barrier): New insn attribute.
9881         (cmp<mode>): Allow SP in comparisons.
9882         (speculation_tracker): New insn.
9883         (speculation_barrier): Add speculation_barrier attribute.
9884         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
9885         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
9886         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
9888 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9890         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
9891         aarch64_track_speculation is true.
9892         (tb<optab><mode>1): Likewise.
9893         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
9894         generate CB[N]Z when tracking speculation.
9895         (aarch64_split_compare_and_swap): Likewise.
9896         (aarch64_split_atomic_op): Likewise.
9898 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9900         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
9902 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9904         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
9905         (speculation_barrier): New insn.
9907 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9909         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
9910         * config/arm/arm.md (speculation_barrier): New expand.
9911         (speculation_barrier_insn): New pattern.
9913 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9915         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
9916         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
9917         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
9918         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
9919         list.
9920         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
9921         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
9922         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
9923         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
9924         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
9925         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
9926         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
9927         * builtins.c (expand_speculation_safe_value): New function.
9928         (expand_builtin): Call it.
9929         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
9930         * doc/extend.texi: Document __builtin_speculation_safe_value.
9931         * doc/md.texi: Document "speculation_barrier" pattern.
9932         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
9933         TARGET_HAVE_SPECULATION_SAFE_VALUE.
9934         * doc/tm.texi: Regenerated.
9935         * target.def (have_speculation_safe_value, speculation_safe_value): New
9936         hooks.
9937         * targhooks.c (default_have_speculation_safe_value): New function.
9938         (default_speculation_safe_value): New function.
9939         * targhooks.h (default_have_speculation_safe_value): Add prototype.
9940         (default_speculation_safe_value): Add prototype.
9942 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9944         * dump-context.h (dump_context::dump_loc): New decl.
9945         * dumpfile.c (dump_context::dump_loc): New member function.
9946         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
9947         and dump_gimple_stmt.
9948         (dump_context::dump_gimple_expr_loc): Likewise, using
9949         dump_gimple_expr.
9950         (dump_context::dump_generic_expr_loc): Likewise, using
9951         dump_generic_expr.
9952         (dump_context::dump_printf_loc_va): Likewise, using
9953         dump_printf_va.
9954         (dump_context::begin_scope): Explicitly using the global function
9955         "dump_loc", rather than the member function.
9957 2018-07-31  Martin Sebor  <msebor@redhat.com>
9959         PR tree-optimization/86741
9960         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
9962 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
9964         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
9965         depend on whether prefetch instructions will be emitted or not.
9966         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
9967         will be emitted or not.
9968         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
9969         (TARGET_SETMEM_PFD): New macros.
9971 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9973         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
9974         (NULL_STMT_VEC_INFO): Delete.
9975         (stmt_vec_info::operator*): Likewise.
9976         (stmt_vec_info::operator gimple *): Likewise.
9977         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
9978         of NULL_STMT_VEC_INFO.
9979         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9980         (vect_reassociating_reduction_p): Likewise.
9981         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9982         (vectorizable_store): Likewise.
9983         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
9984         (vec_info::free_stmt_vec_infos): Likewise.
9986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9988         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
9989         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
9991 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9993         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
9994         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9995         (vec_info::free_stmt_vec_info): New private member functions.
9996         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
9997         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
9998         * tree-parloops.c (gather_scalar_reductions): Remove calls to
9999         set_stmt_vec_info_vec and free_stmt_vec_infos.
10000         * tree-vect-loop.c (_loop_vec_info): Remove call to
10001         set_stmt_vec_info_vec.
10002         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10003         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10004         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10005         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10006         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10007         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10008         with the clearing in the xcalloc.
10009         (stmt_vec_info_vec): Delete.
10010         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10011         (vectorize_loops): Likewise.
10012         (vec_info::~vec_info): Remove argument from call to
10013         free_stmt_vec_infos.
10014         (vec_info::add_stmt): Remove vinfo argument from call to
10015         new_stmt_vec_info.
10017 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10019         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10020         rather than a gimple stmt.
10021         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10022         information for pattern statements when passed the original
10023         statement; instead wait to be passed the pattern statement itself.
10024         Don't call set_vinfo_for_stmt here.
10025         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10026         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10027         stmt_vec_infos here.
10028         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10029         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10030         stmt_vec_infos entry.
10032 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10034         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10035         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10036         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10037         * tree-vect-stmts.c (vectorizable_call): Likewise.
10038         (vectorizable_simd_clone_call): Likewise.
10040 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10042         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10043         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10044         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10045         * tree-vect-loop.c (vect_transform_loop): Likewise.
10046         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10047         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10049 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10051         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10052         (vect_dr_stmt): Delete.
10053         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10054         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10055         of DR_VECT_AUX.
10056         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10057         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10058         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10059         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10060         (vect_analyze_data_refs): Likewise.
10061         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10062         argument.
10063         (vect_find_same_alignment_drs): Likewise.
10064         (vect_slp_analyze_node_dependences): Update calls accordingly.
10065         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10066         instead of DR_VECT_AUX.
10067         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10068         of a vector data references.  Use vec_info::lookup_dr instead of
10069         DR_VECT_AUX.
10070         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10071         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10072         instead of DR_VECT_AUX.
10074 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10076         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10077         dr_vec_info.
10078         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10079         accordingly.
10080         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10081         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10082         (vect_gen_prolog_loop_niters): Likewise.
10084 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10086         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10087         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10088         (vect_known_alignment_in_bytes, vect_dr_behavior)
10089         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10090         instead of data_references.  Update calls to other routines for
10091         which the same change has been made.
10092         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10093         dr_vec_infos instead of stmt_vec_infos.
10094         (vect_analyze_data_ref_dependence): Update call accordingly.
10095         (vect_slp_analyze_data_ref_dependence)
10096         (vect_record_base_alignments): Use DR_VECT_AUX.
10097         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10098         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10099         (vector_alignment_reachable_p, vect_get_data_access_cost)
10100         (vect_peeling_supportable, vect_analyze_group_access_1)
10101         (vect_analyze_group_access, vect_analyze_data_ref_access)
10102         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10103         (vect_compile_time_alias, vect_small_gap_p)
10104         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10105         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10106         instead of data_references.  Update calls to other routines for
10107         which the same change has been made.
10108         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10109         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10110         (vect_slp_analyze_and_verify_node_alignment)
10111         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10112         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10113         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10114         above changes.
10115         (_vect_peel_info::dr): Replace with...
10116         (_vect_peel_info::dr_info): ...this new field.
10117         (vect_peeling_hash_get_most_frequent)
10118         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10119         (vect_peeling_hash_get_lowest_cost):
10120         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10121         routines for which the same change has been made.
10122         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10123         data_reference.
10124         * tree-vect-loop-manip.c (get_misalign_in_elems)
10125         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10126         above changes.
10127         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10128         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10129         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10130         (get_group_load_store_type, get_negative_load_store_type)
10131         (vect_get_data_ptr_increment, vectorizable_store)
10132         (vectorizable_load): Likewise.
10133         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10134         Update calls to other routines for which the same change has been made.
10136 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10138         * tree-vectorizer.h (vec_info::move_dr): New member function.
10139         (dataref_aux): Rename to...
10140         (dr_vec_info): ...this and add "dr" and "stmt" fields.
10141         (_stmt_vec_info::dr_aux): Update accordingly.
10142         (_stmt_vec_info::data_ref_info): Delete.
10143         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10144         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10145         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10146         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10147         of data_ref.
10148         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
10149         (STMT_VINFO_DR_INFO): New macro.
10150         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10151         (set_dr_misalignment): Update after rename of dataref_aux.
10152         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
10153         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10154         initialization of STMT_VINFO_DATA_REF.
10155         * tree-vectorizer.c (vec_info::move_dr): New function.
10156         * tree-vect-patterns.c (vect_recog_bool_pattern)
10157         (vect_recog_mask_conversion_pattern)
10158         (vect_recog_gather_scatter_pattern): Use it.
10159         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10160         the "dr" and "stmt" fields of dr_vec_info instead of
10161         STMT_VINFO_DATA_REF.
10163 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10165         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10166         (is_pattern_stmt_p): Use it.
10167         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10168         on pattern statements.
10170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10172         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10173         original stmt as a stmt_vec_info rather than a gimple stmt.
10174         (vect_pattern_recog_1): Take the statement directly as a
10175         stmt_vec_info, rather than via a gimple_stmt_iterator.
10176         Update call to vect_mark_pattern_stmts.
10177         (vect_pattern_recog): Update calls accordingly.
10179 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10181         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10182         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10183         a vect_def_type for the first argument.
10184         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10185         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
10186         operand if it isn't defined by a vectorized statement.
10187         (vect_build_gather_load_calls): Remove the mask_dt argument and
10188         update calls to vect_get_vec_def_for_stmt_copy.
10189         (vectorizable_bswap): Likewise the dt argument.
10190         (vectorizable_call): Update calls to vectorizable_bswap and
10191         vect_get_vec_def_for_stmt_copy.
10192         (vectorizable_simd_clone_call, vectorizable_assignment)
10193         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10194         (vectorizable_comparison): Update calls to
10195         vect_get_vec_def_for_stmt_copy.
10196         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
10197         vect_is_simple_use.
10198         (vect_get_loop_based_defs): Remove dt argument and update call
10199         to vect_get_vec_def_for_stmt_copy.
10200         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10201         and vect_get_vec_def_for_stmt_copy.
10202         (vectorizable_load): Update calls to vect_build_gather_load_calls
10203         and vect_get_vec_def_for_stmt_copy.
10204         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10205         (vectorizable_reduction, vectorizable_live_operation): Update calls
10206         to vect_get_vec_def_for_stmt_copy.
10208 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10210         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10211         and gimple stmt arguments with a stmt_vec_info.
10212         (vect_record_base_alignments): Update calls accordingly.
10213         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10214         and gimple stmt arguments with a stmt_vec_info.
10215         (vect_build_slp_tree_1): Remove vinfo argument and update call
10216         to vect_record_max_nunits.
10217         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10218         and vect_record_max_nunits.
10220 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10222         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10223         file and take a stmt_vec_info instead of a gimple stmt.
10224         (supportable_widening_operation, vect_finish_replace_stmt)
10225         (vect_finish_stmt_generation, vect_get_store_rhs)
10226         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10227         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10228         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10229         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10230         (vect_create_data_ref_ptr, bump_vector_ptr)
10231         (vect_permute_store_chain, vect_setup_realignment)
10232         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10233         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10234         (vectorizable_reduction, vectorizable_induction)
10235         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10236         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10237         than gimple stmts as arguments.
10238         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10239         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10240         (can_group_stmts_p, vect_check_gather_scatter)
10241         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10242         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10243         (vect_permute_load_chain, vect_shift_permute_load_chain)
10244         (vect_transform_grouped_load)
10245         (vect_record_grouped_load_vectors): Likewise.
10246         * tree-vect-loop.c (vect_fixup_reduc_chain)
10247         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10248         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10249         (vectorizable_reduction, vectorizable_induction)
10250         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10251         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10252         (vect_get_load_store_mask): Likewise.
10253         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10254         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10255         * tree-vect-stmts.c (vect_mark_relevant)
10256         (is_simple_and_all_uses_invariant)
10257         (exist_non_indexing_operands_for_use_p, process_use)
10258         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10259         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10260         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10261         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10262         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10263         (get_negative_load_store_type, get_load_store_type)
10264         (vect_check_load_store_mask, vect_check_store_rhs)
10265         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10266         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10267         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10268         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10269         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10270         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10271         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10272         (supportable_widening_operation): Likewise.
10274 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10276         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10277         a stmt_vec_info instead of a gcall.
10278         (vect_check_gather_scatter): Update call accordingly.
10279         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10280         of a gphi.
10281         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10282         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10283         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10284         instead of a gimple stmt.
10285         (vect_transform_loop): Update calls accordingly.
10286         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10287         stmt_vec_infos instead of gimple stmts.
10288         (vect_analyze_slp_instance): Update use accordingly.
10289         * tree-vect-stmts.c (read_vector_array, write_vector_array)
10290         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10291         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10292         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10293         (vect_gen_widened_results_half, vect_get_loop_based_defs)
10294         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10295         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10296         down to subroutines.
10298 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10300         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10301         of the worklist from a vector of gimple stmts to a vector of
10302         stmt_vec_infos.
10303         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10304         (vect_mark_stmts_to_be_vectorized): Likewise
10306 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10308         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10309         statement before passing it to vect_analyze_stmt.
10310         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10311         the chain of phi vector definitions.  Track the exit phi via its
10312         stmt_vec_info.
10313         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10314         STMT_VINFO_REDUC_DEF.
10315         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10316         stmt_vec_infos to handle the statement chains.
10317         (vect_get_slp_defs): Record the first statement in the node
10318         using a stmt_vec_info.
10319         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10320         statements here and pass their stmt_vec_info down to subroutines.
10321         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10322         down to vect_finish_stmt_generation.
10323         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10324         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10325         stmt_vec_infos to subroutines.
10326         (vect_remove_stores): Use stmt_vec_infos to handle the statement
10327         chains.
10329 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10331         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10332         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10333         (vect_permute_store_chain, vect_setup_realignment)
10334         (vect_permute_load_chain, vect_shift_permute_load_chain)
10335         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10336         stmts internally, and when passing values to other vectorizer routines.
10337         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10338         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10339         (vect_analyze_loop_operations, get_initial_def_for_reduction)
10340         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10341         (vectorizable_reduction, vectorizable_induction)
10342         (vectorizable_live_operation, vect_transform_loop_stmt)
10343         (vect_transform_loop): Likewise.
10344         * tree-vect-patterns.c (vect_reassociating_reduction_p)
10345         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10346         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10347         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10348         (vect_slp_analyze_node_operations_1): Likewise.
10349         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10350         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10351         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10352         (vect_finish_stmt_generation_1, get_group_load_store_type)
10353         (get_load_store_type, vect_build_gather_load_calls)
10354         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10355         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10356         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10357         (vectorizable_store, vectorizable_load, vectorizable_condition)
10358         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10359         (supportable_widening_operation): Likewise.
10360         (vect_get_vector_types_for_stmt): Likewise.
10361         * tree-vectorizer.h (vect_dr_behavior): Likewise.
10363 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10365         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10366         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10367         (vect_permute_store_chain, vect_permute_load_chain)
10368         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10369         repeated stmt_vec_info lookups.
10370         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10371         (vect_update_ivs_after_vectorizer): Likewise.
10372         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10373         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10374         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10375         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10376         (vect_bb_slp_scalar_cost): Likewise.
10377         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10379 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10381         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10382         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10383         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10384         to get gassigns and gcalls, rather than operating on generc gimple
10385         stmts.
10386         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10387         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10388         (vectorizable_load, vect_analyze_stmt): Likewise.
10389         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10391 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10393         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10394         return stmt_vec_infos rather than gimple stmts.  Do not accept
10395         null arguments.
10396         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10397         of a gimple stmt.
10398         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10399         Update use of get_later_stmt.
10400         (vect_get_constant_vectors): Update call accordingly.
10401         (vect_schedule_slp_instance): Likewise
10402         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10403         (vect_slp_analyze_instance_dependence): Likewise.
10404         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10406 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10408         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10409         (stmt_info_for_cost::stmt_info): ...this new field.
10410         (add_stmt_costs): Update accordingly.
10411         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10412         (vect_get_known_peeling_cost): Likewise.
10413         (vect_estimate_min_profitable_iters): Likewise.
10414         * tree-vect-stmts.c (record_stmt_cost): Likewise.
10416 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10418         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10419         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10420         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10421         accordingly.
10422         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10424 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10426         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10427         a gimple stmt to a stmt_vec_info.
10428         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10430 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10432         * tree-vectorizer.h (vec_info::grouped_stores): Change from
10433         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10434         (_loop_vec_info::reduction_chains): Likewise.
10435         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10436         accordingly.
10437         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10439 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10441         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10442         a gimple stmt to a stmt_vec_info.
10443         (_stmt_vec_info::next_element): Likewise.
10444         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10445         (vect_slp_analyze_and_verify_node_alignment)
10446         (vect_analyze_group_access_1, vect_analyze_group_access)
10447         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10448         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10449         (vect_supportable_dr_alignment): Update accordingly.
10450         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10451         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10452         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10453         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10454         * tree-vect-slp.c (vect_build_slp_tree_1)
10455         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10456         (vect_split_slp_store_group, vect_analyze_slp_instance)
10457         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10458         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10459         (get_group_load_store_type, get_load_store_type)
10460         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10461         (vect_transform_stmt, vect_remove_stores): Likewise.
10463 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10465         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10466         than a gimple stmt.
10467         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10468         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10469         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10470         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10471         (vector_alignment_reachable_p, vect_get_data_access_cost)
10472         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10473         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10474         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10475         (vect_analyze_group_access_1, vect_analyze_group_access)
10476         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10477         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10478         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10479         result of vect_dr_stmt and use the stmt_vec_info instead of
10480         the associated gimple stmt.
10481         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10482         (vect_gen_prolog_loop_niters): Likewise.
10483         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10485 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10487         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10488         to a vec<stmt_vec_info>.
10489         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10490         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10491         vec<stmt_vec_info>.
10492         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10493         to a vec<stmt_vec_info>.
10494         (bst_traits::value_type, bst_traits::value_type): Likewise.
10495         (bst_traits::hash): Update accordingly.
10496         (vect_get_and_check_slp_defs): Change the stmts parameter from
10497         a vec<gimple *> to a vec<stmt_vec_info>.
10498         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10499         (vect_build_slp_tree): Likewise.
10500         (vect_build_slp_tree_2): Likewise.  Update uses of
10501         SLP_TREE_SCALAR_STMTS.
10502         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10503         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10504         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10505         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10506         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10507         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10508         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10509         (vect_get_constant_vectors, vect_get_slp_defs)
10510         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10511         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10512         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10513         instead of gimple stmts.
10514         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10515         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10516         (vect_slp_analyze_instance_dependence): Update uses of
10517         SLP_TREE_SCALAR_STMTS.
10518         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10519         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10520         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10521         (get_initial_defs_for_reduction): Likewise.
10522         (vect_create_epilog_for_reduction): Likewise.
10523         (vectorize_fold_left_reduction): Likewise.
10524         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10525         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
10526         (can_vectorize_live_stmts): Likewise.
10528 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10530         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
10531         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10532         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
10533         than gimple stmts.
10534         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
10535         of a gimple stmt.
10536         (gather_scalar_reductions): Update after above interface changes.
10537         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
10538         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
10539         than gimple stmts.
10540         (vect_force_simple_reduction): Likewise.
10541         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
10542         LOOP_VINFO_REDUCTIONS.
10543         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10545 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10547         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
10548         a gimple stmt to a stmt_vec_info.
10549         * tree-vect-loop.c (vect_active_double_reduction_p)
10550         (vect_force_simple_reduction, vectorizable_reduction): Update
10551         accordingly.
10553 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10555         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
10556         vec<gimple *> to a vec<stmt_vec_info>.
10557         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
10558         the reduction_phis argument from a vec<gimple *> to a
10559         vec<stmt_vec_info>.
10560         (vectorizable_reduction): Likewise the phis local variable that
10561         is passed to vect_create_epilog_for_reduction.  Update for new type
10562         of SLP_TREE_VEC_STMTS.
10563         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
10564         (vectorizable_live_operation): Likewise.
10565         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
10566         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
10568 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10570         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
10571         a gimple stmt to a stmt_vec_info.
10572         (vectorizable_condition, vectorizable_live_operation)
10573         (vectorizable_reduction, vectorizable_induction): Pass back the
10574         vectorized statement as a stmt_vec_info.
10575         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
10576         use of STMT_VINFO_VEC_STMT.
10577         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
10578         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
10579         as stmt_vec_infos rather than gimple stmts.
10580         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
10581         to a stmt_vec_info.
10582         (vectorizable_live_operation): Likewise.
10583         (vectorizable_reduction, vectorizable_induction): Likewise,
10584         updating use of STMT_VINFO_VEC_STMT.
10585         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
10586         of STMT_VINFO_VEC_STMT.
10587         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
10588         (vectorizable_simd_clone_call, vectorizable_conversion)
10589         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10590         (vectorizable_store, vectorizable_load, vectorizable_condition)
10591         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
10592         from a gimple stmt to a stmt_vec_info.
10593         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
10594         pointer to a stmt_vec_info to the vectorizable_* routines.
10596 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10598         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
10599         a gimple stmt to a stmt_vec_info.
10600         (is_pattern_stmt_p): Update accordingly.
10601         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
10602         (vect_record_grouped_load_vectors): Likewise.
10603         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10604         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
10605         (vect_model_reduction_cost): Likewise.
10606         (vect_create_epilog_for_reduction): Likewise.
10607         (vectorizable_reduction, vectorizable_induction): Likewise.
10608         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10609         Return the stmt_vec_info for the pattern statement.
10610         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
10611         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
10612         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10613         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
10614         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10615         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
10616         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
10617         (free_stmt_vec_info, vect_is_simple_use): Likewise.
10619 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10621         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
10622         (vect_finish_stmt_generation): Likewise.
10623         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10624         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
10625         (vect_build_gather_load_calls): Use the return value of the above
10626         functions instead of a separate call to vinfo_for_stmt.  Use narrow
10627         scopes for the input gimple stmt and wider scopes for the associated
10628         stmt_vec_info.  Use vec_info::lookup_def when setting these
10629         stmt_vec_infos from an SSA_NAME definition.
10630         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10631         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10632         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10633         (vectorizable_store, vectorizable_load, vectorizable_condition)
10634         (vectorizable_comparison): Likewise.
10635         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
10636         (vectorizable_reduction): Likewise.
10638 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10640         * tree-vectorizer.h (vect_is_simple_use): Add an optional
10641         stmt_vec_info * parameter before the optional gimple **.
10642         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10643         (process_use, vect_get_vec_def_for_operand_1): Update callers.
10644         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
10645         * tree-vect-loop.c (vectorizable_reduction): Likewise.
10646         (vectorizable_live_operation): Likewise.
10647         * tree-vect-patterns.c (type_conversion_p): Likewise.
10648         (vect_look_through_possible_promotion): Likewise.
10649         (vect_recog_rotate_pattern): Likewise.
10650         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10652 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10654         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
10655         a typedef to a wrapper class.
10656         (NULL_STMT_VEC_INFO): New macro.
10657         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
10658         (stmt_vec_info::operator*): New function.
10659         (stmt_vec_info::operator gimple *): Likewise.
10660         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
10661         (add_stmt_costs): Likewise.
10662         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
10663         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10664         (vect_get_known_peeling_cost): Likewise.
10665         (vect_estimate_min_profitable_iters): Likewise.
10666         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10667         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
10668         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10669         (vectorizable_store, free_stmt_vec_infos): Likewise.
10670         (new_stmt_vec_info): Change return type of xcalloc to
10671         _stmt_vec_info *.
10673 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10675         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
10676         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
10677         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
10678         a single_imm_use-based sequence.
10679         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
10681 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10683         * tree-vectorizer.h (vec_info::lookup_def): Declare.
10684         * tree-vectorizer.c (vec_info::lookup_def): New function.
10685         * tree-vect-patterns.c (vect_get_internal_def): Use it.
10686         (vect_widened_op_tree): Likewise.
10687         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10688         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
10689         (vectorizable_reduction): Likewise.
10690         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
10691         of a gimple *.
10692         (vect_is_slp_reduction): Update calls accordingly.  Use
10693         vec_info::lookup_def.
10694         (vect_is_simple_reduction): Likewise
10695         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
10697 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10699         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
10700         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
10701         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
10702         of vinfo_for_stmt.
10703         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
10704         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
10705         (vect_update_vf_for_slp, vect_analyze_loop_operations)
10706         (vect_is_slp_reduction, vectorizable_induction)
10707         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
10708         * tree-vect-patterns.c (vect_init_pattern_stmt):
10709         (vect_determine_min_output_precision_1, vect_determine_precisions)
10710         (vect_pattern_recog): Likewise.
10711         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
10712         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
10713         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
10714         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10715         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
10716         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
10717         info field from a loop to a loop_vec_info.
10719 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10721         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
10722         (vec_info::add_stmt): Declare.
10723         * tree-vectorizer.c (vec_info::add_stmt): New function.
10724         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
10725         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
10726         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10727         (vectorizable_induction): Likewise.
10728         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
10729         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10730         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
10731         (vectorizable_load): Likewise.
10732         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10733         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
10734         (vect_recog_gather_scatter_pattern): Likewise.
10735         (append_pattern_def_seq): Likewise.  Remove a check that is
10736         performed by add_stmt itself.
10738 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10740         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
10741         which make_ssa_name was called with new_stmt before new_stmt
10742         had been created.
10744 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10746         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
10747         split out from...
10748         (vect_is_slp_reduction): ...here...
10749         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
10750         that are already known to be false.
10752 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10754         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
10755         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
10756         STMT_VINFO_NUM_SLP_USES when it's true.
10757         (vect_free_slp_instance): Add a final_p parameter and pass it to
10758         vect_free_slp_tree.
10759         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
10760         (vect_analyze_slp_instance): Likewise.
10761         (vect_slp_analyze_operations): Likewise.
10762         (vect_slp_analyze_bb_1): Likewise.
10763         * tree-vectorizer.c (vec_info): Likewise.
10764         * tree-vect-loop.c (vect_transform_loop): Likewise.
10766 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10768         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10769         function is not called for second and subsequent members of
10770         a reduction group.
10772 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10774         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
10775         cases for nested loops from here to ...
10776         (vect_create_epilog_for_reduction): ...here.  Only call
10777         vect_is_simple_use for inner-loop reductions.
10779 2018-07-31  Martin Liska  <mliska@suse.cz>
10781         PR gcov-profile/85338
10782         PR gcov-profile/85350
10783         PR gcov-profile/85372
10784         * profile.c (struct location_triplet): New.
10785         (struct location_triplet_hash): Likewise.
10786         (output_location): Do not output a BB that
10787         is already recorded for a line.
10788         (branch_prob): Use streamed_locations.
10790 2018-07-31  Martin Liska  <mliska@suse.cz>
10792         PR gcov-profile/85370
10793         * coverage.c (coverage_begin_function): Do not mark target
10794         clones as artificial functions.
10796 2018-07-31  Martin Liska  <mliska@suse.cz>
10798         PR gcov-profile/83813
10799         PR gcov-profile/84758
10800         PR gcov-profile/85217
10801         PR gcov-profile/85332
10802         * profile.c (branch_prob): Do not record GOTO expressions
10803         for GIMPLE statements which locations are already streamed.
10805 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10807         * gcc.c (handle_spec_function): Accept a soft_matched_part
10808         argument, as do_spec_1.  Pass it down to ...
10809         (eval_spec_function): Accept a soft_matched_part argument,
10810         and pass it down to ...
10811         (do_spec_2): Accept a soft_matched_part argument, and pass
10812         it down to do_spec_1.
10813         (do_spec_1): Pass soft_matched_part to handle_spec_function.
10814         (handle_braces): Update call to handle_spec_function.
10815         (driver::set_up_specs): Update calls to do_spec_2.
10816         (compare_debug_dump_opt_spec_function): Likewise.
10817         (compare_debug_self_opt_spec_function): Likewise.
10819 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10821         * common.opt (nolibc): New option.
10822         * doc/invoke.texi (Link Options): Document it.
10823         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
10824         * config/alpha/linux.h: Likewise.
10825         * config/arc/elf.h: Likewise.
10826         * config/arm/uclinux-elf.h: Likewise.
10827         * config/arm/unknown-elf.h: Likewise.
10828         * config/avr/avrlibc.h: Likewise.
10829         * config/bfin/bfin.h: Likewise.
10830         * config/bfin/linux.h: Likewise.
10831         * config/bfin/uclinux.h: Likewise.
10832         * config/darwin.h: Likewise.
10833         * config/darwin10.h: Likewise.
10834         * config/darwin12.h: Likewise.
10835         * config/gnu-user.h: Likewise.
10836         * config/lm32/uclinux-elf.h: Likewise.
10837         * config/pa/pa-hpux11.h: Likewise.
10838         * config/pa/pa64-hpux.h: Likewise.
10839         * config/sparc/sparc.h: Likewise.
10841 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10843         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
10844         undefined variables.
10846 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10848         PR target/86640
10849         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
10850         instead of GEN_INT.
10852 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10854         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10855         terminated string literal.
10857 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10859         PR rtl-optimization/85160
10860         * combine.c (is_just_move): New function.
10861         (try_combine): Allow combining two instructions into two if neither of
10862         the original instructions was a move.
10864 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
10866         PR target/86673
10867         * doc/extend.texi (Global Register Variables): Discourage use of type
10868         qualifiers.
10869         (Local Register Variables): Likewise.
10871 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
10873         PR tree-optimization/86506
10874         * hwint.h (ceil_log2): Resync with hwint.c implementation.
10876 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
10878         PR target/86547
10879         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
10880         hard_regno, make sure no insn between `from` and `to` clobbers it.
10882 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
10883             Tom de Vries  <tdevries@suse.de>
10885         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
10886         (PTX_DEFAULT_RUNTIME_DIM): ... this.
10887         (nvptx_goacc_validate_dims): Set default worker and gang dims to
10888         PTX_DEFAULT_RUNTIME_DIM.
10889         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
10891 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
10893         * config/pa/pa.c (pa_output_addr_vec): Align address table.
10894         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
10895         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
10897 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
10899         * config/rs6000/constraints.md (wG constraint): Delete, no longer
10900         used.
10901         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
10902         predicate to reflect toc fusion has been deleted.
10903         (toc_fusion_mem_raw): Delete, no longer used.
10904         (toc_fusion_mem_wrapped): Likewise.
10905         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
10906         fusion mask bit.
10907         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
10908         Delete, no longer used.
10909         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
10910         meant to be used for toc fusion.
10911         (rs6000_debug_print_mode): Delete toc fusion debugging.
10912         (rs6000_debug_reg_global): Likewise.
10913         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
10914         fusion and secondary reload support that were never used.
10915         (rs6000_option_override_internal): Delete TOC fusion, that was only
10916         partially defined, and it did not work unless you also used the
10917         -mcmodel= switch.
10918         (rs6000_legitimate_address_p): Delete TOC fusion support.
10919         (rs6000_opt_masks): Likewise.
10920         (fusion_wrap_memory_address): Delete function, no longer used.
10921         (fusion_split_address); Delete TOC fusion support.
10922         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
10923         longer used with toc fusion being deleted.
10924         (TARGET_TOC_FUSION_FP): Likewise.
10925         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
10926         UNSPEC.
10927         (toc fusion spliter): Delete TOC fusion support.
10928         (toc_fusionload_<mode>): Likewise.
10929         (toc_fusionload_di): Likewise.
10930         (fusion_gpr_load_<mode>): Delete generator function, this insn no
10931         longer needs to be named.  Rename predicate to delete TOC fusion.
10932         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10933         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10934         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10935         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10936         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
10938 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10940         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
10941         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
10942         __int128 in built-in function prototypes.
10943         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
10944         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
10946 2018-07-27  Martin Sebor  <msebor@redhat.com>
10948         PR tree-optimization/86696
10949         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
10950         types, including enums.
10951         (handle_char_store): Be prepared for the above function to fail.
10953 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
10955         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
10956         when optimization level is lower than 2 or optimize for size.
10958 2018-07-26  Martin Sebor  <msebor@redhat.com>
10960         PR tree-optimization/86043
10961         PR tree-optimization/86042
10962         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
10963         (get_string_cst_length): Rename...
10964         (get_min_string_length): ...to this.  Add argument.
10965         (handle_char_store): Extend to handle multi-character stores by
10966         MEM_REF.
10967         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
10968         * tree.h (initializer_zerop): Add argument.
10970 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
10972         PR middle-end/86660
10973         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
10974         declare target to variables if they have always,{to,from,tofrom} map
10975         kinds.
10977 2018-07-26  Martin Liska  <mliska@suse.cz>
10979         PR lto/86548
10980         * lto-wrapper.c: Add linker_output as prefix
10981         for ltrans_output_file.
10983 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
10985         PR rtl-optimization/85805
10986         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
10987         value for hard registers if that was written in the same mode.
10989 2018-07-26  Martin Liska  <mliska@suse.cz>
10991         PR gcov-profile/86536
10992         * gcov.c (format_gcov): Use printf format %.*f directly
10993         and do not handle special values.
10995 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10997         * common/config/arc/arc-common.c (arc_option_optimization_table):
10998         Update default optimizations for size.
11000 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11002         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11003         * config/arc/constraints.md (CfZ): New constraint.
11004         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11005         (subsf3_fpu): Likewise.
11006         (cmpsf_fpu): Likewise.
11007         (cmpsf_fpu_uneq): Likewise.
11009 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11011         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11012         accesses as well.
11013         (arc_is_uncached_mem_p): uncached applies to both the variable and
11014         the pointer.
11016 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11018         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11019         register names.
11021 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11023         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11024         field "m_scopes" from vec to auto_vec.
11026 2018-07-25  Martin Liska  <mliska@suse.cz>
11028         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11029         return type.
11031 2018-07-25  Richard Biener  <rguenther@suse.de>
11033         PR debug/86654
11034         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11035         special wrt context_die late.
11036         (gen_subprogram_die): Re-use DIEs in local scope.
11038 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11040         PR tree-optimization/86644
11041         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11043 2018-07-25  Martin Liska  <mliska@suse.cz>
11045         PR middle-end/86645
11046         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11047         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11049 2018-07-25  Martin Liska  <mliska@suse.cz>
11051         PR sanitizer/79635
11052         * params.def: Explain ASan abbreviation and provide
11053         a documentation link.
11055 2018-07-24  Martin Sebor  <msebor@redhat.com>
11057         PR tree-optimization/86622
11058         PR tree-optimization/86532
11059         * builtins.h (string_length): Declare.
11060         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11061         (check_access): Be prepared for non-constant length ranges.
11062         (string_length): Make extern.
11063         * expr.c (string_constant): Only handle the minor non-constant
11064         array index.  Use string_constant to compute the length of
11065         a generic string constant.
11067 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11069         PR tree-optimization/86618
11070         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11071         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11073 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11075         PR tree-optimization/86636
11076         * json.cc (json::object::set): Fix comment.  Add assertions.
11077         (json::array::append): Move here from json.h.  Add comment and an
11078         assertion.
11079         (json::string::string): Likewise.
11080         * json.h (json::array::append): Move to json.cc.
11081         (json::string::string): Likewise.
11082         * optinfo-emit-json.cc
11083         (optrecord_json_writer::impl_location_to_json): Assert that we
11084         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11085         wrapper around it.  Expand the location once, rather than three
11086         times.
11087         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11088         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11089         wrappers.
11090         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11091         places.  Fix some overlong lines.
11093 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11095         * config/aarch64/aarch64-simd.md
11096         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11097         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11098         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11099         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11100         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11101         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11102         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11103         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11104         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11106 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11108         PR middle-end/86627
11109         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11110         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11111         and abs_d == d, do the power of two handling if profitable.
11113 2018-07-24  Richard Biener  <rguenther@suse.de>
11115         * match.pd: Add BIT_FIELD_REF canonicalizations.
11117 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11119         PR c/86617
11120         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11122 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11124         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11125         terminated STRING_CST object.
11127 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11129         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11131 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11133         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11134         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11135         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11136         the elements into a register.
11137         (rs6000_split_v4si_init_di_reg): Delete.
11138         (rs6000_split_v4si_init): Delete.
11139         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11140         (vsx_init_v4si): Rewrite as a define_expand.
11142 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11144         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11145         zero_extend argument from memory): New.
11147 2018-07-22  Martin Sebor  <msebor@redhat.com>
11149         PR bootstrap/86621
11150         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11151         diagnosing calls with unknown arguments unless -Walloca-larger-than
11152         is restricted to less than PTRDIFF_MAX bytes.
11154 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
11156         * doc/gcov.texi (Invoking Gcov): Editorial changes.
11158 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11160         * pretty-print.c (text_info::set_location): Remove redundant
11161         "line_table" parameter from call to rich_location::set_range.
11163 2018-07-20  Martin Sebor  <msebor@redhat.com>
11165         PR middle-end/82063
11166         * builtins.c (expand_builtin_alloca): Adjust.
11167         * calls.c (alloc_max_size): Simplify.
11168         * cgraphunit.c (cgraph_node::expand): Adjust.
11169         * common.opt (larger_than_size, warn_frame_larger_than): Remove
11170         variables.
11171         (frame_larger_than_size): Same.
11172         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11173         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11174         Initialize.
11175         * doc/invoke.texi (GCC Command Options): Document option arguments.
11176         Explain byte-size arguments and suffixes.
11177         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11178         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11179         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11180         * doc/options.texi (UInteger): Expand.
11181         (Host_Wide_Int, ByteSize): Document new properties.
11182         * final.c (final_start_function_1): Include sizes in an error message.
11183         * function.c (frame_offset_overflow): Same.
11184         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11185         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11186         Diagnose unbounded alloca calls only for limits of less than
11187         PTRDIFF_MAX.
11188         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
11189         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
11190         for alloca(0).
11191         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
11192         only for limits of less than PTRDIFF_MAX.
11193         * langhooks-def.h (lhd_handle_option): Change function argument
11194         to HOST_WIDE_INT.
11195         * langhooks.c (lhd_handle_option): Same.
11196         * langhooks.h (handle_option): Same.
11197         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11198         ByteSize flags.
11199         (var_type, var_type_struct): Same.
11200         (var_set): Handle ByteSize flag.
11201         * optc-gen.awk: Add comments to output to ease debugging.  Make
11202         use of HOST_WIDE_INT where appropriate.
11203         * opts-gen-save.awk:  Use %lx to format unsigned long.
11204         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11205         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11206         arguments.  Parse bytes-size suffixes.
11207         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11208         (enum_value_to_arg): Same.
11209         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
11210         (handle_option): Adjust.
11211         (generate_option): Change function argument to HOST_WIDE_INT.
11212         (cmdline_handle_error): Adjust.
11213         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11214         (set_option): Change function argument to HOST_WIDE_INT.
11215         (option_enabled): Handle cl_host_wide_int.
11216         (get_option_state): Handle CLVC_SIZE.
11217         (control_warning_option): Same.
11218         * opts.c (common_handle_option): Change function argument to
11219         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11220         OPT_Wvla_larger_than_.
11221         * opts.h (enum cl_var_type): Add an enumerator.
11222         * stor-layout.c (layout_decl): Print a more meaningful warning.
11223         * toplev.c (output_stack_usage): Adjust.
11225 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11227         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11228         call to inline_expand_builtin_string_cmp.
11229         (expand_builtin_strcmp): Likewise.
11230         (expand_builtin_strncmp): Likewise.
11231         (inline_string_cmp): Delete the last parameter, change char_type_node
11232         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11233         two operands.
11234         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11235         the inlining expansion on target where the type of the call has same or
11236         narrower precision than unsigned char.
11238 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11240         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11241         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11242         * common.opt (fsave-optimization-record): New option.
11243         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11244         * doc/invoke.texi (-fsave-optimization-record): New option.
11245         * dumpfile.c: Include "optinfo-emit-json.h".
11246         (struct kv_pair): Move to coretypes.h.
11247         (optgroup_options): Make non-static.
11248         (dump_context::end_scope): Call
11249         optimization_records_maybe_pop_dump_scope.
11250         * dumpfile.h (optgroup_options): New decl.
11251         * json.cc: New file.
11252         * json.h: New file.
11253         * optinfo-emit-json.cc: New file.
11254         * optinfo-emit-json.h: New file.
11255         * optinfo.cc: Include "optinfo-emit-json.h".
11256         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11257         (optinfo_enabled_p): Check optimization_records_enabled_p.
11258         (optinfo_wants_inlining_info_p): Likewise.
11259         * optinfo.h: Update comment.
11260         * profile-count.c (profile_quality_as_string): New function.
11261         * profile-count.h (profile_quality_as_string): New decl.
11262         (profile_count::quality): New accessor.
11263         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11264         and optinfo_emit_json_cc_tests.
11265         * selftest.h (selftest::json_cc_tests): New decl.
11266         (selftest::optinfo_emit_json_cc_tests): New decl.
11267         * toplev.c: Include "optinfo-emit-json.h".
11268         (compile_file): Call optimization_records_finish.
11269         (do_compile): Call optimization_records_start.
11270         * tree-ssa-live.c: Include optinfo.h.
11271         (remove_unused_scope_block_p): Retain inlining information if
11272         optinfo_wants_inlining_info_p returns true.
11274 2018-07-20  Richard Biener  <rguenther@suse.de>
11276         PR debug/86585
11277         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11278         to cover -flto-partition=none.
11280 2018-07-20  Martin Liska  <mliska@suse.cz>
11282         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11283         (get_decl_source_range): Remove unused function.
11285 2018-07-20  Richard Biener  <rguenther@suse.de>
11287         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11288         (struct vn_phi_s): Likewise.
11289         (struct vn_reference_s): Likewise.
11290         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11291         for searching the slot of an entry known to be in the hash itself.
11292         (vn_phi_hasher::equal): Likewise.
11293         (vn_reference_hasher::equal): Likewise.
11294         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11295         globals.
11296         (optimistic_info, current_info): Remove, keeping only valid_info.
11297         (vn_reference_lookup_1): Remove fallback lookup.
11298         (vn_reference_lookup_2): Likewise.
11299         (vn_nary_op_lookup_1): Likewise.
11300         (vn_phi_lookup): Likewise.
11301         (vn_nary_build_or_lookup_1): Make sure to not chain the built
11302         hash element.
11303         (vn_reference_insert): Adjust, chain the inserted hash element
11304         at last_inserted_ref.
11305         (vn_reference_insert_pieces): Likewise.
11306         (visit_reference_op_call): Likewise.
11307         (vn_nary_op_insert_into): Chain the inserted hash element at
11308         last_inserted_nary.
11309         (vn_nary_op_insert_pieces): Adjust.
11310         (vn_nary_op_insert): Likewise.
11311         (vn_nary_op_insert_stmt): Likewise.
11312         (vn_phi_insert): Adjust, chain the inserted hash element at
11313         last_inserted_phi.
11314         (process_scc): Remove clearing and copying the optimistic
11315         table.  Instead remove elements inserted during an optimistic
11316         iteration from the single table we maintain.
11317         (init_scc_vn): Adjust.
11318         (free_scc_vn): Likewise.
11319         (sccvn_dom_walker::record_cond): Likewise.
11320         (sccvn_dom_walker::after_dom_children): Likewise.
11322 2018-07-19  Martin Sebor  <msebor@redhat.com>
11324         PR tree-optimization/84047
11325         PR tree-optimization/83776
11326         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11327         (check_array_bounds): Call it.
11329 2018-07-19  Martin Sebor  <msebor@redhat.com>
11331         * align.h (align_flags): Use member initialization.
11333 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
11335         * Makefile.in (OBJS): Add optinfo.o.
11336         * coretypes.h (class symtab_node): New forward decl.
11337         (struct cgraph_node): New forward decl.
11338         (class varpool_node): New forward decl.
11339         * dump-context.h: New file.
11340         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11341         "tree-pass.h".
11342         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11343         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11344         (set_alt_dump_file): Likewise.
11345         (dump_context::~dump_context): New dtor.
11346         (dump_gimple_stmt): Move implementation to...
11347         (dump_context::dump_gimple_stmt): ...this new member function.
11348         Add the stmt to any pending optinfo, creating one if need be.
11349         (dump_gimple_stmt_loc): Move implementation to...
11350         (dump_context::dump_gimple_stmt_loc): ...this new member function.
11351         Start a new optinfo and add the stmt to it.
11352         (dump_gimple_expr): Move implementation to...
11353         (dump_context::dump_gimple_expr): ...this new member function.
11354         Add the stmt to any pending optinfo, creating one if need be.
11355         (dump_gimple_expr_loc): Move implementation to...
11356         (dump_context::dump_gimple_expr_loc): ...this new member function.
11357         Start a new optinfo and add the stmt to it.
11358         (dump_generic_expr): Move implementation to...
11359         (dump_context::dump_generic_expr): ...this new member function.
11360         Add the tree to any pending optinfo, creating one if need be.
11361         (dump_generic_expr_loc): Move implementation to...
11362         (dump_context::dump_generic_expr_loc): ...this new member
11363         function.  Add the tree to any pending optinfo, creating one if
11364         need be.
11365         (dump_printf): Move implementation to...
11366         (dump_context::dump_printf_va): ...this new member function.  Add
11367         the text to any pending optinfo, creating one if need be.
11368         (dump_printf_loc): Move implementation to...
11369         (dump_context::dump_printf_loc_va): ...this new member function.
11370         Start a new optinfo and add the stmt to it.
11371         (dump_dec): Move implementation to...
11372         (dump_context::dump_dec): ...this new member function.  Add the
11373         value to any pending optinfo, creating one if need be.
11374         (dump_context::dump_symtab_node): New member function.
11375         (dump_context::get_scope_depth): New member function.
11376         (dump_context::begin_scope): New member function.
11377         (dump_context::end_scope): New member function.
11378         (dump_context::ensure_pending_optinfo): New member function.
11379         (dump_context::begin_next_optinfo): New member function.
11380         (dump_context::end_any_optinfo): New member function.
11381         (dump_context::s_current): New global.
11382         (dump_context::s_default): New global.
11383         (dump_scope_depth): Delete global.
11384         (dumpfile_ensure_any_optinfo_are_flushed): New function.
11385         (dump_symtab_node): New function.
11386         (get_dump_scope_depth): Reimplement in terms of dump_context.
11387         (dump_begin_scope): Likewise.
11388         (dump_end_scope): Likewise.
11389         (selftest::temp_dump_context::temp_dump_context): New ctor.
11390         (selftest::temp_dump_context::~temp_dump_context): New dtor.
11391         (selftest::verify_item): New function.
11392         (ASSERT_IS_TEXT): New macro.
11393         (ASSERT_IS_TREE): New macro.
11394         (ASSERT_IS_GIMPLE): New macro.
11395         (selftest::test_capture_of_dump_calls): New test.
11396         (selftest::dumpfile_c_tests): Call it.
11397         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11398         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11399         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11400         descriptive comment.
11401         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11402         (dump_node, dump_bb): Move these unrelated decls.
11403         (class dump_manager): Add leading comment.
11404         * optinfo.cc: New file.
11405         * optinfo.h: New file.
11407 2018-07-19  Michael Collison  <michael.collison@arm.com>
11408             Richard Henderson <rth@redhat.com>
11410         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11411         (subti): Handle op1 zero.
11412         (subvti4, usub4ti4): New.
11413         (*sub<GPI>3_compare1_imm): New.
11414         (sub<GPI>3_carryinCV): New.
11415         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11416         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11418 2018-07-19  Michael Collison  <michael.collison@arm.com>
11419             Richard Henderson <rth@redhat.com>
11421         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11422         (addti3): Create simpler code if low part is already known to be 0.
11423         (addvti4, uaddvti4): New.
11424         (*add<GPI>3_compareC_cconly_imm): New.
11425         (*add<GPI>3_compareC_cconly): New.
11426         (*add<GPI>3_compareC_imm): New.
11427         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11428         handle constants within this pattern..
11429         (*add<GPI>3_compareV_cconly_imm): New.
11430         (*add<GPI>3_compareV_cconly): New.
11431         (*add<GPI>3_compareV_imm): New.
11432         (add<GPI>3_compareV): New.
11433         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11434         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11435         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11436         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11437         with 'comparison' operator.
11438         (*add<GPI>3_compareV_cconly_imm): Ditto.
11439         (*add<GPI>3_compareV_cconly): Ditto.
11440         (*add<GPI>3_compareV_imm): Ditto.
11441         (add<GPI>3_compareV): Ditto.
11442         (add<mode>3_carryinC): Ditto.
11443         (*add<mode>3_carryinC_zero): Ditto.
11444         (*add<mode>3_carryinC): Ditto.
11445         (add<mode>3_carryinV): Ditto.
11446         (*add<mode>3_carryinV_zero): Ditto.
11447         (*add<mode>3_carryinV): Ditto.
11449 2018-07-19  Michael Collison  <michael.collison@arm.com>
11450             Richard Henderson <rth@redhat.com>
11452         * config/aarch64/aarch64-modes.def (CC_V): New.
11453         * config/aarch64/aarch64-protos.h
11454         (aarch64_addti_scratch_regs): Declare
11455         (aarch64_subvti_scratch_regs): Declare.
11456         (aarch64_expand_subvti): Declare.
11457         (aarch64_gen_unlikely_cbranch): Declare
11458         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11459         for signed overflow using CC_Vmode.
11460         (aarch64_get_condition_code_1): Handle CC_Vmode.
11461         (aarch64_gen_unlikely_cbranch): New function.
11462         (aarch64_addti_scratch_regs): New function.
11463         (aarch64_subvti_scratch_regs): New function.
11464         (aarch64_expand_subvti): New function.
11466 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11468         * config/aarch64/aarch64-option-extensions.def: New entry for profile
11469         extension.
11470         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11471         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11472         extension.
11474 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11476         PR target/83009
11477         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11478         address check not strict.
11480 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11482         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11483         Umq with Umn.
11484         (store_pair_lanes<mode>): Likewise.
11485         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11486         enum value 'ADDR_QUERY_LDP_STP_N'.
11487         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11488         (aarch64_print_address_internal): Add declaration.
11489         (aarch64_print_ldpstp_address): Remove.
11490         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11491         (aarch64_print_operand): Change printing of 'y'.
11492         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11493         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11494         'true' rather than '1'.
11495         * config/aarch64/constraints.md (Uml): Likewise.
11496         (Uml): Rename to Umn.
11497         (Umq): Remove.
11499 2018-07-19  Richard Biener  <rguenther@suse.de>
11501         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11502         a trailing array.
11503         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11504         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11505         (vn_reference_hasher): Likewise.
11506         (struct vn_tables_s): Remove obstack and alloc-pool members.
11507         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11508         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11509         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11510         (vn_reference_insert_pieces): Likewise.
11511         (alloc_vn_nary_op_noinit): Adjust.
11512         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11513         (vn_phi_eq): Adjust.
11514         (shared_lookup_phiargs): Remove.
11515         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11516         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11517         (visit_reference_op_call): Likewise.
11518         (copy_nary, copy_phi, copy_reference): Remove.
11519         (process_scc): Rewind the obstack when iterating.  Do not
11520         copy the elements to valid_info but just move them from one
11521         hashtable to the other.
11522         (allocate_vn_table): Adjust.
11523         (free_vn_table): Likewise.
11524         (init_scc_vn): Likewise.
11525         (free_scc_vn): Likewise.
11527 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
11529         PR target/86560
11530         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
11531         indirect_return as function type attribute.
11532         (ix86_attribute_table): Change indirect_return to function
11533         type attribute.
11534         * doc/extend.texi: Update indirect_return attribute.
11536 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
11538         * wide-int.h (widest2_int): New.
11539         * gimple-fold.c (arith_overflowed_p): Use it.
11540         * tree.h (widest2_int_cst): New.
11541         * tree-vrp.c (wide_int_binop_overflow): Rename from
11542         vrp_int_const_binop.
11543         Rewrite to work on trees.
11544         (extract_range_from_multiplicative_op_1): Abstract code to...
11545         (wide_int_range_min_max): ...here.
11546         (wide_int_range_cross_product): ...and here.
11547         (extract_range_from_binary_expr_1): Abstract overflow code to...
11548         (wide_int_range_mult_wrapping): ...here.
11549         * tree-vrp.h (wide_int_range_cross_product): New.
11550         (wide_int_range_mult_wrapping): New.
11552 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
11553             Julia Koval  <julia.koval@intel.com>
11555         * config/i386/x86-tune-costs.h (skylake_memcpy,
11556         skylake_memset): Replace rep_prefix with unrolling for size 512.
11558 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
11560         PR middle-end/86544
11561         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
11562         comparision with EQ_EXPR in last stmt.
11564 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11566         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
11567         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
11568         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
11569         previously known as "PowerPC AltiVec Built-in Functions".  Move
11570         some material to new subsubsections "PowerPC AltiVec Built-in
11571         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
11572         ISA 2.07".
11573         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
11574         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
11575         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
11576         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
11578 2018-07-18  Richard Biener  <rguenther@suse.de>
11580         PR tree-optimization/86557
11581         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
11582         EXACT_DIV_EXPR.
11584 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
11586         * config/s390/s390.c (s390_function_profiler): Generate CFI.
11588 2018-07-17  Jeff Law  <law@redhat.com>
11590         * config/arm/arm.c (get_label_padding): Update for recent
11591         changes to label_to_alignment.
11593         PR tree-optimization/86010
11594         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
11596         * config/mips/mips.c (vr4130_align_insns): Update for recent
11597         changes to label_to_alignment.
11599         * config/frv/frv.c (frv_label_align): Update for recent changes
11600         to label_to_alignment.
11602         * config/nios2/nios2.c (nios2_label_align): Update for recent
11603         changes which dropped ALIGN_LABELS_LOG.
11605 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
11607         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
11608         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
11610 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
11612         * config/arc/arc.c (arc_label_align): Use align_labels instead of
11613         deprecated align_labels_log.
11615 2018-07-17  Richard Biener  <rguenther@suse.de>
11617         PR lto/86456
11618         * dwarf2out.c (init_sections_and_labels): Always generate
11619         a debug_line_str_section for early LTO debug.
11620         (dwarf2out_finish): Reset debug_line_str_hash output early.
11621         Bump counter for extra dwarf5 .debug_loc labels to not conflict
11622         with fat LTO part.
11623         (dwarf2out_early_finish): Output debug_line_str.
11625 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11627         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
11628         index register on z196 or later.
11630 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11632         * config/s390/s390.c (s390_default_align): Set default function
11633         alignment to 16.
11634         (s390_override_options_after_change): Call s390_default align.
11635         (s390_option_override_internal): Call s390_default_align.
11636         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
11638 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11640         PR middle-end/86542
11641         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
11642         to _looptemp_ clauses, other than the first two.
11644 2018-07-17  Martin Liska  <mliska@suse.cz>
11646         * opts.c: Do not enable OPT_falign_* for -Os.
11648 2018-07-17  Martin Liska  <mliska@suse.cz>
11650         * align.h (MAX_CODE_ALIGN): New.
11651         (MAX_CODE_ALIGN_VALUE): New.
11652         * common/config/i386/i386-common.c (ix86_handle_option):
11653         (MAX_CODE_ALIGN): Moved to align.h.
11654         * final.c (MAX_CODE_ALIGN): Likewise.
11655         * opts.c (parse_and_check_align_values):
11656         (MAX_CODE_ALIGN): Likewise.
11657         (MAX_CODE_ALIGN_VALUE): Likewise.
11659 2018-07-17  Martin Liska  <mliska@suse.cz>
11661         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
11662         in order to fulfil coding style.
11663         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
11664         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11665         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11666         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
11667         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
11668         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
11669         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
11670         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
11671         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11673 2018-07-17  Martin Liska  <mliska@suse.cz>
11675         * align.h: New file.
11676         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
11677         directly.
11678         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
11679         align_flags of label_to_alignment.
11680         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
11681         align_flags class.
11682         * config/m68k/m68k.c: Do not use removed align_labels_value and
11683         align_loops_value.
11684         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
11685         (LOOP_ALIGN): Likewise.
11686         (LABEL_ALIGN): Likewise.
11687         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11688         Remove not used macro.
11689         (rs6000_loop_align): Change return type to align_flags.
11690         (rs6000_loop_align_max_skip): Remove.
11691         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
11692         Change return type to align_flags.
11693         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11694         Remove not used macro.
11695         (rs6000_loop_align):  Change return type to align_flags.
11696         (rs6000_loop_align_max_skip): Remove.
11697         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
11698         * config/rx/rx-protos.h (rx_align_for_label): Make it
11699         static function.
11700         * config/rx/rx.c (rx_align_for_label): Change return type
11701         to align_flags.
11702         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
11703         macro definitions.
11704         into align_flags class.
11705         (LABEL_ALIGN): Likewise.
11706         (LOOP_ALIGN): Likewise.
11707         * config/s390/s390.c (s390_label_align): Use align_flags
11708         class member.
11709         (s390_asm_output_function_label): Likewise.
11710         * config/sh/sh.c (sh_override_options_after_change):
11711         Use align_flags class directly without macros.
11712         (find_barrier): Likewise.
11713         (barrier_align): Likewise.
11714         (sh_loop_align): Likewise.
11715         * config/spu/spu.c (spu_option_override):
11716         Use align_flags_tuple::get_value instead of removed macros.
11717         (spu_sched_init): Likewise.
11718         * config/spu/spu.h (GTY): Likewise.
11719         * config/visium/visium.c (visium_option_override):
11720         Set "8" as default secondary alignment.
11721         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
11722         in order to guarantee secondary alignment of 8.
11723         * coretypes.h: Include align.h header file.
11724         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
11725         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
11726         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
11727         * doc/tm.texi.in: Likewise.
11728         * final.c (struct label_alignment): Remove not used structure.
11729         (LABEL_ALIGN): Change type to align_flags.
11730         (LOOP_ALIGN): Likewise.
11731         (JUMP_ALIGN): Likewise.
11732         (default_loop_align_max_skip): Remove.
11733         (default_label_align_max_skip): Likewise.
11734         (default_jump_align_max_skip): Likewise.
11735         (default_label_align_after_barrier_max_skip):
11736         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
11737         (LABEL_TO_MAX_SKIP): Remove.
11738         (label_to_alignment): Return align_flags type instead of integer.
11739         (label_to_max_skip): Remove.
11740         (align_fuzz): Use align_flags type.
11741         (compute_alignments): Use align_flags type and use align_flags::max
11742         to combine multiple alignments.
11743         (grow_label_align): Grow vec instead of C array.
11744         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
11745         (shorten_branches):  Use align_flags type and use align_flags::max
11746         to combine multiple alignments.
11747         (final_scan_insn_1): Remove usage of secondary alignment that comes
11748         from label alignment, but instead use proper secondary alignment
11749         which is computed in grow_label_align.
11750         * flags.h (struct align_flags_tuple): Move to align.h.
11751         (struct align_flags): Likewise.
11752         (state_align_loops): Rename to align_loops.
11753         (state_align_jumps): Rename to align_jumps.
11754         (state_align_labels): Rename to align_labels.
11755         (state_align_functions): Rename to align_functions.
11756         (align_loops_log): Remove.
11757         (align_jumps_log): Remove.
11758         (align_labels_log): Remove.
11759         (align_functions_log): Remove.
11760         (align_loops_max_skip): Remove.
11761         (align_jumps_max_skip): Remove.
11762         (align_labels_max_skip): Remove.
11763         (align_functions_max_skip): Remove.
11764         (align_loops_value): Remove.
11765         (align_jumps_value): Remove.
11766         (align_labels_value): Remove.
11767         (align_functions_value): Remove.
11768         * output.h (label_to_alignment): Change return type to align_flags.
11769         (label_to_max_skip): Remove.
11770         * target.def: Remove loop_align_max_skip, label_align_max_skip,
11771         jump_align_max_skip macros.
11772         * targhooks.h (default_loop_align_max_skip): Remove.
11773         (default_label_align_max_skip): Likewise.
11774         (default_jump_align_max_skip): Likewise.
11775         (default_label_align_after_barrier_max_skip): Remove.
11776         * toplev.c (read_log_maxskip): Use ::normalize function.
11777         (parse_N_M): Remove not used argument and also call ::normalize.
11778         (parse_alignment_opts): Do not pass unused arguments.
11779         * varasm.c (assemble_start_function): Use directly align_functions
11780         instead of removed macros.
11781         * system.h: Do not poison removed macros.
11783 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11785         PR middle-end/86539
11786         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
11787         and cond temporaries don't have reference type if iterator has
11788         pointer type.  For init use &for_pre_body instead of pre_p if
11789         for_pre_body is non-empty.
11791 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11793         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
11794         double-double modes to SFmode directly directly.
11795         (trunc<mode>sf2_fprs): Delete.
11797 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11799         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
11800         for conversions between IFmode and the decimal floating point modes.
11801         (init_float128_ieee): Use the correct names for conversions between
11802         KFmode and the decimal floating point modes.
11804 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11806         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
11807         for the conversions between TDmode and IFmode.
11808         (init_float128_ieee): Use more correct names for the conversions
11809         between TDmode and KFmode.
11811 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
11813         PR tree-optimization/86526
11814         * builtins.c (expand_builtin_memcmp): Formatting fixes.
11815         (inline_expand_builtin_string_cmp): Likewise.
11816         (inline_string_cmp): Likewise.  Use c_readstr instead of
11817         builtin_memcpy_read_str.  Add unit_mode temporary.
11819 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11821         PR middle-end/86528
11822         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
11823         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
11825 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11827         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
11828         Alphabetize prototypes of built-in functions, separating out
11829         built-in functions that are listed in this section but should be
11830         described elsewhere.
11832 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
11834         PR target/86511
11835         * expmed.c (emit_store_flag): Do not emit setcc followed by a
11836         conditional move when trapping comparison was split to a
11837         non-trapping one (and vice versa).
11839 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11841         * config/s390/s390.c (s390_function_profiler): Generate nops
11842         instead of profiler call sequences.
11843         * config/s390/s390.opt: Add the new option.
11845 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11847         * config/s390/s390.c (s390_function_profiler): Generate
11848         __mcount_loc section.
11849         * config/s390/s390.opt: Add the new option.
11851 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11853         * common.opt: Add the new warning.
11854         * config/s390/s390.c (s390_function_profiler): Emit "brasl
11855         %r0,__fentry__" when -mfentry is specified.
11856         (s390_option_override_internal): Disallow -mfentry for 31-bit
11857         CPUs.
11858         * config/s390/s390.opt: Add the new option.
11860 2018-07-16  Richard Biener  <rguenther@suse.de>
11862         PR lto/86523
11863         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
11864         for function-local FUNCTION_DECL and RESULT_DECL immediately.
11866 2018-07-16  Martin Liska  <mliska@suse.cz>
11868         PR ipa/86529
11869         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
11870         to ::get_create.
11872 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11874         * config/arc/arcHS.md: Update ARCHS scheduling rules.
11876 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11878         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
11879         for ARCHS4x.
11880         * config/arc/arc-cpus.def (hs4x): New cpu.
11881         (hs4xd): Likewise.
11882         * config/arc/arc-tables.opt: Regenerate.
11883         * config/arc/arc.c (arc_sched_issue_rate): New function.
11884         (TARGET_SCHED_ISSUE_RATE): Define.
11885         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
11886         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
11887         fpu_cvt.
11888         (attr tune): Add ARCHS4x tune values.
11889         (attr tune_dspmpy): Define.
11890         (*tst): Correct instruction type.
11891         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
11892         * config/arc/arcHS4x.md: New file.
11893         * config/arc/fpu.md: Update instruction type attributes.
11894         * config/arc/t-multilib: Regenerate.
11896 2018-07-16  Tom de Vries  <tdevries@suse.de>
11898         PR debug/86455
11899         * var-tracking.c (vt_initialize): Fix pre_dec handling.
11901 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
11903         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
11904         early clobber.
11906 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
11908         PR tree-optimization/86514
11909         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
11910         conversion to a boolean type from a type with greater precision.
11912 2018-07-16  Tom de Vries  <tdevries@suse.de>
11914         * var-tracking.c (vt_initialize): Print adjusted insn slim if
11915         dump_flags request TDF_SLIM.
11917 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
11919         * fold-const.c (int_const_binop_1): Abstract...
11920         (wide_int_binop): ...wide int code here.
11921         (poly_int_binop): ...poly int code here.
11922         Abstract the rest of int_const_binop_1 into int_const_binop.
11923         * fold-const.h (wide_int_binop): New.
11924         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
11925         Remove useless PLUS/MINUS_EXPR case.
11926         (zero_nonzero_bits_from_vr): Move wide int code...
11927         (zero_nonzero_bits_from_bounds): ...here.
11928         (extract_range_from_binary_expr_1): Move mask optimization code...
11929         (range_easy_mask_min_max): ...here.
11930         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
11931         (range_easy_mask_min_max): New.
11933 2018-07-15  Jeff Law  <law@redhat.com>
11935         PR target/85993
11936         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
11937         block.
11939 2018-07-14  Jim Wilson  <jimw@sifive.com>
11941         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
11943 2018-07-14  Paul Koning  <ni1d@arrl.net>
11945         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
11947 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
11949         * lto-streamer-out.c (copy_function_or_variable): Dump info about
11950         copying section.
11952 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
11953             Steve Munroe  <munroesj52@gmail.com>
11955         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
11956         (_mm_andnot_si128): Likewise.
11957         (_mm_or_si128): Likewise.
11958         (_mm_xor_si128): Likewise.
11960 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
11962         PR middle-end/78809
11963         * builtins.c (expand_builtin_memcmp): Inline the calls first
11964         when result_eq is false.
11965         (expand_builtin_strcmp): Inline the calls first.
11966         (expand_builtin_strncmp): Likewise.
11967         (inline_string_cmp): New routine. Expand a string compare
11968         call by using a sequence of char comparison.
11969         (inline_expand_builtin_string_cmp): New routine. Inline expansion
11970         a call to str(n)cmp/memcmp.
11971         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
11972         New option.
11973         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
11975 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
11977         * config/arm/driver-arm.c: Include arm-native.h.
11978         (host_detect_local_cpu): Use auto-generated data tables.
11979         (vendors, arm_cpu_table): Delete.  Move part information to ...
11980         * config/arm/arm-cpus.in: ... here.
11981         * config/arm/parsecpu.awk (gen_native): New function.
11982         (vendor, part): New CPU fields.
11983         (END): Add support for building the native CPU detection tables.
11984         * config/arm/t-arm (arm-native.h): Add build rule.
11985         (driver-arm.o): Add dependency on arm-native.h.
11987 2018-07-13  Richard Biener  <rguenther@suse.de>
11989         PR middle-end/85974
11990         * match.pd (addr1 - addr2): Allow either of the operand to
11991         have a conversion.
11993 2018-07-13  Tom de Vries  <tdevries@suse.de>
11995         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
11996         in remap_ssa_name.
11998 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12000         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12001         arrays instead of numbered variables.
12003 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12005         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12006         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12008 2018-07-13  Richard Biener  <rguenther@suse.de>
12010         PR debug/86452
12011         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12012         instead of get_context_die.
12014 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12015             Richard Biener  <rguenther@suse.de>
12017         PR middle-end/86489
12018         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12019         that the loop latch destination where phi is defined.
12021 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12023         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12024         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12025         (riscv_set_current_function): Extract getting interrupt type to new
12026         function.
12027         (riscv_get_interrupt_type): New function.
12028         (riscv_merge_decl_attributes): New function, checking interrupt type is
12029         same.
12030         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12032 2018-07-12  Paul Koning  <ni1d@arrl.net>
12034         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12035         directive.
12037 2018-07-12  Paul Koning  <ni1d@arrl.net>
12039         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12040         zero reference, add doloop_end instead.
12041         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12042         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12043         detail for doloop_end.
12045 2018-07-12  Martin Sebor  <msebor@redhat.com>
12047         PR c/86453
12048         * attribs.c (decl_attributes): Reject conflicting attributes before
12049         calling attribute handlers.
12051 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12053         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12054          parameter.
12055         (gcc::dump_manager::get_dump_file_name): likewise.
12056         (dump_begin): Likewise.
12057         * dumpfile.h (dump_begin): Update prototype.
12058         (gcc::dump_manager::get_dump_file_name,
12059         gcc::dump_manager::get_dump_file_name): Update prototype.
12061 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12063         * internal-fn.h (vectorizable_internal_fn_p): New function.
12064         * tree-vect-slp.c (compatible_calls_p): Likewise.
12065         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12066         to internal functions.
12067         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12069 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12071         * fold-const.h (inverse_conditions_p): Declare.
12072         * fold-const.c (inverse_conditions_p): New function.
12073         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12074         that test the inverse condition of a conditional internal function.
12075         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12076         * internal-fn.c (internal_fn_mask_index): Handle conditional
12077         internal functions.
12078         (vectorized_internal_fn_supported_p): New function.
12079         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12080         (any_pred_load_store): Replace with...
12081         (need_to_predicate): ...this new variable.
12082         (redundant_ssa_names): New variable.
12083         (ifcvt_can_use_mask_load_store): Move initial checks to...
12084         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12085         for which a conditional internal function exists.
12086         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12087         instead of ifcvt_can_use_mask_load_store.  Update after variable
12088         name change.
12089         (predicate_load_or_store): New function, split out from
12090         predicate_mem_writes.
12091         (check_redundant_cond_expr): New function.
12092         (value_available_p): Likewise.
12093         (predicate_rhs_code): Likewise.
12094         (predicate_mem_writes): Rename to...
12095         (predicate_statements): ...this.  Use predicate_load_or_store
12096         and predicate_rhs_code.
12097         (combine_blocks, tree_if_conversion): Update after above name changes.
12098         (ifcvt_local_dce): Handle redundant_ssa_names.
12099         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12100         general conditional functions.
12101         * tree-vect-stmts.c (vectorizable_call): Likewise.
12103 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12104             Alan Hayward  <alan.hayward@arm.com>
12105             David Sherwood  <david.sherwood@arm.com>
12107         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12108         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12109         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12110         plus and minus and convert them into IFN_COND_FMA-based sequences.
12111         (convert_mult_to_fma): Handle conditional plus and minus.
12113 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12115         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12116         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12117         (cond_fnms_optab): New optabs.
12118         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12119         internal functions.
12120         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12121         * internal-fn.h (get_conditional_internal_fn): Declare.
12122         (get_unconditional_internal_fn): Likewise.
12123         * internal-fn.c (cond_ternary_direct): New macro.
12124         (expand_cond_ternary_optab_fn): Likewise.
12125         (direct_cond_ternary_optab_supported_p): Likewise.
12126         (FOR_EACH_COND_FN_PAIR): Likewise.
12127         (get_conditional_internal_fn): New function.
12128         (get_unconditional_internal_fn): Likewise.
12129         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12130         (gimple_match_op::gimple_match_op): Add a new overload for 5
12131         operands.
12132         (gimple_match_op::set_op): Likewise.
12133         (gimple_resimplify5): Declare.
12134         * genmatch.c (decision_tree::gen): Generate simplifications for
12135         5 operands.
12136         * gimple-match-head.c (gimple_simplify): Define an overload for
12137         5 operands.  Handle calls with 5 arguments in the top-level overload.
12138         (convert_conditional_op): Handle conversions from unconditional
12139         internal functions to conditional ones.
12140         (gimple_resimplify5): New function.
12141         (build_call_internal): Pass a fifth operand.
12142         (maybe_push_res_to_seq): Likewise.
12143         (try_conditional_simplification): Try converting conditional
12144         internal functions to unconditional internal functions.
12145         Handle 3-operand unconditional forms.
12146         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12147         Define ternary equivalents of the current rules for binary conditional
12148         internal functions.
12149         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12150         ternary operations.
12151         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12152         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12153         (optab): Handle them.
12154         (SVE_COND_FP_TERNARY): New int iterator.
12155         (sve_fmla_op, sve_fmad_op): New int attributes.
12156         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12157         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12158         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12160 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12162         * target.def (preferred_else_value): New target hook.
12163         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12164         * doc/tm.texi: Regenerate.
12165         * targhooks.h (default_preferred_else_value): Declare.
12166         * targhooks.c (default_preferred_else_value): New function.
12167         * internal-fn.h (conditional_internal_fn_code): Declare.
12168         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12169         (get_conditional_internal_fn): Use it.
12170         (conditional_internal_fn_code): New function.
12171         * gimple-match.h (gimple_match_cond): New struct.
12172         (gimple_match_op): Add a cond member function.
12173         (gimple_match_op::gimple_match_op): Update all forms to take a
12174         gimple_match_cond.
12175         * genmatch.c (expr::gen_transform): Use the same condition as res_op
12176         for the suboperation, but don't specify a particular else_value.
12177         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12178         (visit_nary_op, visit_reference_op_load): Pass
12179         gimple_match_cond::UNCOND to the gimple_match_op constructor.
12180         * gimple-match-head.c: Include tree-eh.h
12181         (convert_conditional_op): New function.
12182         (maybe_resimplify_conditional_op): Likewise.
12183         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12184         (gimple_resimplify2): Likewise.
12185         (gimple_resimplify3): Likewise.
12186         (gimple_resimplify4): Likewise.
12187         (maybe_push_res_to_seq): Return null for conditional operations.
12188         (try_conditional_simplification): New function.
12189         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
12190         constructor.
12191         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12192         IFN_COND_* call.
12193         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12194         function.
12195         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12197 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12199         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12200         DECL_FCONTEXT
12201         (hash_tree): Do not hash DECL_FCONTEXT
12202         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12203         Do not stream DECL_FCONTEXT.
12204         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12205         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12207 2018-07-12  Richard Biener  <rguenther@suse.de>
12209         PR debug/86462
12210         * dwarf2out.c (gen_block_die): Only output blocks when they have
12211         at least one !DECL_IGNORED_P variable.
12213 2018-07-12  Richard Biener  <rguenther@suse.de>
12215         PR target/84829
12216         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12217         Remove -mieee-fp handling.
12219 2018-07-12  Richard Biener  <rguenther@suse.de>
12221         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12222         left-over from last patch.
12224 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12226         PR tree-optimization/86492
12227         * gimple-ssa-store-merging.c
12228         (imm_store_chain_info::coalesce_immediate_stores): Call
12229         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12231 2018-07-12  Richard Biener  <rguenther@suse.de>
12233         PR middle-end/86479
12234         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12235         move possibly trapping operations into the conditional.
12237 2018-07-12  Richard Biener  <rguenther@suse.de>
12239         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12240         (vn_lookup_simplify_result): Remove recursion limit applied
12241         here.
12242         (vn_nary_build_or_lookup_1): Adjust.
12243         (try_to_simplify): Likewise.
12244         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12245         here.
12246         (gimple_resimplify2): Likewise.
12247         (gimple_resimplify3): Likewise.
12248         (gimple_resimplify4): Likewise.
12250 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12252         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12253         Use __mmask64 type instead of __mmask8 for __M argument.
12254         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12255         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12256         __U argument.
12257         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12258         __mmask16 for __M argument.
12259         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12260         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12261         to __mmask16 instead of __mmask8.
12262         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12263         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12264         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12265         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12266         instead of __mmask16 for __U argument.
12267         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12268         __mmask16 instead of __mmask8 for __U argument.
12269         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12270         __U argument.
12271         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12272         __mmask16.
12273         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12274         argument.
12275         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12276         __U argument.
12277         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12278         __mmask16.
12279         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12280         of __mmask16.
12281         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12282         __U argument.
12283         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12284         __U argument.
12285         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12286         __U argument.
12287         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12288         __U argument.
12289         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12290         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12291         return type as well as __M argument type and all casts from __mmask8
12292         to __mmask32.
12293         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12294         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12295         return type as well as __M argument type and all casts from __mmask8
12296         to __mmask16.
12297         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12298         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12299         return type as well as __M argument type and all casts from __mmask8
12300         to __mmask32.
12301         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12302         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12303         return type as well as __M argument type and all casts from __mmask8
12304         to __mmask16.
12305         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12306         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12307         __mmask16.
12309 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
12311         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12312         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12313         for __U argument.
12315 2018-07-11  Paul Koning  <ni1d@arrl.net>
12317         * doc/md.texi (define_subst): Document how multiple occurrences of
12318         the same argument in the replacement pattern are handled.
12320 2018-07-11  Paul Koning  <ni1d@arrl.net>
12322         * doc/extend.texi (Common Variable Attributes): Move "mode" into
12323         alphabetical order.
12324         (Common Type Attributes): Add "mode" attribute.
12326 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
12328         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12329         stream DECL_ORIGINAL_TYPE.
12330         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12331         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12332         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12333         Do not walk original type.
12334         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12335         external decls.
12336         (write_ts_decl_non_common_tree_pointers): Do not stream
12337         DECL_ORIGINAL_TYPE
12338         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12339         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12341 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12343         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12344         thread twice from the same starting edge.
12346 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12348         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12349         * gimple.c (gimple_call_nonnull_result_p): ...here...
12350         (gimple_call_nonnull_arg): ...and here.
12351         * gimple.h (gimple_call_nonnull_result_p): New.
12352         (gimple_call_nonnull_arg): New.
12354 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
12356         * config/arm/arm-cpus.in: Move information from fpu field of each
12357         cpu definition to the isa field.
12358         * config/arm/parsecpu.awk (fpu): Delete match rule.
12359         (gen_comm_data): Don't add bits from the CPU's FPU entry.
12361 2018-07-11  Richard Biener  <rguenther@suse.de>
12363         PR debug/86457
12364         * dwarf2out.c (init_sections_and_labels): Use
12365         output_asm_line_debug_info consistently.
12366         (dwarf2out_early_finish): Likewise.
12367         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12368         type units.
12370 2018-07-11  Richard Biener  <rguenther@suse.de>
12372         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12373         Rework father_bb setting in a way to avoid propagating constants
12374         multiple times on a loop body.
12376 2018-07-10  Mark Wielaard  <mark@klomp.org>
12378         PR debug/86459
12379         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12381 2018-07-10  Richard Biener  <rguenther@suse.de>
12383         * hash-map.h (hash_map::iterator::operator*): Return
12384         references to key and value.
12386 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
12388         PR c++/86443
12389         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12390         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12391         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12392         (gimplify_omp_for): For composite loops, move outer
12393         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12394         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
12395         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
12396         TREE_LIST for both the original class iterator and the "last" helper
12397         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
12398         loop, remember has_decl_expr from outer composite loops for the
12399         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12401 2018-07-09  Martin Sebor  <msebor@redhat.com>
12403         PR middle-end/77357
12404         PR middle-end/86428
12405         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12406         accessing implicitly initialized array elements.
12407         * expr.c (string_constant): Handle string initializers of
12408         character arrays within aggregates.
12409         * gimple-fold.c (fold_array_ctor_reference): Add argument.
12410         Store element offset.  As a special case, handle zero size.
12411         (fold_nonarray_ctor_reference): Same.
12412         (fold_ctor_reference): Add argument.  Store subobject offset.
12413         * gimple-fold.h (fold_ctor_reference): Add argument.
12415 2018-07-09  Paul Koning  <ni1d@arrl.net>
12417         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12418         (pdp11_insn_cost): New function.
12419         (pdp11_md_asm_adjust): New function.
12420         (TARGET_INVALID_WITHIN_DOLOOP): Define.
12421         (pdp11_rtx_costs): Update to match machine better.
12422         (output_addr_const_pdp11): Correct format mismatch warnings.
12423         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12424         * config/pdp11/pdp11.md: General change to add base_cost and/or
12425         length attributes for use by new pdp11_insn_cost function.
12426         (MIN_BRANCH): Correct definition.
12427         (MIN_SOB): Ditto.
12428         (doloop_end): Use standard pattern name for looping pattern.
12429         (doloop_end_nocc): New.
12430         (movsf): Add another constraint alternative.
12431         (zero_extendqihi2): Add constraint alternatives for not in place
12432         extend.
12433         (zero_extendhisi2): Remove.
12434         (shift patterns): Add CC handling variants.
12435         (bswaphi2): New.
12436         (bswapsi2): New.
12437         (rothi3): New.
12438         (define_peephole2): New peephole to recognize mov that sets CC for
12439         subsequent test.
12441 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12443         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12444         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12445         wi::add.
12447 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
12449         PR c/86420
12450         * real.c (real_nextafter): Return true if result is denormal.
12452 2018-07-09  Martin Liska  <mliska@suse.cz>
12454         * common.opt: Add back wrongly removed attribute.
12456 2018-07-09  Richard Biener  <rguenther@suse.de>
12458         PR debug/86413
12459         * dwarf2out.c (gen_block_die): For an early generated DIE
12460         always output high/low PC attributes.
12462 2018-07-09  Tom de Vries  <tdevries@suse.de>
12464         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12465         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12466         onto VAR_DECL with abstract origin.
12468 2018-07-07  Jim Wilson  <jimw@sifive.com>
12470         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12472 2018-07-07  Tom de Vries  <tdevries@suse.de>
12474         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12476 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
12478         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12479         overflow_type.
12480         (combine_bound): Use wide-int overflow calculation instead of
12481         rolling our own.
12482         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12483         overflow_type.
12484         * fold-const.c (int_const_binop_2): Same.
12485         (extract_muldiv_1): Same.
12486         (fold_div_compare): Same.
12487         (fold_abs_const): Same.
12488         * match.pd: Same.
12489         * poly-int.h (add): Same.
12490         (sub): Same.
12491         (neg): Same.
12492         (mul): Same.
12493         * predict.c (predict_iv_comparison): Same.
12494         * profile-count.c (slow_safe_scale_64bit): Same.
12495         * simplify-rtx.c (simplify_const_binary_operation): Same.
12496         * tree-chrec.c (tree_fold_binomial): Same.
12497         * tree-data-ref.c (split_constant_offset_1): Same.
12498         * tree-if-conv.c (idx_within_array_bound): Same.
12499         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12500         * tree-ssa-phiopt.c (minmax_replacement): Same.
12501         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12502         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12503         * vr-values.c (vr_values::adjust_range_with_scev): Same.
12504         * wide-int.cc (wi::add_large): Same.
12505         (wi::mul_internal): Same.
12506         (wi::sub_large): Same.
12507         (wi::divmod_internal): Same.
12508         * wide-int.h: Change overflow type to overflow_type for neg, add,
12509         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12510         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12511         mul_internal, divmod_internal.
12512         (overflow_type): New enum.
12513         (accumulate_overflow): New.
12515 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
12517         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12518         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12520 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12522         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12523         argument is checked for zero before entering loop, avoid checking again.
12525 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12527         * gimplify.h (generic_expr_could_trap_p): Set as global function.
12528         * gimplify.h (generic_expr_could_trap_p): Likwise.
12529         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
12531 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
12533         PR tree-optimization/86401
12534         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
12535         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
12536         (fold_bit_and_mask): ... here.  New helper function for match.pd.
12537         * fold-const.h (fold_bit_and_mask): Declare.
12538         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
12540 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
12542         PR target/86324
12543         * target.def (translate_mode_attribute): New hook.
12544         * targhooks.h (default_translate_mode_attribute): Declare.
12545         * targhooks.c (default_translate_mode_attribute): New function.
12546         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
12547         * doc/tm.texi: Regenerate.
12548         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
12549         (rs6000_translate_mode_attribute): New function.
12551 2018-07-06  Paul Koning  <ni1d@arrl.net>
12553         * doc/md.texi (define_split): Document DONE and FAIL.
12554         (define_peephole2): Ditto.
12556 2018-07-05  Jeff Law  <law@redhat.com>
12558         PR tree-optimization/86010
12559         * tree-ssa-dse.c (compute_trims): More aggressively trim at
12560         both the head and tail of mem* and str* calls.
12562 2018-07-05  Jim Wilson  <jimw@sifive.com>
12564         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
12566 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
12568         * config/aarch64/aarch64-simd.md: correct flags text for
12569         MIN_EXPR replacement.
12571 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
12573         * configure: Regenerated.
12575 2018-07-05  Carl Love  <cel@us.ibm.com>
12577         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
12578         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
12579         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
12580         VSX_BUILTIN_DOUBLEL_V4SF.
12582 2018-07-05  Martin Sebor  <msebor@redhat.com>
12584         PR c++/86400
12585         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
12586         than its domain to compute its the upper bound of a char array.
12588 2018-07-05  Nathan Sidwell  <nathan@acm.org>
12590         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
12591         * doc/cpp.texi: Update comment.
12592         * doc/tm.texi: Rebuilt.
12593         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
12594         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
12595         * doc/extend.texi (Backwards Compatibility): Clarify it is system
12596         headers affected by extern "C".
12597         * system.h: Poison NO_IMPLICIT_EXTERN_C.
12598         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
12599         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
12600         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
12601         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
12602         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
12603         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
12604         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
12605         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
12606         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
12607         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
12608         config/sparc/sp64-elf.h, config/spu/spu.h,
12609         config/stormy16/stormy16.h, config/v850/v850.h,
12610         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
12611         define NO_IMPLICIT_EXTERN_C.
12612         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
12614 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12616         PR target/84711
12617         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12618         instead of GET_MODE_SIZE when comparing Units.
12620 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12622         PR target/84711
12623         * rtlanal.c (set_noop_p): Constrain on mode change,
12624         include hard-reg-set.h
12626 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12628         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
12630 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
12632         Revert
12633         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12635         PR sanitizer/84250
12636         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12637         libasan.
12638         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12640 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12642         PR sanitizer/84250
12643         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12644         libasan.
12645         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12647 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
12649         PR middle-end/86380
12650         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
12652 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
12654         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
12655         neg_*_op* variables.
12657 2018-07-04  Martin Liska  <mliska@suse.cz>
12659         * tree-switch-conversion.c: Define
12660         max_ratio_for_speed and max_ratio_for_size constants.
12662 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12663             Martin Liska  <mliska@suse.cz>
12665         PR middle-end/66240
12666         PR target/45996
12667         PR c/84100
12668         * common.opt: Rename align options with 'str_' prefix.
12669         * common/config/i386/i386-common.c (set_malign_value): New
12670         function.
12671         (ix86_handle_option): Use it to set -falign-* options/
12672         * config/aarch64/aarch64-protos.h (struct tune_params): Change
12673         type from int to string.
12674         * config/aarch64/aarch64.c: Update default values from int
12675         to string.
12676         * config/alpha/alpha.c (alpha_override_options_after_change):
12677         Likewise.
12678         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
12679         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12680         max skip conditionally.
12681         * config/i386/freebsd.h (SUBALIGN_LOG): New.
12682         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12683         max skip conditionally.
12684         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12685         max skip conditionally.
12686         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
12687         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12688         max skip conditionally.
12689         * config/i386/i386.c (struct ptt): Change type from int to
12690         string.
12691         (ix86_default_align): Set default values.
12692         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
12693         max skip conditionally.
12694         * config/i386/iamcu.h (SUBALIGN_LOG): New.
12695         (ASM_OUTPUT_MAX_SKIP_ALIGN):
12696         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
12697         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12698         max skip conditionally.
12699         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
12700         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
12701         * config/i386/x86-64.h (SUBALIGN_LOG): New.
12702         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12703         max skip conditionally.
12704         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12705         * config/ia64/ia64.c (ia64_option_override): Set default values
12706         for alignment options.
12707         * config/m68k/m68k.c: Handle new str_align_* options.
12708         * config/mips/mips.c (mips_set_compression_mode): Change
12709         type of constants.
12710         (mips_option_override): Set default values for options.
12711         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
12712         Likewise.
12713         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12714         Likewise.
12715         * config/rx/rx.c (rx_option_override): Likewise.
12716         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
12717         (LABEL_ALIGN): Use align_labels_log.
12718         (LOOP_ALIGN): Use align_loops_align.
12719         * config/s390/s390.c (s390_asm_output_function_label): Use new
12720         macros.
12721         * config/sh/sh.c (sh_override_options_after_change):
12722         Change type of constants.
12723         * config/spu/spu.c (spu_sched_init): Likewise.
12724         * config/sparc/sparc.c (sparc_option_override): Set default
12725         values for options.
12726         * config/visium/visium.c (visium_option_override): Likewise.
12727         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
12728         emit p2align format with last argument if it's not needed.
12729         * doc/invoke.texi: Document extended format of -falign-*.
12730         * final.c: Use align_labels alignment.
12731         * flags.h (struct target_flag_state): Change type to use
12732         align_flags.
12733         (struct align_flags_tuple): New.
12734         (struct align_flags): Likewise.
12735         (align_loops_log): Redefine macro to use new types.
12736         (align_loops_max_skip): Redefine macro to use new types.
12737         (align_jumps_log): Redefine macro to use new types.
12738         (align_jumps_max_skip): Redefine macro to use new types.
12739         (align_labels_log): Redefine macro to use new types.
12740         (align_labels_max_skip): Redefine macro to use new types.
12741         (align_functions_log): Redefine macro to use new types.
12742         (align_loops): Redefine macro to use new types.
12743         (align_jumps): Redefine macro to use new types.
12744         (align_labels): Redefine macro to use new types.
12745         (align_functions): Redefine macro to use new types.
12746         (align_functions_max_skip): Redefine macro to use new types.
12747         (align_loops_value): New macro.
12748         (align_jumps_value): New macro.
12749         (align_labels_value): New macro.
12750         (align_functions_value): New macro.
12751         * function.c (invoke_set_current_function_hook): Propagate
12752         alignment values from flags to global variables default in
12753         topleev.h.
12754         * ipa-icf.c (sem_function::equals_wpa): Use
12755         cl_optimization_option_eq instead of memcmp.
12756         * lto-streamer.h (cl_optimization_stream_out): Support streaming
12757         of string types.
12758         (cl_optimization_stream_in): Likewise.
12759         * optc-save-gen.awk: Support strings in cl_optimization.
12760         * opth-gen.awk: Likewise.
12761         * opts.c (finish_options): Remove error checking of invalid
12762         value ranges.
12763         (MAX_CODE_ALIGN): Remove.
12764         (MAX_CODE_ALIGN_VALUE): Likewise.
12765         (parse_and_check_align_values): New function.
12766         (check_alignment_argument): Likewise.
12767         (common_handle_option): Use check_alignment_argument.
12768         * opts.h (parse_and_check_align_values): Declare.
12769         * toplev.c (init_alignments): Remove.
12770         (read_log_maxskip): New.
12771         (parse_N_M): Likewise.
12772         (parse_alignment_opts): Likewise.
12773         (backend_init_target): Remove usage of init_alignments.
12774         * toplev.h (parse_alignment_opts): Declare.
12775         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
12776         argument.
12777         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
12778         * tree.c (cl_option_hasher::equal): New.
12779         * varasm.c: Use new global macros.
12781 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12783         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
12784         Use a simpler align directive also if MAXSKIP = ALIGN-1.
12785         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12786         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12787         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12788         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12789         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
12790         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
12791         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
12792         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12793         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12794         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12795         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12797 2018-07-04  Martin Liska  <mliska@suse.cz>
12798             Jonathan Wakely  <jwakely@redhat.com>
12800         * coverage.c: Use correct type.
12801         * doc/invoke.texi: Language correction.
12803 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
12805         PR target/85620
12806         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
12807         ENDBRANCH for non-tail call which may return via indirect branch.
12808         * doc/extend.texi: Document indirect_return attribute.
12810 2018-07-03  Martin Sebor  <msebor@redhat.com>
12812         PR tree-optimization/86274
12813         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
12814         precondition.
12815         (format_floating): Correct handling of infinities and NaNs.
12817 2018-07-03  Martin Sebor  <msebor@redhat.com>
12819         * print-tree.c (print_real_cst): New function.
12820         (print_node_brief): Call it.
12821         (print_node): Ditto.
12823 2018-07-03  Jeff Law  <law@redhat.com>
12825         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
12826         into a single pattern.
12828         * config/h8300/h8300.md (ors code_iterator): New.
12829         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
12830         a single pattern and single splitter.
12831         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
12832         (iorqi3_1, xorqi3_1): Likewise.
12833         (iorqi3, xorqi3 expanders): Similarly.
12835         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
12836         (movmd_internal) into a single pattern using the P mode iterator.
12837         (movmd splitters): Similarly.
12838         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
12839         (movsd splitters): Similarly.
12841         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
12842         ADDB, ADDW and ADDL into a single ADD attribute which selects the
12843         right table based on the size of the operand.
12844         * config/h8300/h8300.md (length_table): Corresponding changes. All
12845         references to "addb", "addw" and "addl" changed to "add".
12846         (btst patterns): Merge two variants into a single pattern.
12847         (tstqi, tsthi): Likewise.
12848         (addhi3_incdec, addsi3_incdec): Likewise.
12849         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
12850         (mulhi3, mulsi3): Likewise.
12851         (udivhi3, udivsi3): Likewise.
12852         (divhi3, divsi3): Likewise.
12853         (andorqi3, andorhi3, andorsi3): Likewise.
12855 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
12857         PR target/85694
12858         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
12859         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
12861 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12863         PR tree-optimization/85694
12864         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
12865         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
12866         UNSPEC_URHADD.
12867         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
12868         (<u>avg<mode>3_ceil): New patterns.
12870 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
12872         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
12873         scan-tree-dump directive.
12874         * gcc.dg/vect/slp-perm-2.c: Likewise.
12875         * gcc.dg/vect/slp-perm-3.c: Likewise.
12876         * gcc.dg/vect/slp-perm-5.c: Likewise.
12877         * gcc.dg/vect/slp-perm-6.c: Likewise.
12878         * gcc.dg/vect/slp-perm-7.c: Likewise.
12879         * gcc.dg/vect/slp-perm-8.c: Likewise.
12881 2018-07-03  Marek Polacek  <polacek@redhat.com>
12883         PR middle-end/86202
12884         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
12886 2018-07-03  Richard Biener  <rguenther@suse.de>
12888         PR ipa/86389
12889         * tree-ssa-structalias.c (find_func_clobbers): Properly
12890         handle indirect calls.
12892 2018-07-03  Jeff Law  <law@redhat.com>
12894         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
12895         (shifts): New code iterator.
12896         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
12897         expander.  Fix HImode handling on H8/SX.
12898         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
12899         (subqi3, subhi3, subsi3 expanders): Likewise.
12900         (andqi3, andhi3, andsi3 expanders): Likewise.
12901         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
12902         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
12903         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
12904         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
12905         (zero_extendqihi2, zero_extendqisi2): Likewise.
12906         (extendqihi2, extendqisi2): Likewise.
12907         (rotlqi3, rotlhi3, rotlsi3): Likewise.
12908         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
12909         (rotlqi3_1, rotlhi3_1): Likewise.
12910         (logicalhi3_sn, logicalsi3_sn): Likewise.
12911         (logicalhi3, logicalsi3): Likewise.
12913 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12915         * tree-vect-patterns.c (vect_recog_rotate_pattern)
12916         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12917         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
12918         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
12919         type to append_pattern_def_seq instead of creating a stmt_vec_info
12920         directly.
12921         (build_mask_conversion): Likewise.  Remove vinfo argument.
12922         (vect_add_conversion_to_patterm): Likewise, renaming to...
12923         (vect_add_conversion_to_pattern): ...this.
12924         (vect_recog_mask_conversion_pattern): Update call to
12925         build_mask_conversion.  Pass the vector type to
12926         append_pattern_def_seq here too.
12927         (vect_recog_gather_scatter_pattern): Update call to
12928         vect_add_conversion_to_pattern.
12930 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12932         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
12933         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12934         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
12935         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
12936         STMT_VINFO_PATTERN_DEF_SEQ to null here.
12937         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
12938         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
12939         append_pattern_def_seq instead of new_pattern_def_seq.
12940         (vect_recog_divmod_pattern): Do both of the above.
12941         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
12942         is null.
12944 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12946         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
12947         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
12948         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
12949         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
12950         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
12951         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12952         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
12953         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12954         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12955         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
12956         parameter with a single stmt_vec_info.
12957         (vect_recog_func_ptr): Likewise.
12958         (vect_recog_gather_scatter_pattern): Likewise, folding in...
12959         (vect_try_gather_scatter_pattern): ...this.
12960         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
12961         the stmt_vec_info of the statement to be matched.  Don't clear
12962         STMT_VINFO_RELATED_STMT.
12963         (vect_pattern_recog): Update call accordingly.
12965 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12967         PR tree-optimization/85694
12968         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
12969         (uavgM3_ceil): Document new optabs.
12970         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
12971         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
12972         functions.
12973         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
12974         (savg_ceil_optab): New optabs.
12975         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
12976         (vect_vect_recog_func_ptrs): Add it.
12977         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
12978         constant directly from the associated lhs.
12980 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12982         * tree-vect-patterns.c (vect_split_statement): New function.
12983         (vect_convert_input): Use it to try to split an existing cast.
12985 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12987         * poly-int.h (print_hex): New function.
12988         * dumpfile.h (dump_dec, dump_hex): Declare.
12989         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
12990         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
12991         min_input_precision, operation_precision and operation_sign.
12992         * tree-vect-patterns.c (vect_get_range_info): New function.
12993         (vect_same_loop_or_bb_p, vect_single_imm_use)
12994         (vect_operation_fits_smaller_type): Delete.
12995         (vect_look_through_possible_promotion): Add an optional
12996         single_use_p parameter.
12997         (vect_recog_over_widening_pattern): Rewrite to use new
12998         stmt_vec_info infomration.  Handle one operation at a time.
12999         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13000         (vect_truncatable_operation_p, vect_set_operation_type)
13001         (vect_set_min_input_precision): New functions.
13002         (vect_determine_min_output_precision_1): Likewise.
13003         (vect_determine_min_output_precision): Likewise.
13004         (vect_determine_precisions_from_range): Likewise.
13005         (vect_determine_precisions_from_users): Likewise.
13006         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13007         (vect_vect_recog_func_ptrs): Put over_widening first.
13008         Add cast_forwprop.
13009         (vect_pattern_recog): Call vect_determine_precisions.
13011 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13013         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13014         statements that have been replaced by further pattern statements.
13015         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13017 2018-07-03  Richard Biener  <rguenther@suse.de>
13019         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13020         always set *dt.  Dump vectype in vectype overload.
13021         * dumpfile.h (dump_gimple_expr): New function.
13022         (dump_gimple_expr_loc): Likewise.
13023         * dumpfile.c (dump_gimple_expr): New function.
13024         (dump_gimple_expr_loc): Likewise.
13026 2018-07-02  Jeff Law  <law@redhat.com>
13028         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13029         the H8/300, H8/300H and H8/S variants into a single pattern.
13030         (movhi_h8300, movqi_h8300hs): Similarly.
13031         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13032         (QHI mode iterator): New.
13034         * config/h8300/h8300.md: Remove trailing whitespace.
13036 2018-07-02  Jim Wilson  <jimw@sifive.com>
13038         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13039         instead of emit_insn for interrupt returns.
13040         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13041         (riscv_sret, riscv_uret): Likewise.
13043 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13045         * pretty-print.c (selftest::test_pp_format): Move save and restore
13046         of quotes to class auto_fix_quotes, and add an instance.
13047         * selftest.c: Include "intl.h".
13048         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13049         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13050         * selftest.h (selftest::auto_fix_quotes): New class.
13052 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13054         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13055         (aarch64_sve_prepare_conditional_op): Remove.
13056         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13057         Allow aarch64_simd_reg_or_zero as select operand; remove
13058         the aarch64_sve_prepare_conditional_op call.
13059         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13060         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13061         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13062         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13063         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13064         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13065         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13066         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13067         and a splitters to match all of the *_any patterns.
13068         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13070         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13071         (SVE_COND_FP_BINARY_REV): Remove.
13072         (sve_int_op_rev, sve_fp_op_rev): New.
13073         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13074         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13075         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13076         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13077         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13078         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13079         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13080         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13081         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13083         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13084         Remove match_dup 1 from the inner unspec.
13085         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13087         * config/aarch64/aarch64.md (movprfx): New attr.
13088         (length): Default movprfx to 8.
13089         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13090         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13091         (*<su>mul<SVE_I>3_highpart): Likewise.
13092         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13093         (*v<ASHIFT><SVE_I>3): Likewise.
13094         (*<su><MAXMIN><SVE_I>3): Likewise.
13095         (*<su><MAXMIN><SVE_F>3): Likewise.
13096         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13097         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13098         (*div<SVE_F>4): Likewise.
13100 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13102         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13103         in dump string.
13105 2018-07-02  Richard Biener  <rguenther@suse.de>
13107         PR tree-optimization/86363
13108         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13109         memset argument refers to a non-variable address.
13111 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13113         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13114         {PLUS,MINUS}_EXPR code to...
13115         (adjust_symbolic_bound): ...here,
13116         (combine_bound): ...here,
13117         (set_value_range_with_overflow): ...and here.
13119 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13121         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13122         code...
13123         (extract_range_from_abs_expr): ...here.
13125 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13127         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13128         -fno-omit-frame-pointer when not optimizing.
13130 2018-07-02  Martin Liska  <mliska@suse.cz>
13132         PR ipa/86279
13133         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13134         (propagate_nothrow): Likewise.
13136 2018-07-02  Martin Liska  <mliska@suse.cz>
13138         PR ipa/86323
13139         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13141 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13143         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13144         function in r262149, changing "loc" param from source_location to
13145         const dump_location_t &.
13146         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13147         declaration, as above.
13149 2018-07-01  Paul Koning  <ni1d@arrl.net>
13151         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13152         -munit-asm, -mgnu-asm, -mdec-asm.
13153         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13154         (pdp11_output_labelref): New.
13155         (pdp11_output_def): New.
13156         (pdp11_output_addr_vec_elt): New.
13157         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
13158         %# and %@ format codes.
13159         (pdp11_option_override): New.
13160         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13161         (pdp11_output_ident): New.
13162         (pdp11_asm_named_section): New.
13163         (pdp11_asm_init_sections): New.
13164         (pdp11_file_start): New.
13165         (pdp11_file_end): New.
13166         (output_ascii): Use .ascii/.asciz for -mdec-asm.
13167         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
13168         %o, like %c but octal.
13169         (pdp11_option_override): New.
13170         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13171         -mdec-asm.
13172         (DATA_SECTION_ASM_OP): Ditto.
13173         (READONLY_DATA_SECTION_ASM_OP): New.
13174         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13175         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13176         (ASM_OUTPUT_LABELREF): Ditto.
13177         (ASM_OUTPUT_DEF): Ditto.
13178         (ASM_OUTPUT_EXTERNAL): New.
13179         (ASM_OUTPUT_SOURCE_FILENAME): New.
13180         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13181         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13182         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
13183         %# and %@ format codes.
13184         * config/pdp11/pdp11.opt (mgnu-asm): New.
13185         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13186         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13187         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13189 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
13191         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13192         dereferencing path[] beyond its length.
13193         (debug_path): New.
13194         (debug_all_paths): New.
13195         (rewire_first_differing_edge): New.
13196         (adjust_paths_after_duplication): New.
13197         (duplicate_thread_path): Call adjust_paths_after_duplication.
13198         Add new argument.
13199         (thread_through_all_blocks): Add new argument to
13200         duplicate_thread_path.
13202 2018-06-30  Jim Wilson  <jimw@sifive.com>
13204         * config/riscv/predicates.md (p2m1_shift_operand): New.
13205         (high_mask_shift_operand): New.
13206         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13207         pattern using p2m1_shift_operand.
13208         (lshsi3_zero_extend_3+2): New combiner pattern using
13209         high_mask_shift_operand.
13211 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13213         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13214         split out from...
13215         (vect_recog_rotate_pattern): ...here.
13216         (vect_convert_input): Try to insert casts of invariants in the
13217         preheader.
13218         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13219         preheader to be empty.
13221 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13223         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13224         vector type.  If given, install it in the new statement's
13225         STMT_VINFO_VECTYPE.
13226         (vect_element_precision): New function.
13227         (vect_unpromoted_value): New struct.
13228         (vect_unpromoted_value::vect_unpromoted_value): New function.
13229         (vect_unpromoted_value::set_op): Likewise.
13230         (vect_look_through_possible_promotion): Likewise.
13231         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13232         (vect_widened_op_tree, vect_convert_input): Likewise.
13233         (vect_convert_inputs, vect_convert_output): Likewise.
13234         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13235         to handle the optional cast of the multiplication result and
13236         vect_widened_op_tree to detect the widened multiplication itself.
13237         Do not require the input and output of promotion casts to have
13238         the same sign, but base the signedness of the operation on the
13239         input rather than the result.  If the pattern includes two
13240         promotions, check that those promotions have the same sign.
13241         Do not restrict the MULT_EXPR handling to a double-width result;
13242         handle quadruple-width results and wider.  Use vect_convert_inputs
13243         to convert the inputs to the common type.
13244         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13245         to handle the optional cast of the ABS result.  Also allow a sign
13246         change or a sign extension between the ABS and MINUS.
13247         Use vect_widened_op_tree to detect the widened subtraction and use
13248         vect_convert_inputs to convert the inputs to the common type.
13249         (vect_handle_widen_op_by_const): Delete.
13250         (vect_recog_widen_op_pattern): New function.
13251         (vect_recog_widen_mult_pattern): Use it.
13252         (vect_recog_widen_shift_pattern): Likewise.
13253         (vect_recog_widen_sum_pattern): Use
13254         vect_look_through_possible_promotion to handle the promoted
13255         PLUS_EXPR operand.
13257 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13259         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13260         the containing gimple_seq *.
13261         * gimple-iterator.h (gsi_for_stmt): Declare it.
13262         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13263         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13264         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13265         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13266         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13267         checks.
13268         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13269         split out from...
13270         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
13271         statement being replaced is part of an existing pattern
13272         definition sequence, inserting the new pattern statements before
13273         the original one.
13274         (vect_pattern_recog_1): Don't return a bool.  If the statement
13275         is already part of a pattern, instead apply pattern matching
13276         to the pattern definition statements.  Don't clear the
13277         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13278         (vect_pattern_recog): Don't break after the first match;
13279         continue processing the pattern definition statements instead.
13280         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13282 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13284         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13285         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13286         (vect_recog_widen_sum_pattern): Use it.
13288 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13290         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13291         phi is not a pattern statement and has not been replaced by
13292         a pattern statement.
13293         * tree-vect-patterns.c (type_conversion_p): Don't check
13294         STMT_VINFO_IN_PATTERN_P.
13295         (vect_recog_vector_vector_shift_pattern): Likewise.
13296         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13297         the pattern statement rather than the original statement; check
13298         directly for a WIDEN_MULT_EXPR here.
13299         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13300         vect_is_simple_use to return the pattern statement rather
13301         than the original statement; use is_pattern_stmt_p to check
13302         for such a pattern statement.
13303         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13304         to return the pattern statement rather than the original statement;
13305         don't do the same transformation here.
13306         (vect_is_simple_use): If the defining statement has been replaced
13307         by a pattern statement, return the pattern statement instead.
13308         Remove the corresponding (local) transformation from the vectype
13309         overload.
13311 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13313         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13314         end and default to null.
13315         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13316         (vectorizable_reduction): Update calls accordingly, dropping the
13317         gimple ** argument if the passed-back statement isn't needed.
13318         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13319         (vect_recog_rotate_pattern): Likewise.
13320         (vect_recog_mask_conversion_pattern): Likewise.
13321         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13322         (vect_mask_constant_operand_p): Likewise.
13323         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13324         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13325         (get_group_load_store_type, get_load_store_type): Likewise.
13326         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13327         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13328         (vectorizable_conversion, vectorizable_assignment): Likewise.
13329         (vectorizable_shift, vectorizable_operation): Likewise.
13330         (vectorizable_store, vect_is_simple_cond): Likewise.
13331         (vectorizable_condition, vectorizable_comparison): Likewise.
13332         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13333         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13334         and move it to the end.  Cope with null def_stmt_outs.
13336 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13338         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13340 2018-06-29  Jeff Law  <law@redhat.com>
13342         * config/v850/v850.c (v850_legitimate_address_p): Handle large
13343         displacements for TARGET_V850E2V3 and newer.
13344         (TARGET_LRA_P): Remove.  Defaults to LRA now.
13345         * config/v850/v850.md (sign23byte_load): Remove.
13346         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13347         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13349 2018-06-29  Martin Liska  <mliska@suse.cz>
13351         PR lto/85759
13352         * coverage.c (coverage_init): Mangle full path name.
13353         * doc/invoke.texi: Document the change.
13354         * gcov-io.c (mangle_path): New.
13355         * gcov-io.h (mangle_path): Likewise.
13356         * gcov.c (mangle_name): Use mangle_path for path mangling.
13358 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13360         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13361         if starting source register is not even.
13363 2018-06-29  Martin Liska  <mliska@suse.cz>
13365         PR tree-optimization/86263
13366         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13367         Make edge redirection.
13369 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13371         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13372         (dump_scope_depth): New variable.
13373         (get_dump_scope_depth): New function.
13374         (dump_begin_scope): New function.
13375         (dump_end_scope): New function.
13376         * dumpfile.h (get_dump_scope_depth): New declaration.
13377         (dump_begin_scope): New declaration.
13378         (dump_end_scope): New declaration.
13379         (class auto_dump_scope): New class.
13380         (AUTO_DUMP_SCOPE): New macro.
13381         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13382         AUTO_DUMP_SCOPE.
13384 2018-06-29  Richard Biener  <rguenther@suse.de>
13386         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13387         compute_all_dependences succeeds.
13388         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13389         exceed --param loop-max-datarefs-for-datadeps.
13391 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
13393         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13395 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
13397         PR target/86348
13398         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13399         alternative 0 in preferred_for_speed attribute.
13401 2018-06-28  Paul Koning  <ni1d@arrl.net>
13403         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13404         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13405         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13406         * config/pdp11/pdp11.md: Correct "length" attribute calculation
13407         for shift insn patterns.
13409 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
13411         * cgraph.c (cgraph_node::get_body): Replace assignments to
13412         "dump_file" with calls to set_dump_file.
13413         * dumpfile.c (alt_dump_file): Make static, and group with...
13414         (alt_flags): ...this definition.
13415         (dumps_are_enabled): New variable.
13416         (refresh_dumps_are_enabled): New function.
13417         (set_dump_file): New function.
13418         (set_alt_dump_file): New function.
13419         (gcc::dump_manager::dump_start): Replace assignments to
13420         "dump_file" and "alt_dump_file" with calls to set_dump_file and
13421         set_alt_dump_file.
13422         (gcc::dump_manager::dump_finish): Likewise.
13423         * dumpfile.h (alt_dump_file): Delete decl.
13424         (dumps_are_enabled): New variable decl.
13425         (set_dump_file): New function decl.
13426         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13427         global.
13428         * tree-nested.c (lower_nested_functions): Replace assignments to
13429         "dump_file" with calls to set_dump_file.
13431 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13433         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13434         goto_locus of each outgoing edge of each basic block.
13436 2018-06-28  Richard Biener  <rguenther@suse.de>
13438         * dwarf2out.c (decl_scope_table): Remove.
13439         (push_decl_scope): Likewise.
13440         (pop_decl_scope): Likewise.
13441         (gen_type_die_for_member): Do not call push/pop_decl_scope.
13442         (gen_struct_or_union_type_die): Likewise.
13443         (gen_tagged_type_die): Likewise.
13444         (dwarf2out_init): Do not initialize decl_scope_table.
13445         (dwarf2out_c_finalize): Do not free it.
13447 2018-06-28  Richard Biener  <rguenther@suse.de>
13449         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13450         deciding whether to not re-use a DIE.
13452 2018-06-28  Richard Biener  <rguenther@suse.de>
13454         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13455         DW_AT_abstract_origin attribute.
13457 2018-06-28  Martin Liska  <mliska@suse.cz>
13459         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13460         Use newly introduced constants.
13461         * tree-switch-conversion.h (struct jump_table_cluster):
13462         Define max_ratio_for_size and max_ratio_for_speed.
13464 2018-06-28  Martin Liska  <mliska@suse.cz>
13466         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13467         Add new checking assert to catch invalid state.
13468         (jump_table_cluster::can_be_handled): Handle single case
13469         clusters.
13470         (jump_table_cluster::is_beneficial): Bail out for such case.
13471         (bit_test_cluster::find_bit_tests):
13472         Add new checking assert to catch invalid state.
13473         (bit_test_cluster::can_be_handled): Handle single case
13474         clusters.
13475         (bit_test_cluster::is_beneficial): Bail out for such case.
13476         (switch_decision_tree::analyze_switch_statement):
13477         Fix comment.
13479 2018-06-28  Martin Liska  <mliska@suse.cz>
13481         * common.opt: Introduce -completion option.
13482         * gcc.c (driver_handle_option): Handle it.
13483         (driver::main): Print completions if completion
13484         is set.
13485         * opt-suggestions.c (option_proposer::get_completions):
13486         New function.
13487         (option_proposer::suggest_completion): Likewise.
13488         (option_proposer::find_param_completions): Likewise.
13489         (verify_autocompletions): Likewise.
13490         (test_completion_valid_options): Likewise.
13491         (test_completion_valid_params): Likewise.
13492         (in_completion_p): Likewise.
13493         (empty_completion_p): Likewise.
13494         (test_completion_partial_match): Likewise.
13495         (test_completion_garbage): Likewise.
13496         (opt_proposer_c_tests): Likewise.
13497         * opt-suggestions.h: Declare new functions.
13498         * opts.c (common_handle_option): Handle OPT__completion_.
13499         * selftest-run-tests.c (selftest::run_tests): Add
13500         opt_proposer_c_tests.
13501         * selftest.c (assert_str_startswith): New.
13502         * selftest.h (assert_str_startswith): Likewise.
13503         (opt_proposer_c_tests): New.
13504         (ASSERT_STR_STARTSWITH): Likewise.
13506 2018-06-28  Martin Liska  <mliska@suse.cz>
13508         * Makefile.in: Add opt-suggestions.o.
13509         * gcc-main.c: Include opt-suggestions.h.
13510         * gcc.c (driver::driver): Likewise.
13511         (driver::~driver): Remove m_option_suggestions.
13512         (driver::build_option_suggestions): Moved to option_proposer.
13513         (driver::suggest_option): Likewise.
13514         (driver::handle_unrecognized_options): Use option_proposer.
13515         * gcc.h (class driver): Add new memver m_option_proposer.
13516         * opt-suggestions.c: New file.
13517         * opt-suggestions.h: New file.
13519 2018-06-28  Martin Liska  <mliska@suse.cz>
13521         * vec.h (class auto_string_vec): New (moved from auto_argvec).
13522         (auto_string_vec::~auto_string_vec): Likewise.
13524 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13526         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
13527         prevent_decl_creation_for_types fields up and add reset_location field.
13528         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
13529         statement if id->reset_location is true.
13530         (copy_edges_for_bb): Do not set goto_locus on the new edges if
13531         id->reset_location is true.
13532         (copy_phis_for_bb): Force input_location on the arguments if
13533         id->reset_location is true.
13534         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
13535         is set on the function to be inlined.
13537 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
13539         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
13541 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
13543         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
13544         registers for Pmode.
13545         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
13546         hard registers for the clobbered pseudo.
13548 2018-06-27  Paul Koning  <ni1d@arrl.net>
13550         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13551         mutually exclusive options.
13552         * config/pdp11/constraints.md (h): New constraint.
13553         (O): Update definition to match shift code generation.
13554         (D): New constraint.
13555         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
13556         (CCFP): Remove.
13557         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
13558         function.
13559         (output_jump): Change arguments.
13560         (pdp11_fixed_cc_regs): New function.
13561         (pdp11_cc_mode): Ditto.
13562         (pdp11_expand_shift): Ditto.
13563         (pdp11_assemble_shift): Ditto.
13564         (pdp11_small_shift): Ditto.
13565         (pdp11_branch_cost): Remove.
13566         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
13567         from output.
13568         (pdp11_register_move_cost): Update for CC registers.
13569         (pdp11_rtx_costs): Add case for LSHIFTRT.
13570         (pdp11_output_jump): Add CCNZ mode conditional branches.
13571         (notice_update_cc_on_set): Remove.
13572         (pdp11_cc_mode): New function.
13573         (simple_memory_operand): Correct pre/post decrement case.
13574         (no_side_effect_operand): New function.
13575         (pdp11_regno_reg_class): Add CC_REGS class.
13576         (pdp11_fixed_cc_regs): New function.
13577         (pdp11_small_shift): New function.
13578         (pdp11_expand_shift): New function to expand shift insns.
13579         (pdp11_assemble_shift): New function to output shifts.
13580         (pdp11_branch_cost): Remove.
13581         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
13582         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
13583         (WCHAR_TYPE): Ditto.
13584         (PTRDIFF_TYPE): Ditto.
13585         (ADJUST_INSN_LENGTH): New macro.
13586         (FIXED_REGISTERS): Add CC registers.
13587         (CALL_USED_REGISTERS): Ditto.
13588         (reg_class): Ditto.
13589         (REG_CLASS_NAMES): Ditto.
13590         (REG_CLASS_CONTENTS): Ditto.
13591         (SELECT_CC_MODE): Use new function.
13592         (TARGET_FLAGS_REGNUM): New macro.
13593         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
13594         (cc0_reg_rtx): Remove.
13595         (CC_STATUS_MDEP): Remove.
13596         (CC_STATUS_MDEFP_INIT): Remove.
13597         (CC_IN_FPU): Remove.
13598         (NOTICE_UPDATE_CC): Remove.
13599         (REGISTER_NAMES): Add CC registers.
13600         (BRANCH_COST): Change to constant 1.
13601         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
13602         handling.
13603         * config/pdp11/pdp11.opt (mbcopy): Remove.
13604         (mbcopy-builtin): Remove.
13605         (mbranch-cheap): Remove.
13606         (mbranch-expensive): Remove.
13607         * config/pdp11/predicates.md (expand_shift_operand): Update to
13608         match shift code generation.
13609         (ccnz_operator): New predicate.
13610         * doc/invoke.texi (PDP-11 Options): Remove deleted options
13611         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
13612         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
13613         exclusive options.
13614         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
13615         description of O constraint.
13617 2018-06-27  Jeff Law  <law@redhat.com>
13618             Austin Law  <austinklaw@gmail.com>
13620         * config/v850/v850.md (addsi3_set_flags): New pattern.
13621         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
13622         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
13623         (zero_extendhisi2_v850_set_flags): Likewise.
13624         (zero_extendqisi2_v850_set_flags): Likewise.
13625         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
13626         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
13627         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
13629         * config/v850/v850-protos.h (notice_update_cc): Remove.
13630         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
13631         (v850_print_operand): Handle 'D' and "d".
13632         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
13633         Add handling of arithmetic/logical operations compared against zero.
13634         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
13635         Do not look at v850_compare_op, instead get mode from last argument.
13636         (v850_gen_compare): Remove
13637         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
13638         after reload for prologue insns.
13639         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
13640         patterns.
13641         (construct_save_jarl): Likewise.
13642         (TARGET_FLAGS_REGNUM): Define.
13643         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
13644         (NOTICE_UPDATE_CC): Remove.
13645         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
13646         than cc0.  Conditionalize on reload_completed.
13647         (cmpsi_insn, setfcc_insn): Likewise.
13648         (tst1 splitter): Turn into define_and_split which sets the flags
13649         after reload.
13650         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
13651         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
13652         (cstoresf4, cstoredf4): Clobber the flags.
13653         (cmpsi, cmpsf, cmpdf): Remove expanders.
13654         (setf_insn): Remove pattern.
13655         (addsi3): Turn into define_and_split which clobbers the flags after
13656         reload and a suitable pattern (addsi3_clobber_flags) for use after
13657         reload.
13658         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
13659         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
13660         (ashrsi3, ashrsi3_v850e2): Likewise.
13661         (bins): Clobber the flags.
13662         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
13663         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
13664         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
13665         (call_value_internal_short, call_value_internal_long): Likewise.
13666         (callt_save_interrupt, callt_return_interrupt): Likewise.
13667         (save_interrupt, return_interrupt): Likewise.
13668         (callt_save_all_interrupt, save_all_interrupt): Likewise.
13669         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
13670         (restore_all_interrupt, _restore_all_interrupt): Likewise.
13671         (All FP comparisons): Only allow after reload has completed.
13672         (trfsr): Likewise.
13673         (divh, divhu): Tweak output template.
13674         (branch_z_normal, branch_z_invert): Remove
13675         (branch_nz_normal, branch_nz_invert): Likewise.
13676         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
13678         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
13679         * config/v850/v850.c (notice_update_cc): Remove.
13680         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
13681         (CC_NO_CARRY): Likewise.
13682         (NOTICE_UPDATE_CC): Define to nothing.
13683         * config/v850/v850.md: Remove block comment on cc0 handling
13684         Remove "cc" attribute from all patterns.  Remove cc_status handling
13685         from all patterns.  Minor formatting fixes.
13687 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13689         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
13690         (cortex-a76.cortex-a55): Likewise.
13691         * config/aarch64/aarch64-tune.md: Regenerate.
13692         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
13693         cortex-a76.cortex-a55.
13695 2018-06-27  Jeff Law  <law@redhat.com>
13697         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
13698         (MULTILIB_DIRNAMES): Similarly.
13700 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
13702         * gimple.h (gimple_return_retbnd): Delete.
13703         (gimple_return_set_retbnd): Likewise.
13704         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
13705         gimple_return_set_retbnd.
13706         * gimple-pretty-print.c (dump_gimple_return): Remove call to
13707         gimple_return_retbnd and adjust.
13708         * tree-inline.h (struct copy_body_data): Remove retbnd field.
13709         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
13710         Explicitly return NULL in a couple more cases.  Move assertion
13711         on debug statements and remove unreachable code.
13712         (reset_debug_binding): Do not test id->retbnd.
13713         (expand_call_inline): Do not set it.
13715 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
13717         * configure.ac: Add --disable-gcov option.
13718         * configure: Regenerate.
13719         * Makefile.in: Honour @enable_gcov@.
13720         * doc/install.texi: Document --disable-gcov.
13722 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13724         * config/arm/arm-cpus.in (cortex-a76): New entry.
13725         (cortex-a76.cortex-a55): Likewise.
13726         * config/arm/arm-tables.opt: Regenerate.
13727         * config/arm/arm-tune.md: Likewise.
13728         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
13729         * doc/invoke.texi (ARM Options): Document cortex-a76 and
13730         cortex-a76.cortex-a55.
13732 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
13734         PR target/85769
13735         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
13737 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
13739         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
13740         comment.
13741         (EPILOGUE_USES): Likewise.
13743 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
13745         * tree-inline.c (remap_location): New function extracted from...
13746         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
13747         (copy_phis_for_bb): ...here.  Call remap_location.
13748         (copy_cfg_body): Adjust call to copy_edges_for_bb.
13750 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
13752         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
13753         unaligned vsx for 16B memset.
13755 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
13757         PR target/86285
13758         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
13759         ieee128_float_type_node to long_double_type_node unless
13760         TARGET_LONG_DOUBLE_128 is set.
13762 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
13764         * cfgloop.c (get_loop_location): Convert return type from
13765         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
13766         by implicit construction from rtx_insn *, and using
13767         dump_user_location_t::from_function_decl for the fallback case.
13768         * cfgloop.h (get_loop_location): Convert return type from
13769         location_t to dump_user_location_t.
13770         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
13771         dump_printf_loc to pass in a dump_location_t rather than a
13772         location_t, via the gimple stmt.
13773         * coverage.c (get_coverage_counts): Update calls to
13774         dump_printf_loc to pass in dump_location_t rather than a
13775         location_t.
13776         * doc/optinfo.texi (Dump types): Convert example of
13777         dump_printf_loc from taking "locus" to taking "insn".  Update
13778         description of the "_loc" calls to cover dump_location_t.
13779         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
13780         "selftest.h".
13781         (dump_user_location_t::dump_user_location_t): New constructors,
13782         from gimple *stmt and rtx_insn *.
13783         (dump_user_location_t::from_function_decl): New function.
13784         (dump_loc): Make static.
13785         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
13786         const dump_location_t &.
13787         (dump_generic_expr_loc): Delete.
13788         (dump_printf_loc): Convert param "loc" from location_t to
13789         const dump_location_t &.
13790         (selftest::test_impl_location): New function.
13791         (selftest::dumpfile_c_tests): New function.
13792         * dumpfile.h: Include "profile-count.h".
13793         (class dump_user_location_t): New class.
13794         (struct dump_impl_location_t): New struct.
13795         (class dump_location_t): New class.
13796         (dump_printf_loc): Convert 2nd param from source_location to
13797         const dump_location_t &.
13798         (dump_generic_expr_loc): Delete.
13799         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
13800         const dump_location_t &.
13801         * gimple-fold.c (fold_gimple_assign): Update call to
13802         dump_printf_loc to pass in a dump_location_t rather than a
13803         location_t, via the gimple stmt.
13804         (gimple_fold_call): Likewise.
13805         * gimple-loop-interchange.cc
13806         (loop_cand::analyze_iloop_reduction_var): Update for change to
13807         check_reduction_path.
13808         (tree_loop_interchange::interchange): Update for change to
13809         find_loop_location.
13810         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
13811         change in return-type of find_loop_location.
13812         (graphite_regenerate_ast_isl): Likewise.
13813         * graphite-optimize-isl.c (optimize_isl): Likewise.
13814         * graphite.c (graphite_transform_loops): Likewise.
13815         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
13816         pass in a dump_location_t rather than a location_t, via the
13817         gimple stmt.
13818         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13819         * ipa.c (walk_polymorphic_call_targets): Likewise.
13820         * loop-unroll.c (report_unroll): Convert "locus" param from
13821         location_t to dump_location_t.
13822         (decide_unrolling): Update for change to get_loop_location's
13823         return type.
13824         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
13825         location_t to dump_user_location_t.
13826         (grid_find_single_omp_among_assignments_1): Updates calls to
13827         dump_printf_loc to pass in a dump_location_t rather than a
13828         location_t, via the gimple stmt.
13829         (grid_parallel_clauses_gridifiable): Convert "tloc" from
13830         location_t to dump_location_t.  Updates calls to dump_printf_loc
13831         to pass in a dump_location_t rather than a location_t, via the
13832         gimple stmt.
13833         (grid_inner_loop_gridifiable_p): Likewise.
13834         (grid_dist_follows_simple_pattern): Likewise.
13835         (grid_gfor_follows_tiling_pattern): Likewise.
13836         (grid_target_follows_gridifiable_pattern): Likewise.
13837         (grid_attempt_target_gridification): Convert initialization
13838         of local "grid" from memset to zero-initialization; FIXME: does
13839         this require C++11?  Update call to dump_printf_loc to pass in a
13840         optinfo_location rather than a location_t, via the gimple stmt.
13841         * profile.c (read_profile_edge_counts): Updates call to
13842         dump_printf_loc to pass in a dump_location_t rather than a
13843         location_t
13844         (compute_branch_probabilities): Likewise.
13845         * selftest-run-tests.c (selftest::run_tests): Call
13846         dumpfile_c_tests.
13847         * selftest.h (dumpfile_c_tests): New decl.
13848         * tree-loop-distribution.c (pass_loop_distribution::execute):
13849         Update for change in return type of find_loop_location.
13850         * tree-parloops.c (parallelize_loops): Likewise.
13851         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
13852         "locus" from location_t to dump_user_location_t.
13853         (canonicalize_loop_induction_variables): Likewise.
13854         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
13855         for change in return type of find_loop_location.
13856         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
13857         to dump_printf_loc to pass in a dump_location_t rather than a
13858         location_t, via the stmt.
13859         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
13860         Likewise.
13861         * tree-vect-loop-manip.c (find_loop_location): Convert return
13862         type from source_location to dump_user_location_t.
13863         (vect_do_peeling): Update for above change.
13864         (vect_loop_versioning): Update for change in type of
13865         vect_location.
13866         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
13867         from location_t to dump_user_location_t.
13868         (vect_estimate_min_profitable_iters): Update for change in type
13869         of vect_location.
13870         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
13871         location_t to dump_location_t.
13872         (vect_slp_bb): Update for change in type of vect_location.
13873         * tree-vectorizer.c (vect_location): Convert from source_location
13874         to dump_user_location_t.
13875         (try_vectorize_loop_1): Update for change in vect_location's type.
13876         (vectorize_loops): Likewise.
13877         (increase_alignment): Likewise.
13878         * tree-vectorizer.h (vect_location): Convert from source_location
13879         to dump_user_location_t.
13880         (find_loop_location): Convert return type from source_location to
13881         dump_user_location_t.
13882         (check_reduction_path): Convert 1st param from location_t to
13883         dump_user_location_t.
13884         * value-prof.c (check_counter): Update call to dump_printf_loc to
13885         pass in a dump_user_location_t rather than a location_t; update
13886         call to error_at for change in type of "locus".
13887         (check_ic_target): Update call to dump_printf_loc to
13888         pass in a dump_user_location_t rather than a location_t, via the
13889         call_stmt.
13891 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
13893         * config/s390/s390.h (enum processor_flags): Do not use
13894         default tune parameter when -march was specified.
13896 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
13898         PR target/86314
13899         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
13900         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
13901         operands.
13903 2018-06-26  Richard Biener  <rguenther@suse.de>
13905         PR tree-optimization/86287
13906         PR bootstrap/86316
13907         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
13908         (vect_analyze_loop): Initialize n_stmts.
13910 2018-06-26  Richard Biener  <rguenther@suse.de>
13912         PR middle-end/86271
13913         * fold-const.c (fold_convertible_p): Pointer extension
13914         isn't valid.
13916 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
13918         PR debug/86064
13919         * dwarf2out.c (loc_list_has_views): Adjust comments.
13920         (dw_loc_list): Split single cross-partition range with
13921         nonzero locview.
13923 2018-06-25  Jeff Law  <law@redhat.com>
13925         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
13926         on -mbig-switch by default.
13928         * config/v850/predicates.md (const_float_1_operand): Fix match_code
13929         test.
13930         (const_float_0_operand): Remove unused predicate.
13931         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
13932         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
13933         (recipsf2): New expander.  Original pattern now called
13934         (recipsf2_insn).
13935         (recipdf2, recipdf2_insn): Similarly.
13936         (rsqrtsf2, rsqrtsf2_insn): Similarly
13937         (rsqrtdf2, rsqrtdf2_insn): Similarly
13939 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
13941         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
13942         Simplify logic for FreeBSD (twice).
13944 2018-06-25  Martin Sebor  <msebor@redhat.com>
13946         PR tree-optimization/86204
13947         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
13948         a strnlen result if it's less than the length of the string.
13950 2018-06-25  Martin Sebor  <msebor@redhat.com>
13952         PR tree-optimization/85700
13953         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
13954         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
13955         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
13957 2018-06-25  Martin Sebor  <msebor@redhat.com>
13959         * doc/extend.texi (Zero-length arrays): Update and clarify.
13961 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
13963         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
13964         added IEEE/IBM long double multilib support on PowerPC little
13965         endian Linux systems.
13966         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
13967         (MULTILIB_DEFAULTS): Likewise.
13968         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13969         Likewise.
13970         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
13971         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
13972         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
13974 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
13976         PR middle-end/86311
13977         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
13978         (REORDER_45): Likewise.
13980 2018-06-25  Jeff Law  <law@redhat.com>
13982         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
13983         dividend to 32 bits.  Adjust length.
13984         (udivmodhi4): Cleanup output template.  Fix length.
13986 2018-06-25  Carl Love  <cel@us.ibm.com>
13988         * config/rs6000/vsx.md: Change word selector to prefered location.
13990 2018-06-25  Richard Biener  <rguenther@suse.de>
13992         PR tree-optimization/86304
13993         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
13994         epilogue-if-converted loops as well.
13996 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
13998         * lto-section-out.c (lto_begin_section): Do not print section
13999         name for noaddr and unnumbered dumps.
14001 2018-06-25  Richard Biener  <rguenther@suse.de>
14003         * tree-vectorizer.h (struct vec_info_shared): New structure
14004         with parts split out from struct vec_info and loop_nest from
14005         struct _loop_vec_info.
14006         (struct vec_info): Adjust accordingly.
14007         (struct _loop_vec_info): Likewise.
14008         (LOOP_VINFO_LOOP_NEST): Adjust.
14009         (LOOP_VINFO_DATAREFS): Likewise.
14010         (LOOP_VINFO_DDRS): Likewise.
14011         (struct _bb_vec_info): Likewise.
14012         (BB_VINFO_DATAREFS): Likewise.
14013         (BB_VINFO_DDRS): Likewise.
14014         (struct _stmt_vec_info): Add dr_aux member.
14015         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14016         (DR_MISALIGNMENT_UNINITIALIZED): New.
14017         (set_dr_misalignment): Adjust.
14018         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14019         (vect_analyze_loop): Adjust prototype.
14020         (vect_analyze_loop_form): Likewise.
14021         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14022         Compute dependences lazily.
14023         (vect_record_base_alignments): Use shared datarefs/ddrs.
14024         (vect_verify_datarefs_alignment): Likewise.
14025         (vect_analyze_data_refs_alignment): Likewise.
14026         (vect_analyze_data_ref_accesses): Likewise.
14027         (vect_analyze_data_refs): Likewise.
14028         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14029         constructor parameter for shared part.
14030         (vect_analyze_loop_form): Pass in shared part and adjust.
14031         (vect_analyze_loop_2): Pass in storage for the number of
14032         stmts.  Move loop nest finding to the caller.  Compute
14033         datarefs lazily.
14034         (vect_analyze_loop): Pass in shared part.
14035         (vect_transform_loop): Verify shared datarefs are unchanged.
14036         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14037         constructor parameter for shared part.
14038         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14039         (vect_slp_bb): Verify shared datarefs are unchanged before
14040         transform.
14041         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14042         change.
14043         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14044         DR_MISALIGNMENT_UNINITIALIZED.
14045         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14046         parameter for shared part.
14047         (vec_info::~vec_info): Adjust.
14048         (vec_info_shared::vec_info_shared): New.
14049         (vec_info_shared::~vec_info_shared): Likewise.
14050         (vec_info_shared::save_datarefs): Likewise.
14051         (vec_info_shared::check_datarefs): Likewise.
14052         (try_vectorize_loop_1): Construct shared part live for analyses
14053         of a single loop for multiple vector sizes.
14054         * tree-parloops.c (gather_scalar_reductions): Adjust.
14056 2018-06-25  Richard Biener  <rguenther@suse.de>
14058         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14059         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14060         (vect_analyze_data_refs): Remove similar code from here and
14061         simplify accordingly.
14063 2018-06-25  Richard Biener  <rguenther@suse.de>
14065         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14066         for reverse storage order accesses rather than asserting
14067         they cannot happen here.
14069 2018-06-25  Tom de Vries  <tdevries@suse.de>
14071         PR debug/86257
14072         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14073         Use data16 instead of .byte for insn prefix.
14075 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14077         PR C++/86082
14078         * parser.c (make_char_string_pack): Pass this literal chars
14079         through cpp_interpret_string.
14080         (cp_parser_userdef_numeric_literal): Check the result of
14081         make_char_string_pack.
14083 2018-06-24  Maya Rashish  <coypu@sdf.org>
14085         * ginclude/stddef.h: Simplify conditions around avoiding
14086         re-definition of __size_t.
14088 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14090         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14091         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14093 2018-06-22  Maya Rashish  <coypu@sdf.org>
14095         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14096         made to OpenBSD and NetBSD through the years.
14098 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14100         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14101         behavior of vec_pack (vector double, vector double) to match
14102         behavior of vec_float2 (vector double, vector double).
14104 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14106         * gimplify.c (gimplify_function_tree): Prevent creation
14107         of a trampoline for the address of the current function
14108         passed to entry/exit instrumentation hooks.
14110 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14112         PR target/86222
14113         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14114         correctly.
14116 2018-06-22  Martin Liska  <mliska@suse.cz>
14118         PR tree-optimization/86263
14119         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14120         Bail out if is_enabled is false.
14121         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14122         New declaration.
14123         (jump_table_cluster::is_enabled): New function.
14125 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14127         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14128         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14129         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14130         (lto_input_ts_binfo_tree_pointers): Likewise.
14131         * tree-streamer-out.c (streamer_write_tree_bitfields,
14132         write_ts_binfo_tree_pointers): Likewise.
14133         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14135 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14137         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14139 2018-06-22  Martin Liska  <mliska@suse.cz>
14141         * symbol-summary.h (get): Make it pure and inline move
14142         functionality from ::get function.
14143         (get): Remove and inline into ::get and ::get_create.
14144         (get_create): Move code from ::get function.
14146 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14148         PR target/85994
14149         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14150         -x assembler-with-cpp.
14152 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14154         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14155         _FILE_OFFSET_BITS=64 for C++.
14157 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
14159         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14160         conversion insn that shows up when pr85657-3.c is compiled using
14161         IEEE 128-bit long double.
14162         (neg<mode>2_internal): Use the correct mode to check whether the
14163         mode is IBM extended.
14164         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14165         multiply and divide external functions from being created more
14166         than once.
14168 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14170         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14171         functions.
14172         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
14173         the edge can be forwarded.
14174         (cfg_layout_merge_blocks): Likewise.
14176 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14178         * except.c (finish_eh_generation): Commit edge insertions only after
14179         the EH edges have been redirected from post-landing to landing pads.
14181 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14183         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14184         create_tmp_var_for to create the FRAME decl.
14185         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14187 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14189         * tree-inline.c (copy_edges_for_bb): Minor tweak.
14190         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14191         debug statement when resetting its value.
14192         (expand_call_inline): Copy the locus of the call onto the assignment
14193         of the return value, if any.  Use local variable in more cases.
14195 2018-06-21  Martin Liska  <mliska@suse.cz>
14197         * ipa-pure-const.c (propagate_nothrow): Use
14198         funct_state_summaries->get.
14199         (dump_malloc_lattice): Likewise.
14200         (propagate_malloc): Likewise.
14202 2018-06-21  Richard Biener  <rguenther@suse.de>
14204         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14205         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14206         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14207         comment.
14208         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14209         BLOCK_ABSTRACT_ORIGIN unconditionally.
14211 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14213         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14214         deleting it.
14215         * ipa-reference.c (ipa_reference_c_finalize): Delete
14216         ipa_ref_opt_sum_summaries and set it to NULL.
14218 2018-06-21  Tom de Vries  <tdevries@suse.de>
14220         PR tree-optimization/85859
14221         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14222         test with comment from bb_no_side_effects_p.
14224 2018-06-21  Richard Biener  <rguenther@suse.de>
14226         PR tree-optimization/86232
14227         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14228         max for constant niter.
14230 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14232         * config/aarch64/aarch64-simd.md
14233         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14235 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14237         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14238         Make opernads of the unspec commutative.
14240 2018-06-21  Richard Biener  <rguenther@suse.de>
14242         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14243         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14244         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14245         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14246         NULL DR_STEP.
14247         (vect_record_base_alignments): Do not record base alignment
14248         for gathers or scatters.
14249         (vect_compute_data_ref_alignment): Drop return value that is always
14250         true.  Bail out early for gathers or scatters.
14251         (vect_enhance_data_refs_alignment): Bail out early for gathers
14252         or scatters.
14253         (vect_find_same_alignment_drs): Likewise.
14254         (vect_analyze_data_refs_alignment): Remove dead code.
14255         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14256         (vect_analyze_data_refs): For possible gathers or scatters do
14257         not create an alternate DR, just check their possible validity
14258         and mark them.  Adjust DECL_NONALIASED handling to not rely
14259         on DR_BASE_ADDRESS.
14260         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14261         update inits of gathers or scatters.
14262         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14263         Also copy gather/scatter flag to pattern vinfo.
14265 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14267         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14268         behavior of vec_packsu (vector unsigned long long, vector unsigned
14269         long long) to match behavior of vec_packs with same signature.
14271 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
14272             Thomas Schwinge <thomas@codesourcery.com>
14273             Cesar Philippidis  <cesar@codesourcery.com>
14275         * gimplify.c (gimplify_scan_omp_clauses): Add support for
14276         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14277         (gimplify_adjust_omp_clauses): Likewise.
14278         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14279         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14280         (gimplify_omp_target_update): Update handling of acc update and
14281         enter/exit data.
14282         * omp-low.c (install_var_field): Remove unused parameter
14283         base_pointers_restrict.
14284         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14285         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14286         FINALIZE}
14287         (omp_target_base_pointers_restrict_p): Delete.
14288         (scan_omp_target): Update call to scan_sharing_clauses.
14289         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14290         FINALIZE}.
14291         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14292         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14293         (convert_local_omp_clauses): Likewise.
14294         * tree-pretty-print.c (dump_omp_clause): Likewise.
14295         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
14296         FINALIZE}.
14297         (omp_clause_code_name): Likewise.
14299 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14301         PR debug/86194
14302         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14303         be narrowed.
14305         PR tree-optimization/86231
14306         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
14307         anti-range don't overwrite *vr0min before using it to compute *vr0max.
14309 2018-06-20  Tom de Vries  <tdevries@suse.de>
14311         PR tree-optimization/86097
14312         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14313         iv type if signedness of iv type is not the same as that of *nit.
14315 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14317         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
14318         EDGE_EH edges, verify they are all EDGE_EH.
14320 2018-06-20  Maya Rashish  <coypu@sdf.org>
14322         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14324 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14326         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14327         * config/aarch64/aarch64.c (xgene1_tunings): Add
14328         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14329         (aarch64_mode_valid_for_sched_fusion_p):
14330         Allow 16-byte modes.
14331         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14332         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14333         128-bit modes.
14334         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14335         New pattern.
14336         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14337         * config/aarch64/iterators.md (VQ2): New mode iterator.
14339 2018-06-20  Martin Liska  <mliska@suse.cz>
14341         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14342         Change default ratio from 10 to 8.
14344 2018-06-20  Martin Liska  <mliska@suse.cz>
14346         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14347         New.
14348         (bit_test_cluster::find_bit_tests): Likewise.
14349         (switch_decision_tree::analyze_switch_statement): Find clusters.
14350         * tree-switch-conversion.h (struct jump_table_cluster): Document
14351         hierarchy.
14353 2018-06-20  Martin Liska  <mliska@suse.cz>
14355         * tree-switch-conversion.c (switch_conversion::collect):
14356         Record m_uniq property.
14357         (switch_conversion::expand): Bail out for special conditions.
14358         (group_cluster::~group_cluster): New.
14359         (group_cluster::group_cluster): Likewise.
14360         (group_cluster::dump): Likewise.
14361         (jump_table_cluster::emit): New.
14362         (switch_decision_tree::fix_phi_operands_for_edges): New.
14363         (struct case_node): Remove struct.
14364         (jump_table_cluster::can_be_handled): New.
14365         (case_values_threshold): Moved to header.
14366         (reset_out_edges_aux): Likewise.
14367         (jump_table_cluster::is_beneficial): New.
14368         (bit_test_cluster::can_be_handled): Likewise.
14369         (add_case_node): Remove.
14370         (bit_test_cluster::is_beneficial): New.
14371         (case_bit_test::cmp): New.
14372         (bit_test_cluster::emit): New.
14373         (expand_switch_as_decision_tree_p): Remove.
14374         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14375         (fix_phi_operands_for_edge): Likewise.
14376         (switch_decision_tree::analyze_switch_statement): New.
14377         (compute_cases_per_edge): Move ...
14378         (switch_decision_tree::compute_cases_per_edge): ... here.
14379         (try_switch_expansion): Likewise.
14380         (switch_decision_tree::try_switch_expansion): Likewise.
14381         (record_phi_operand_mapping): Likewise.
14382         (switch_decision_tree::record_phi_operand_mapping): Likewise.
14383         (emit_case_decision_tree): Likewise.
14384         (switch_decision_tree::emit): Likewise.
14385         (balance_case_nodes): Likewise.
14386         (switch_decision_tree::balance_case_nodes): Likewise.
14387         (dump_case_nodes): Likewise.
14388         (switch_decision_tree::dump_case_nodes): Likewise.
14389         (emit_jump): Likewise.
14390         (switch_decision_tree::emit_jump): Likewise.
14391         (emit_cmp_and_jump_insns): Likewise.
14392         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14393         (emit_case_nodes): Likewise.
14394         (switch_decision_tree::emit_case_nodes): Likewise.
14395         (conditional_probability): Remove.
14396         * tree-switch-conversion.h (enum cluster_type): New.
14397         (PRINT_CASE): New.
14398         (struct cluster): Likewise.
14399         (cluster::cluster): Likewise.
14400         (struct simple_cluster): Likewise.
14401         (simple_cluster::simple_cluster): Likewise.
14402         (struct group_cluster): Likewise.
14403         (struct jump_table_cluster): Likewise.
14404         (struct bit_test_cluster): Likewise.
14405         (struct min_cluster_item): Likewise.
14406         (struct case_tree_node): Likewise.
14407         (case_tree_node::case_tree_node): Likewise.
14408         (jump_table_cluster::case_values_threshold): Likewise.
14409         (struct case_bit_test): Likewise.
14410         (struct switch_decision_tree): Likewise.
14411         (struct switch_conversion): Likewise.
14412         (switch_decision_tree::reset_out_edges_aux): Likewise.
14414 2018-06-20  Martin Liska  <mliska@suse.cz>
14416         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14417         (hoist_edge_and_branch_if_true): Likewise.
14418         (expand_switch_using_bit_tests_p): Likewise.
14419         (struct case_bit_test): Likewise.
14420         (case_bit_test_cmp): Likewise.
14421         (emit_case_bit_tests): Likewise.
14422         (switch_conversion::switch_conversion): New class.
14423         (struct switch_conv_info): Remove old struct.
14424         (collect_switch_conv_info): More to ...
14425         (switch_conversion::collect): ... this.
14426         (check_range): Likewise.
14427         (switch_conversion::check_range): Likewise.
14428         (check_all_empty_except_final): Likewise.
14429         (switch_conversion::check_all_empty_except_final): Likewise.
14430         (check_final_bb): Likewise.
14431         (switch_conversion::check_final_bb): Likewise.
14432         (create_temp_arrays): Likewise.
14433         (switch_conversion::create_temp_arrays): Likewise.
14434         (free_temp_arrays): Likewise.
14435         (gather_default_values): Likewise.
14436         (switch_conversion::gather_default_values): Likewise.
14437         (build_constructors): Likewise.
14438         (switch_conversion::build_constructors): Likewise.
14439         (constructor_contains_same_values_p): Likewise.
14440         (switch_conversion::contains_same_values_p): Likewise.
14441         (array_value_type): Likewise.
14442         (switch_conversion::array_value_type): Likewise.
14443         (build_one_array): Likewise.
14444         (switch_conversion::build_one_array): Likewise.
14445         (build_arrays): Likewise.
14446         (switch_conversion::build_arrays): Likewise.
14447         (gen_def_assigns): Likewise.
14448         (switch_conversion::gen_def_assigns): Likewise.
14449         (prune_bbs): Likewise.
14450         (switch_conversion::prune_bbs): Likewise.
14451         (fix_phi_nodes): Likewise.
14452         (switch_conversion::fix_phi_nodes): Likewise.
14453         (gen_inbound_check): Likewise.
14454         (switch_conversion::gen_inbound_check): Likewise.
14455         (process_switch): Use the newly created class.
14456         (switch_conversion::expand): New.
14457         (switch_conversion::~switch_conversion): New.
14458         * tree-switch-conversion.h: New file.
14460 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14462         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14463         tree-vect-patterns.c.
14464         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14465         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
14466         (vect_recog_sad_pattern): Likewise.
14467         (vect_recog_widen_sum_pattern): Likewise.
14468         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
14469         (vect_recog_widen_shift_pattern): Remove the type_in argument.
14470         (vect_recog_rotate_pattern): Likewise.
14471         (vect_recog_mult_pattern): Likewise.
14472         (vect_recog_vector_vector_shift_pattern): Likewise.
14473         (vect_recog_divmod_pattern): Likewise.
14474         (vect_recog_mixed_size_cond_pattern): Likewise.
14475         (vect_recog_bool_pattern): Likewise.
14476         (vect_recog_mask_conversion_pattern): Likewise.
14477         (vect_try_gather_scatter_pattern): Likewise.
14478         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
14479         (vect_recog_over_widening_pattern): Likewise.
14480         (vect_recog_gather_scatter_pattern): Likewise.
14481         (vect_recog_func_ptr): Move from tree-vectorizer.h
14482         (vect_vect_recog_func_ptrs): Move further down the file.
14483         (vect_recog_func): Likewise.  Remove the third argument.
14484         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14485         (vect_pattern_recog_1): Expect the pattern function to do any
14486         necessary target tests.  Also expect it to provide a vector type.
14487         Remove the type_in handling.
14489 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14491         * tree-vect-patterns.c (vect_pattern_detected): New function.
14492         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14493         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14494         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14495         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14496         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14497         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14498         (vect_recog_mask_conversion_pattern)
14499         (vect_try_gather_scatter_pattern): Likewise.
14501 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14503         * tree-vect-patterns.c (vect_get_internal_def): New function.
14504         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14505         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14506         (search_type_for_mask_1): Use it.
14508 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14510         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14511         redundant WIDEN_SUM_EXPR handling.
14512         (vect_recog_sad_pattern): Likewise.
14514 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14516         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14517         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14518         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14519         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14520         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14522 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14524         * tree-vect-stmts.c (vectorizable_call): Make sure that we
14525         use the stmt_vec_info of the original bb statement for the
14526         new zero assignment, even if the call is part of a pattern.
14528 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14530         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
14531         that the sequence is attached to the original statement rather
14532         than the pattern statement.
14533         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
14534         PATTERN_DEF_SEQ from the original statement rather than
14535         the main pattern statement.
14536         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
14537         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
14538         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
14540 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14542         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
14543         definition statements before the early exit for statements that aren't
14544         live or relevant.
14545         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
14546         split out from...
14547         (vect_transform_loop): ...here.  Process pattern definition
14548         statements without first checking whether the main pattern
14549         statement is live or relevant.
14551 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14553         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
14554         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
14556 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
14558         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
14559         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
14560         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
14561         (expand_block_compare): Change select_block_compare_mode call.
14562         (expand_strncmp_align_check): Use new functions, fix comment.
14563         (emit_final_str_compare_gpr): New function.
14564         (expand_strn_compare): Refactor and clean up code.
14565         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
14567 2018-06-19  Tony Reix  <tony.reix@atos.com>
14568             Damien Bergamini  <damien.bergamini@atos.com>
14569             David Edelsohn  <dje.gcc@gmail.com>
14571         * collect2.c (static_obj): New variable.
14572         (static_libs): New variable.
14573         (is_in_list): Uncomment declaration.
14574         (main): Track AIX libraries linked statically.
14575         (is_in_list): Uncomment definition.
14576         (scan_prog_file): Don't add AIX shared libraries initializer
14577         to constructor list if linking statically.
14579 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
14581         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
14582         constant.
14583         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
14585 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
14587         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
14588         blocks.
14590 2018-06-19  Martin Liska  <mliska@suse.cz>
14592         * config/i386/i386.c (ix86_can_inline_p): Do not use
14593         ipa_fn_summaries::get_create.
14594         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
14595         get.
14596         (devirtualization_time_bonus): Likewise.
14597         (ipcp_propagate_stage): Likewise.
14598         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14599         (edge_set_predicate): Likewise.
14600         (evaluate_conditions_for_known_args): Likewise.
14601         (evaluate_properties_for_edge): Likewise.
14602         (ipa_call_summary::reset): Tranform to ...
14603         (ipa_call_summary::~ipa_call_summary): ... this.
14604         (ipa_fn_summary::reset): Transform to ...
14605         (ipa_fn_summary::~ipa_fn_summary): ... this.
14606         (ipa_fn_summary_t::remove): Rename to ...
14607         (ipa_fn_summary_t::remove_callees): ... this.
14608         (ipa_fn_summary_t::duplicate): Use placement new
14609         instead of memory copy.
14610         (ipa_call_summary_t::duplicate): Likewise.
14611         (ipa_call_summary_t::remove): Remove.
14612         (dump_ipa_call_summary): Change get_create to get.
14613         (ipa_dump_fn_summary): Dump only when summary exists.
14614         (analyze_function_body): Use symbol_summary::get instead
14615         of get_create.
14616         (compute_fn_summary): Likewise.
14617         (estimate_edge_devirt_benefit): Likewise.
14618         (estimate_edge_size_and_time): Likewise.
14619         (inline_update_callee_summaries): Likewise.
14620         (remap_edge_change_prob): Likewise.
14621         (remap_edge_summaries): Likewise.
14622         (ipa_merge_fn_summary_after_inlining): Likewise.
14623         (write_ipa_call_summary): Likewise.
14624         (ipa_fn_summary_write): Likewise.
14625         (ipa_free_fn_summary): Likewise.
14626         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
14627         (struct ipa_call_summary): Likewise.
14628         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
14629         of get_create.
14630         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
14631         (estimate_size_after_inlining): Likewise.
14632         (estimate_growth): Likewise.
14633         (growth_likely_positive): Likewise.
14634         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14635         (inline_call): Likewise.
14636         * ipa-inline.c (caller_growth_limits): Likewise.
14637         (can_inline_edge_p): Likewise.
14638         (can_inline_edge_by_limits_p): Likewise.
14639         (compute_uninlined_call_time): Likewise.
14640         (compute_inlined_call_time): Likewise.
14641         (want_inline_small_function_p): Likewise.
14642         (edge_badness): Likewise.
14643         (update_caller_keys): Likewise.
14644         (update_callee_keys): Likewise.
14645         (inline_small_functions): Likewise.
14646         (inline_to_all_callers_1): Likewise.
14647         (dump_overall_stats): Likewise.
14648         (early_inline_small_functions): Likewise.
14649         (early_inliner): Likewise.
14650         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14651         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14652         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14653         * ipa-split.c (execute_split_functions): Likewise.
14654         * symbol-summary.h: Likewise.
14655         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14657 2018-06-19  Richard Biener  <rguenther@suse.de>
14659         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
14660         (vectorize_loops): ... here.  Fix dbgcnt handling.
14661         (try_vectorize_loop): Wrap try_vectorize_loop_1.
14663 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
14665         PR target/86197
14666         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
14667         ieee128 argument takes up only one (vector) register, not two (floating
14668         point) registers.
14670 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14672         * gimplify.c (gimplify_init_constructor): Really never clear for an
14673         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
14675 2018-06-19  Richard Biener  <rguenther@suse.de>
14677         PR tree-optimization/86179
14678         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
14679         after failed recognition.
14681 2018-06-18  Martin Sebor  <msebor@redhat.com>
14683         PR middle-end/85602
14684         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
14685         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
14686         Handle integer subtraction.
14687         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
14688         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
14690 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14692         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
14693         param from rtx to rtx_insn *.
14694         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
14695         param.
14696         (frv_ifcvt_modify_insn): Likwise.
14697         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
14698         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
14699         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
14700         as_a <rtx_insn *> cast to local "unprotected_region" once
14701         it's been established that it's not NULL or pc_rtx.
14702         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
14703         param "sethi" from rtx to rtx_insn *.
14704         (nds32_group_float_insns): Likewise for param "insn".
14705         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
14706         param.
14707         (vax_output_int_subtract): Likewise.
14708         * config/vax/vax.c (vax_output_int_add): Likewise for param
14709         "insn".
14710         (vax_output_int_subtract): Likewise.
14711         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
14712         (emit_pattern_after): Likewise for param "after".
14713         (emit_insn_after): Likewise.
14714         (emit_jump_insn_after): Likewise.
14715         (emit_call_insn_after): Likewise.
14716         (emit_debug_insn_after): Likewise.
14717         (emit_pattern_before): Likewise for param "before".
14718         (emit_insn_before): Likewise.
14719         (emit_jump_insn_before): Likewise.
14720         * final.c (get_insn_template): Likewise for param "insn", removing
14721         a cast.
14722         * output.h (get_insn_template): Likewise for 2nd param.
14723         * rtl.h (emit_insn_before): Likewise.
14724         (emit_jump_insn_before): Likewise.
14725         (emit_debug_insn_before_noloc): Likewise.
14726         (emit_insn_after): Likewise.
14727         (emit_jump_insn_after): Likewise.
14728         (emit_call_insn_after): Likewise.
14729         (emit_debug_insn_after): Likewise.
14730         (set_insn_deleted): Likewise for param.
14732 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
14734         PR target/85358
14735         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
14736         floating point modes, so that IFmode is numerically greater than
14737         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
14738         to declare the ordering.  This prevents IFmode from being
14739         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
14740         machine.  Include rs6000-modes.h to share the fractional values
14741         between genmodes* and the rest of the compiler.
14742         (IFmode): Likewise.
14743         (KFmode): Likewise.
14744         (TFmode): Likewise.
14745         * config/rs6000/rs6000-modes.h: New file.
14746         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
14747         meaning of rs6000_long_double_size so that 126..128 selects an
14748         appropriate 128-bit floating point type.
14749         (rs6000_option_override_internal): Likewise.
14750         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
14751         (TARGET_LONG_DOUBLE_128): Change the meaning of
14752         rs6000_long_double_size so that 126..128 selects an appropriate
14753         128-bit floating point type.
14754         (LONG_DOUBLE_TYPE_SIZE): Update comment.
14755         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
14756         source and destination to match the standard usage.
14757         (truncifkf2): Likewise.
14758         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
14759         ISA 2.07 to use an explicit clobber, instead of passing in a
14760         temporary.
14761         (copysign<mode>3_soft): Likewise.
14763 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14765         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14766         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
14767         (vect_slp_analyze_instance_dependence): Likewise.
14768         (vect_enhance_data_refs_alignment): Likewise.
14769         (vect_analyze_data_refs_alignment): Likewise.
14770         (vect_slp_analyze_and_verify_instance_alignment
14771         (vect_analyze_data_ref_accesses): Likewise.
14772         (vect_prune_runtime_alias_test_list): Likewise.
14773         (vect_analyze_data_refs): Likewise.
14774         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
14775         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14776         (vect_analyze_scalar_cycles_1): Likewise.
14777         (vect_get_loop_niters): Likewise.
14778         (vect_analyze_loop_form_1): Likewise.
14779         (vect_update_vf_for_slp): Likewise.
14780         (vect_analyze_loop_operations): Likewise.
14781         (vect_analyze_loop): Likewise.
14782         (vectorizable_induction): Likewise.
14783         (vect_transform_loop): Likewise.
14784         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
14785         * tree-vect-slp.c (vect_analyze_slp): Likewise.
14786         (vect_make_slp_decision): Likewise.
14787         (vect_detect_hybrid_slp): Likewise.
14788         (vect_slp_analyze_operations): Likewise.
14789         (vect_slp_bb): Likewise.
14790         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
14791         (vectorizable_bswap): Likewise.
14792         (vectorizable_call): Likewise.
14793         (vectorizable_simd_clone_call): Likewise.
14794         (vectorizable_conversion): Likewise.
14795         (vectorizable_assignment): Likewise.
14796         (vectorizable_shift): Likewise.
14797         (vectorizable_operation): Likewise.
14798         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
14800 2018-06-18  Martin Sebor  <msebor@redhat.com>
14802         PR tree-optimization/81384
14803         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
14804         * builtins.c (expand_builtin_strnlen): New function.
14805         (expand_builtin): Call it.
14806         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
14807         * builtins.def (BUILT_IN_STRNLEN): New.
14808         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
14809         Warn for bounds in excess of maximum object size.
14810         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
14811         single-value ranges.  Handle strnlen.
14812         (handle_builtin_strlen): Handle strnlen.
14813         (strlen_check_and_optimize_stmt): Same.
14814         * doc/extend.texi (Other Builtins): Document strnlen.
14816 2018-06-18  Maya Rashish  <coypu@sdf.org>
14818         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
14819         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
14820         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
14822         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
14823         here to ...
14824         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
14826 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14828         * tree.c (escaped_string::escape): Replace cast to char * by
14829         const_cast<char *> (unescaped).
14831 2018-06-18  Nick Clifton  <nickc@redhat.com>
14833         PR 84195
14834         * tree.c (escaped_string): New class.  Converts an unescaped
14835         string into its escaped equivalent.
14836         (warn_deprecated_use): Use the new class to convert the
14837         deprecation message, if present.
14838         (test_escaped_strings): New self test.
14839         (test_c_tests): Add test_escaped_strings.
14840         * doc/extend.texi (deprecated): Add a note that the
14841         deprecation message is affected by the -fmessage-length
14842         option, and that control characters will be escaped.
14843         (#pragma GCC error): Document this pragma.
14844         (#pragma GCC warning): Likewise.
14845         * doc/invoke.texi (-fmessage-length): Document this option's
14846         effect on the #warning and #error preprocessor directives and
14847         the deprecated attribute.
14849 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
14851         * tree.c (decl_value_expr_lookup): Revert latest change.
14852         (decl_value_expr_insert): Likewise.
14854 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
14856         * gimplify.c (nonlocal_vlas): Delete.
14857         (nonlocal_vla_vars): Likewise.
14858         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
14859         referenced VLAs.
14860         (gimplify_body): Do not create and destroy nonlocal_vlas.
14861         * tree-nested.c: Include diagnostic.h.
14862         (use_pointer_in_frame): Tweak.
14863         (lookup_field_for_decl): Add assertion and declare the transformation.
14864         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
14865         internal error when the reference is in a wrong context.  Do not
14866         create a debug decl by default.
14867         (note_nonlocal_block_vlas): Delete.
14868         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
14869         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
14870         create a debug decl by default.
14871         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
14872         call is in a wrong context.
14873         (fixup_vla_decls): New function.
14874         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
14875         debug variables were created.
14876         * tree.c (decl_value_expr_lookup): Add checking assertion.
14877         (decl_value_expr_insert): Likewise.
14879 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14881         PR middle-end/82479
14882         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
14883         * tree-scalar-evolution.c (interpret_expr): Likewise.
14884         (expression_expensive_p): Likewise.
14885         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
14886         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
14887         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
14888         (ssa_defined_by_minus_one_stmt_p): New.
14890 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14892         PR middle-end/64946
14893         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
14894         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14895         * dojump.c (do_jump): Likewise.
14896         * expr.c (expand_expr_real_2): Check operand type's sign.
14897         * fold-const.c (const_unop): Handle ABSU_EXPR.
14898         (fold_abs_const): Likewise.
14899         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
14900         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
14901         (strip_sign_op_1): Likesise.
14902         * match.pd: Add new pattern to generate ABSU_EXPR.
14903         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
14904         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14905         * tree-eh.c (operation_could_trap_helper_p): Likewise.
14906         * tree-inline.c (estimate_operator_cost): Likewise.
14907         * tree-pretty-print.c (dump_generic_node): Likewise.
14908         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
14909         * tree.def (ABSU_EXPR): New.
14911 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
14913         PR middle-end/86095
14914         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
14915         documented as preserved for backward compatibility only.
14916         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
14918         PR rtl-optimization/86108
14919         * bb-reorder.c (create_forwarder_block): Renamed to ...
14920         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
14921         jump from new landing pad to the second part.
14922         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
14923         Adjust callers.
14925 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
14927         PR middle-end/85878
14928         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
14929         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
14930         Only call store_expr for halves if the mode is the same.
14932         PR middle-end/86123
14933         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
14934         Fix up comment formatting.
14936 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14938         * typed-splay-tree.h (typed_splay_tree::remove): New function.
14939         (typed_splay_tree::closure,
14940         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
14941         (typed_splay_tree::typed_splay_tree,
14942         typed_splay_tree::operator =): Declared private.
14943         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
14944         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
14945         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
14946         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
14947         typed_splay_tree::splay_tree_splay,
14948         typed_splay_tree::splay_tree_foreach_helper,
14949         typed_splay_tree::splay_tree_insert,
14950         typed_splay_tree::splay_tree_remove,
14951         typed_splay_tree::splay_tree_lookup,
14952         typed_splay_tree::splay_tree_predecessor,
14953         typed_splay_tree::splay_tree_successor,
14954         typed_splay_tree::splay_tree_min,
14955         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
14956         (typed_splay_tree::root, typed_splay_tree::comp,
14957         typed_splay_tree::delete_key,
14958         typed_splay_tree::delete_value): New data members.
14959         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
14960         typed_splay_tree::remove.
14962 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
14964         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
14965         -mginv and -mno-ginv to the assembler.
14966         * config/mips/mips.opt (-mcrc): New option.
14967         (-mginv): Likewise.
14968         * doc/invoke.text (-mcrc): Document.
14969         (-mginv): Likewise.
14971 2018-06-15  Nick Clifton  <nickc@redhat.com>
14973         PR 84195
14974         * tree.c (escaped_string): New class.  Converts an unescaped
14975         string into its escaped equivalent.
14976         (warn_deprecated_use): Use the new class to convert the
14977         deprecation message, if present.
14978         (test_escaped_strings): New self test.
14979         (test_c_tests): Add test_escaped_strings.
14980         * doc/extend.texi (deprecated): Add a note that the
14981         deprecation message is affected by the -fmessage-length
14982         option, and that control characters will be escaped.
14983         (#pragma GCC error): Document this pragma.
14984         (#pragma GCC warning): Likewise.
14985         * doc/invoke.texi (-fmessage-length): Document this option's
14986         effect on the #warning and #error preprocessor directives and
14987         the deprecated attribute.
14989 2018-06-15  Richard Biener  <rguenther@suse.de>
14991         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
14992         here, also noting vector size used.
14993         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
14994         size used in MSG_OPTIMIZED_LOCATIONS dump.
14995         (pass_slp_vectorize::execute): Adjust.
14997 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
14999         PR target/85968
15000         * config/arc/arc.c (arc_return_address_register): Fix
15001         if-condition.
15003 2018-06-15  Richard Biener  <rguenther@suse.de>
15005         PR middle-end/86159
15006         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15007         leave useless conversion stripping to force_gimple_operand_gsi.
15008         (gimplify_build2): Likewise.
15009         (gimplify_build1): Likewise.
15011 2018-06-15  Richard Biener  <rguenther@suse.de>
15013         PR middle-end/86076
15014         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15015         before adjusting their block.
15017 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15019         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15020         multilibs for *-*-rtems*.
15021         * config/riscv/t-rtems: New file.
15023 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15025         PR middle-end/86122
15026         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15027         unsigned_type_for returns NULL.
15029         PR target/85945
15030         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15031         subregs of multi-word pseudos unless the float mode has word size.
15033 2018-06-14  Richard Biener  <rguenther@suse.de>
15035         PR middle-end/86139
15036         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15037         duplicate and harmful type_hash_canon.
15038         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15040 2018-06-14  Richard Biener  <rguenther@suse.de>
15042         PR ipa/86124
15043         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15044         NULL cgraph_node.
15046 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15048         * config/rtems.h (STDINT_LONG32): Define.
15050 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15051             Prachi Godbole  <prachi.godbole@imgtec.com>
15053         * config/mips/mips-cpus.def: Define P6600.
15054         * config/mips/mips-tables.opt: Regenerate.
15055         * config/mips/mips.c (mips_ucbranch_type): New enum.
15056         (mips_rtx_cost_data): Add support for P6600.
15057         (mips_issue_rate): Likewise.
15058         (mips_multipass_dfa_lookahead): Likewise.
15059         (mips_avoid_hazard): Likewise.
15060         (mips_reorg_process_insns): Likewise.
15061         (mips_classify_branch_p6600): New function.
15062         * config/mips/mips.h (TUNE_P6600): New define.
15063         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15064         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15065         * config/mips/mips.md: Include p6600.md.
15066         (processor): Add p6600.
15067         * config/mips/p6600.md: New file.
15068         * doc/invoke.texi: Add p6600 to supported architectures.
15070 2018-06-13  Martin Sebor  <msebor@redhat.com>
15072         PR tree-optimization/86114
15073         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15074         of integer types.
15075         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15077 2018-06-13  Richard Biener  <rguenther@suse.de>
15079         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15080         Properly set vector type of the intermediate stmt.
15081         * tree-vect-stmts.c (vectorizable_operation): The destination
15082         var always has vectype_out type.
15084 2018-06-13  Jeff Law  <law@redhat.com>
15086         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15087         integer 0 for argument to print_rtl_with_bb.
15088         (rl78_reorg): Likewise.
15090 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15092         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15093         from rtx to rtx_insn *.
15094         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15095         "label".
15096         (add_sched_insns_for_speculation): Likewise for local "target",
15097         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15098         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15099         from rtx_insn ** to rtx_code_label **.
15100         (reorg_emit_nops): Likewise.
15101         (c6x_reorg): Likewise for local "call_labels".
15102         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15103         rtx to rtx_insn *.
15104         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15105         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15106         the loops over LABEL_REFS.
15107         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15108         braf_label.
15109         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15110         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15111         (split_branches): Strengthen local "olabel" from rtx to
15112         rtx_insn *, adding a safe_as_a cast.
15113         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15114         to "rtx_insn *".
15115         (add_insn_after): Likewise for first two params.
15116         (add_insn_before): Likewise.
15117         (remove_insn): Likewise for param.
15118         (emit_pattern_before_noloc): Likewise for second and third params.
15119         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15120         (emit_call_insn_before_noloc): Likewise.
15121         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15122         to "rtx_insn *".
15123         (emit_barrier_before): Likewise.
15124         (emit_label_before): Strengthen "label" param from "rtx" to
15125         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15126         "rtx_insn *".
15127         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15128         "rtx_insn *".
15129         (emit_pattern_after_noloc): Likewise.
15130         (emit_insn_after_noloc): Likewise.
15131         (emit_jump_insn_after_noloc): Likewise.
15132         (emit_call_insn_after_noloc): Likewise.
15133         (emit_debug_insn_after_noloc): Likewise.
15134         (emit_barrier_after): Likewise.
15135         (emit_label_after): Likewise for both params.
15136         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
15137         "loc" param from "int" to "location_t".
15138         (emit_insn_after_setloc): Likewise.
15139         (emit_jump_insn_after_setloc): Likewise.
15140         (emit_call_insn_after_setloc): Likewise.
15141         (emit_debug_insn_after_setloc): Likewise.
15142         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
15143         "loc" param from "int" to "location_t".
15144         (emit_pattern_before): Convert NULL_RTX to NULL.
15145         (emit_insn_before_setloc): Convert "loc" param from "int" to
15146         "location_t".
15147         (emit_jump_insn_before_setloc): Likewise.
15148         (emit_call_insn_before_setloc): Likewise.
15149         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15150         rtx_insn *.  Convert "loc" param from "int" to "location_t".
15151         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15152         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15153         Convert 3rd param from "int" to "location_t".
15154         (emit_barrier_before, emit_barrier_after, next_real_insn):
15155         Strengthen param from rtx to rtx_insn *.
15156         (emit_label_before): Strengthen 1st param from "rtx" to
15157         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
15158         "rtx_insn *".
15159         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15160         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15161         Strengthen 2nd param from "rtx" to "rtx_insn *".
15162         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15163         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15164         Likewise. Convert 3rd param from "int" to "location_t".
15165         (emit_label_after): Strengthen 1st param from "rtx" to
15166         "rtx_code_label *".
15167         (next_real_insn, remove_insn): Strengthen param from "rtx" to
15168         "rtx_insn *".
15169         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15170         from "rtx" to "rtx_insn *".
15172 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
15174         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15175         bodies streamed in with -Q.
15176         * dumpfile.c (dump_files): Add lto-stream-out dump file.
15177         * dumpfile.h (tree_dump_index): Add lto_stream_out.
15178         * gimple-streamer-out.c: Include gimple-pretty-print.h
15179         (output_bb): Dump stmts streamed.
15180         * lto-section-out.c: Include print-tree.h
15181         (lto_begin_section): Dump sections created.
15182         (lto_output_decl_index): Dump decl encoded.
15183         * lto-streamer-out.c: Include print-tree.h
15184         (create_output_block): Dump output block created.
15185         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15186         (output_function): Dump function output.
15187         (output_constructor): Dump constructor streamed.
15188         (write_global_stream): Output indexes encoded.
15189         (produce_asm_for_decls): Dump streams encoded.
15190         * lto-streamer.c (streamer_dump_file): New global var.
15191         * lto-streamer.h (streamer_dump_file): Declare.
15192         * passes.c (ipa_write_summaries): Initialize streamer dump.
15193         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15194         in.
15196 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
15198         PR target/86048
15199         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15200         offsets for register save directives.  Emit a second batch of save
15201         directives, if need be, when the function accesses prior frames.
15203 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15205         * config/arc/fpu.md (fmasf4): Force operand to register.
15206         (fnmasf4): Likewise.
15208 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15210         * config/arc/arc-protos.h (arc_pad_return): Remove.
15211         * config/arc/arc.c (machine_function): Remove force_short_suffix
15212         and size_reason.
15213         (arc_print_operand): Adjust printing of '&'.
15214         (arc_verify_short): Remove conditional printing of short suffix.
15215         (arc_final_prescan_insn): Remove reference to size_reason.
15216         (pad_return): New function.
15217         (arc_reorg): Call pad_return.
15218         (arc_pad_return): Remove.
15219         (arc_init_machine_status): Remove reference to force_short_suffix.
15220         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15221         (attr length): When attribute iscompact is true force to 2
15222         regardless; in the case of maybe check if we want to force the
15223         instruction to have 4 bytes length.
15224         (nopv): Change it to generate 4 byte long nop as well.
15225         (blockage): New pattern.
15226         (simple_return): Remove call to arc_pad_return.
15227         (p_return_i): Likewise.
15229 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15231         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15233 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15235         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15236         ARC cores.
15238 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15240         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15241         for ARC700 and ARCv2.
15243 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15245         PR target/86076
15246         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15247         operands[2] instead of operands[1].
15250 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15252         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15253         case, check whether the outer register overlaps an unallocatable
15254         register, not just whether it fits the required class.
15256 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15258         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15259         are poly_ints.
15260         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15261         (memrefs_conflict_p): Likewise.
15262         (init_alias_analysis): Likewise.
15263         * cfgexpand.c (expand_debug_expr): Likewise.
15264         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15265         * cse.c (fold_rtx): Likewise.
15266         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15267         * expr.c (emit_block_move_hints): Likewise.
15268         (clear_storage_hints, push_block, emit_push_insn): Likewise.
15269         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15270         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15271         (emit_group_store): Likewise.
15272         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
15273         to read the PRE/POST_MODIFY increment.
15274         * calls.c (store_one_arg): Use strip_offset.
15275         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15276         poly_int_rtx_p.
15277         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15278         by a VEC_SELECT.
15279         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15280         (simplify_binary_operation_1): Extend CONST_INT handling to
15281         poly_int_rtx_p.
15282         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15283         than a HOST_WIDE_INT.
15284         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15285         poly_int64.
15286         (adjust_mems, add_stores): Update accodingly.
15287         (vt_canonicalize_addr): Track polynomial offsets.
15288         (emit_note_insn_var_location): Likewise.
15289         (vt_add_function_parameter): Likewise.
15290         (vt_initialize): Likewise.
15292 2018-06-12  Jeff Law  <law@redhat.com>
15294         * config.gcc (alpha*-*-freebsd*): Remove.
15295         * config/alpha/freebsd.h: Remove.
15297 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
15299         PR other/69968
15300         * spellcheck-tree.c (levenshtein_distance): Rename to...
15301         (get_edit_distance): ...this, and update for underlying renaming.
15302         * spellcheck-tree.h (levenshtein_distance): Rename to...
15303         (get_edit_distance): ...this.
15304         * spellcheck.c (levenshtein_distance): Rename to...
15305         (get_edit_distance): ...this.  Convert from Levenshtein distance
15306         to Damerau-Levenshtein distance by supporting transpositions of
15307         adjacent characters.  Rename "v1" to "v_next" and "v0" to
15308         "v_one_ago".
15309         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15310         (selftest::test_edit_distance_unit_test_oneway): ...this, and
15311         update for underlying renaming.
15312         (selftest::levenshtein_distance_unit_test): Rename to...
15313         (selftest::test_get_edit_distance_unit): ...this, and update for
15314         underlying renaming.
15315         (selftest::test_find_closest_string): Add example from PR 69968
15316         where transposition helps
15317         (selftest::test_metric_conditions): Update for renaming.
15318         (selftest::test_metric_conditions): Likewise.
15319         (selftest::spellcheck_c_tests): Likewise.
15320         * spellcheck.h (levenshtein_distance): Rename both overloads to...
15321         (get_edit_distance): ...this.
15322         (best_match::consider): Update for renaming.
15324 2018-06-12  Martin Sebor  <msebor@redhat.com>
15326         PR tree-optimization/85259
15327         * builtins.c (compute_objsize): Handle constant offsets.
15328         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15329         true iff a warning has been issued.
15330         * gimple.h (gimple_nonartificial_location): New function.
15331         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15332         gimple_nonartificial_location and handle -Wno-system-headers.
15333         (handle_builtin_stxncpy): Same.
15335 2018-06-12  Martin Sebor  <msebor@redhat.com>
15337         PR c/85931
15338         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15340 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
15342         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15343         BUILTIN_VEC_XST entries for pointer to double and long long.
15345 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
15347         PR target/85990
15348         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15349         Update comments.
15350         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15351         Likewise.
15353 2018-06-12  Martin Liska  <mliska@suse.cz>
15355         * doc/options.texi: Document IntegerRange.
15357 2018-06-12  Martin Liska  <mliska@suse.cz>
15359         * config/i386/i386.opt: Make MPX-related options as Deprecated.
15360         * opt-functions.awk: Handle Deprecated flag.
15361         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15362         and report error.
15363         (read_cmdline_option): Report warning for a deprecated option.
15364         * opts.h (struct cl_option): Add new field cl_deprecated.
15365         (CL_ERR_DEPRECATED): New.
15367 2018-06-12  Martin Liska  <mliska@suse.cz>
15369         * doc/options.texi: Document Deprecated option flag.
15371 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15373         * config/arc/arc-arch.h (arc_extras): New enum.
15374         (arc_cpu_t):Add field extra.
15375         (arc_cpu_types): Consider the extras.
15376         * config/arc/arc-cpus.def: Add extras info.
15377         * config/arc/arc-opts.h (processor_type): Consider extra field.
15378         * config/arc/arc.c (arc_override_options): Handle extra field.
15380 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15382         * config/arc/arc-arch.h: Update ARC_OPTX macro.
15383         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15384         field.
15385         * config/arc/arc.c (arc_init): Update pic warning.
15386         (irq_range): Update irq range parsing warnings.
15387         (arc_override_options): Update various warning messages.
15388         (arc_handle_aux_attribute): Likewise.
15390 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
15392         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15394 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15396         * doc/sourcebuild.texi: Document usage of line number 0 in verify
15397         compiler messages directives.
15399 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
15401         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15402         * config/mips/mips-tables.opt: Regenerate.
15403         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15404         mips64r6.
15405         * doc/invoke.texi: Document -march=i6500.
15407 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
15409         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15410         (i6400_gpmul): Add cpu_unit.
15411         (i6400_gpdiv): Likewise.
15412         (i6400_msa_add_d): Update reservations.
15413         (i6400_msa_int_add) Likewise.
15414         (i6400_msa_short_logic3) Likewise.
15415         (i6400_msa_short_logic2) Likewise.
15416         (i6400_msa_short_logic) Likewise.
15417         (i6400_msa_move) Likewise.
15418         (i6400_msa_cmp) Likewise.
15419         (i6400_msa_short_float2) Likewise.
15420         (i6400_msa_div_d) Likewise.
15421         (i6400_msa_long_logic1) Likewise.
15422         (i6400_msa_long_logic2) Likewise.
15423         (i6400_msa_mult) Likewise.
15424         (i6400_msa_long_float2) Likewise.
15425         (i6400_msa_long_float4) Likewise.
15426         (i6400_msa_long_float5) Likewise.
15427         (i6400_msa_long_float8) Likewise.
15428         (i6400_fpu_fadd): Include frint type.
15429         (i6400_fpu_store): New define_insn_reservation.
15430         (i6400_fpu_load): Likewise.
15431         (i6400_fpu_move): Likewise.
15432         (i6400_fpu_fcmp): Likewise.
15433         (i6400_fpu_fmadd): Likewise.
15434         (i6400_int_mult): Include imul3nc type and update reservation.
15435         (i6400_int_div): Include idiv3 type and update reservation.
15436         (i6400_int_load): Update to check type not move_type.
15437         (i6400_int_store): Likewise.
15438         (i6400_int_prefetch): Set zero latency.
15440 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
15442         * gcc.c: Document new %@{...} sequence.
15443         (LINK_COMMAND_SPEC): Use it for the -L switches.
15444         (cpp_unique_options): Use it for the -I switches.
15445         (at_file_argbuf): New global variable.
15446         (in_at_file): Likewise.
15447         (alloc_args): Create at_file_argbuf.
15448         (clear_args): Truncate at_file_argbuf.
15449         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15450         (open_at_file): New function.
15451         (close_at_file): Likewise.
15452         (create_at_file): Delete.
15453         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15454         <'o'>: Likewise.
15455         <'@'>: New case.
15456         (validate_switches_from_spec): Deal with %@{...} sequence.
15457         (validate_switches): Likewise.
15458         (driver::finalize): Call clear_args.
15460 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
15462         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15464 2018-06-11  Martin Sebor  <msebor@redhat.com>
15466         * doc/invoke.texi (-Wall): List -Wc++17-compat.
15467         (Wno-class-memaccess): Add @opindex.
15468         (Wno-templates, Wno-multiple-inheritance): Same.
15469         (Wno-virtual-inheritance, Wno-namespaces): Same.
15470         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15471         (Wno-format-overflow, Wno-format-truncation): Same.
15472         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15473         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15474         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15475         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15476         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15477         (Wno-misspelled-isr): Same.
15479 2018-06-11  Martin Sebor  <msebor@redhat.com>
15481         * PR tree-optimization/86083
15482         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15484 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
15486         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15488 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
15490         PR target/85755
15491         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15492         on the correct operand.
15493         (*movdi_internal64): Ditto.
15495 2018-06-11  Martin Liska  <mliska@suse.cz>
15497         PR tree-optimization/86089
15498         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15500 2018-06-11  Julia Koval  <julia.koval@intel.com>
15502         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15503         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15504         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15506 2018-06-11  Olivier Hainque  <hainque@adacore.com>
15508         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15509         for Ada with strict dwarf2.
15511 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
15513         PR target/85755
15514         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15515         addresses.
15517 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
15519         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15521 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
15523         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15524         TARGET_ELF.
15526 2018-06-08  Martin Liska  <mliska@suse.cz>
15528         * tree-cfg.h (debug_function): Fix argument type to match
15529         implementation.
15531 2018-06-08  Martin Liska  <mliska@suse.cz>
15533         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
15534         Remove usage of MPX-related (and removed) fields.
15535         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
15537 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
15539         * cfg.c (debug): Use TDF_NONE rather than 0.
15540         * cfghooks.c (debug): Likewise.
15541         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
15542         (struct dump_option_value_info): Convert to...
15543         (struct kv_pair): ...this template type.
15544         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
15545         rather than 0.
15546         (optinfo_verbosity_options): Likewise.
15547         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
15548         OPTGROUP_NONE.
15549         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
15550         than int for "optgroup_flags" param.
15551         (dump_generic_expr_loc): Use dump_flags_t rather than int for
15552         "dump_kind" param.
15553         (dump_dec): Likewise.
15554         (dump_finish): Use TDF_NONE rather than 0.
15555         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
15556         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
15557         than 0.  Update for change to option_ptr.
15558         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
15559         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
15560         0.  Update for changes to optinfo_verbosity_options and
15561         optgroup_options.
15562         (opt_info_switch_p): Convert optgroup_flags from int to
15563         optgroup_flags_t.
15564         (dump_basic_block): Use dump_flags_t rather than int
15565         for "dump_kind" param.
15566         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
15567         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
15568         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
15569         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
15570         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
15571         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
15572         TDF_NONE): Convert from macros to...
15573         (enum dump_flag): ...this new enum.
15574         (dump_flags_t): Update to use enum.
15575         (operator|, operator&, operator~, operator|=, operator&=):
15576         Implement for dump_flags_t.
15577         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
15578         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
15579         Convert from macros to...
15580         (enum optgroup_flag): ...this new enum.
15581         (optgroup_flags_t): New typedef.
15582         (operator|, operator|=): Implement for optgroup_flags_t.
15583         (struct dump_file_info): Convert field "alt_flags" to
15584         dump_flags_t.  Convert field "optgroup_flags" to
15585         optgroup_flags_t.
15586         (dump_basic_block): Use dump_flags_t rather than int for param.
15587         (dump_generic_expr_loc): Likewise.
15588         (dump_dec): Likewise.
15589         (dump_register): Convert param "optgroup_flags" to
15590         optgroup_flags_t.
15591         (opt_info_enable_passes): Likewise.
15592         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
15593         than 0.
15594         * gimple-pretty-print.c (debug): Likewise.
15595         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
15596         (merged_store_group::apply_stores): Likewise.
15597         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
15598         * gimple.c (verify_gimple_pp): Likewise.
15599         * graphite-poly.c (print_pbb_body): Likewise.
15600         * passes.c (pass_manager::register_one_dump_file): Convert
15601         local "optgroup_flags" to optgroup_flags_t.
15602         * print-tree.c (print_node): Use TDF_NONE rather than 0.
15603         (debug): Likewise.
15604         (debug_body): Likewise.
15605         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
15606         to optgroup_flags_t.
15607         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
15608         than 0.
15609         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
15610         (convert_mult_to_fma): Likewise.
15611         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
15612         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
15613         * tree-vect-data-refs.c (dump_lower_bound): Convert param
15614         "dump_kind" to dump_flags_t.
15616 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15618         * config/rs6000/rs6000.c (min, max): Delete.
15620 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15622         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
15623         -mabi=spe and -mabi=no-spe.
15625 2018-06-08  Martin Liska  <mliska@suse.cz>
15627         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
15628         where we expect an existing summary.
15630 2018-06-08  Martin Liska  <mliska@suse.cz>
15632         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
15633         * ipa-inline.h (estimate_edge_growth): Likewise.
15635 2018-06-08  Martin Liska  <mliska@suse.cz>
15637         * cgraph.c (function_version_hasher::hash): Use
15638         cgraph_node::get_uid ().
15639         (function_version_hasher::equal):
15640         * cgraph.h (cgraph_node::get_uid): New method.
15641         * ipa-inline.c (update_caller_keys): Use
15642         cgraph_node::get_uid ().
15643         (update_callee_keys): Likewise.
15644         * ipa-utils.c (searchc): Likewise.
15645         (ipa_reduced_postorder): Likewise.
15646         * lto-cgraph.c (input_node): Likewise.
15647         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
15648         * symbol-summary.h (symtab_insertion): Likewise.
15649         (symtab_removal): Likewise.
15650         (symtab_duplication): Likewise.
15651         * tree-pretty-print.c (dump_function_header): Likewise.
15652         * tree-sra.c (convert_callers_for_node): Likewise.
15654 2018-06-08  Martin Liska  <mliska@suse.cz>
15656         * cgraph.c (symbol_table::create_edge): Always assign a new
15657         unique number.
15658         (symbol_table::free_edge): Do not recycle numbers.
15659         * cgraph.h (cgraph_edge::get): New method.
15660         * symbol-summary.h (symtab_removal): Use it.
15661         (symtab_duplication): Likewise.
15662         (call_summary::hashable_uid): Remove.
15664 2018-06-08  Martin Liska  <mliska@suse.cz>
15666         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
15667         (initialize_growth_caches): Remove.
15668         (free_growth_caches): Likewise.
15669         (do_estimate_edge_time): Use edge_growth_cache.
15670         (do_estimate_edge_size): Likewise.
15671         (do_estimate_edge_hints): Likewise.
15672         * ipa-inline.c (reset_edge_caches): Likewise.
15673         (recursive_inlining): Likewise.
15674         (inline_small_functions): Likewise.
15675         * ipa-inline.h (initialize_growth_caches): Remove.
15676         (estimate_edge_size): Likewise.
15677         (estimate_edge_time): Likewise.
15678         (estimate_edge_hints): Likewise.
15679         (reset_edge_growth_cache): Likewise.
15680         * symbol-summary.h (call_summary::remove): New method.
15682 2018-06-08  Martin Liska  <mliska@suse.cz>
15684         * ipa-cp.c (class edge_clone_summary): New summary.
15685         (grow_edge_clone_vectors): Remove.
15686         (ipcp_edge_duplication_hook): Remove.
15687         (class edge_clone_summary_t): New call_summary class.
15688         (ipcp_edge_removal_hook): Remove.
15689         (edge_clone_summary_t::duplicate): New function.
15690         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
15691         (create_specialized_node): Likewise.
15692         (ipcp_driver): Initialize edge_clone_summaries and do not
15693         register hooks.
15695 2018-06-08  Martin Liska  <mliska@suse.cz>
15697         * symbol-summary.h (get): New function.
15698         (call_summary::m_initialize_when_cloning): New class member.
15700 2018-06-08  Martin Liska  <mliska@suse.cz>
15702         * cgraph.c (cgraph_node::remove): Do not recycle uid.
15703         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
15704         (symbol_table::allocate_cgraph_symbol): Do not set uid.
15705         * passes.c (uid_hash_t): Record removed_nodes by their uids.
15706         (remove_cgraph_node_from_order): Use the removed_nodes set.
15707         (do_per_function_toporder): Likwise.
15708         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
15709         instead of summary_uid.
15710         (symtab_removal): Likewise.
15711         (symtab_duplication): Likewise.
15713 2018-06-08  Martin Liska  <mliska@suse.cz>
15715         * ipa-cp.c (ipcp_store_bits_results): Use
15716         ipcp_transformation_sum.
15717         (ipcp_store_vr_results): Likewise.
15718         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
15719         to ...
15720         (ipcp_transformation_initialize): ... this.
15721         (ipa_set_node_agg_value_chain):
15722         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
15723         (write_ipcp_transformation_info): Likewise.
15724         (read_ipcp_transformation_info): Likewise.
15725         (ipcp_update_bits): Likewise.
15726         (ipcp_update_vr): Likewise.
15727         (ipcp_transform_function): Likewise.
15728         * ipa-prop.h: Rename ipcp_transformation_summary to
15729         ipcp_transformation.
15730         (class ipcp_transformation_t): New function summary.
15731         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
15732         (ipa_get_agg_replacements_for_node): Likewise.
15734 2018-06-08  Martin Liska  <mliska@suse.cz>
15736         * ipa-pure-const.c (struct funct_state_d): Do it class instead
15737         of struct.
15738         (class funct_state_summary_t): New function_summary class.
15739         (has_function_state): Remove.
15740         (get_function_state): Likewise.
15741         (set_function_state): Likewise.
15742         (add_new_function): Likewise.
15743         (funct_state_summary_t::insert): New function.
15744         (duplicate_node_data): Remove.
15745         (remove_node_data): Remove.
15746         (funct_state_summary_t::duplicate): New function.
15747         (register_hooks): Create new funct_state_summaries.
15748         (pure_const_generate_summary): Use it.
15749         (pure_const_write_summary): Likewise.
15750         (pure_const_read_summary): Likewise.
15751         (propagate_pure_const): Likewise.
15752         (propagate_nothrow): Likewise.
15753         (dump_malloc_lattice): Likewise.
15754         (propagate_malloc): Likewise.
15755         (execute): Do not register hooks, just remove summary
15756         instead.
15757         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
15758         constructor.
15760 2018-06-08  Martin Liska  <mliska@suse.cz>
15762         * ipa-reference.c (remove_node_data): Remove.
15763         (duplicate_node_data): Likewise.
15764         (class ipa_ref_var_info_summary_t): New class.
15765         (class ipa_ref_opt_summary_t): Likewise.
15766         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
15767         (get_reference_optimization_summary): Use
15768         ipa_ref_opt_sum_summaries.
15769         (set_reference_vars_info): Remove.
15770         (set_reference_optimization_summary): Likewise.
15771         (ipa_init): Create summaries.
15772         (init_function_info): Use function summary.
15773         (ipa_ref_opt_summary_t::duplicate): New function.
15774         (ipa_ref_opt_summary_t::remove): New function.
15775         (get_read_write_all_from_node): Fix GNU coding style.
15776         (propagate): Use function summary.
15777         (write_node_summary_p): Fix GNU coding style.
15778         (stream_out_bitmap): Likewise.
15779         (ipa_reference_read_optimization_summary): Use function summary.
15780         (ipa_reference_c_finalize): Do not release hooks.
15782 2018-06-08  Martin Liska  <mliska@suse.cz>
15784         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
15785         (analyze_function_body): Extract multiple calls of get_create.
15786         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15787         * ipa-inline.c (recursive_inlining): Use ::get method.
15788         * ipa-inline.h (estimate_edge_growth): Likewise.
15790 2018-06-08  Martin Liska  <mliska@suse.cz>
15792         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
15793         HSA_INVALID.
15794         (hsa_function_summary::hsa_function_summary): Use the new enum
15795         value.
15796         (hsa_gpu_implementation_p): Use hsa_summaries::get.
15797         * hsa-gen.c (hsa_get_host_function): Likewise.
15798         (get_brig_function_name): Likewise.
15799         * ipa-hsa.c (process_hsa_functions): Likewise.
15800         (ipa_hsa_write_summary): Likewise.
15801         * symbol-summary.h (symtab_duplication): Use ::get function/
15802         (get): New function.
15804 2018-06-08  Martin Liska  <mliska@suse.cz>
15806         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
15807         of get.
15808         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
15809         (hsa_register_kernel): Likewise.
15810         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
15811         * hsa-gen.c (hsa_get_host_function): Likewise.
15812         (get_brig_function_name): Likewise.
15813         (generate_hsa): Likewise.
15814         (pass_gen_hsail::execute): Likewise.
15815         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
15816         (devirtualization_time_bonus): Likewise.
15817         (ipcp_propagate_stage): Likewise.
15818         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15819         (edge_set_predicate): Likewise.
15820         (evaluate_conditions_for_known_args): Likewise.
15821         (evaluate_properties_for_edge): Likewise.
15822         (ipa_fn_summary::reset): Likewise.
15823         (ipa_fn_summary_t::duplicate): Likewise.
15824         (dump_ipa_call_summary): Likewise.
15825         (ipa_dump_fn_summary): Likewise.
15826         (analyze_function_body): Likewise.
15827         (compute_fn_summary): Likewise.
15828         (estimate_edge_devirt_benefit): Likewise.
15829         (estimate_edge_size_and_time): Likewise.
15830         (estimate_calls_size_and_time): Likewise.
15831         (estimate_node_size_and_time): Likewise.
15832         (inline_update_callee_summaries): Likewise.
15833         (remap_edge_change_prob): Likewise.
15834         (remap_edge_summaries): Likewise.
15835         (ipa_merge_fn_summary_after_inlining): Likewise.
15836         (ipa_update_overall_fn_summary): Likewise.
15837         (read_ipa_call_summary): Likewise.
15838         (inline_read_section): Likewise.
15839         (write_ipa_call_summary): Likewise.
15840         (ipa_fn_summary_write): Likewise.
15841         (ipa_free_fn_summary): Likewise.
15842         * ipa-hsa.c (process_hsa_functions): Likewise.
15843         (ipa_hsa_write_summary): Likewise.
15844         (ipa_hsa_read_section): Likewise.
15845         * ipa-icf.c (sem_function::merge): Likewise.
15846         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15847         (do_estimate_edge_time): Likewise.
15848         (estimate_size_after_inlining): Likewise.
15849         (estimate_growth): Likewise.
15850         (growth_likely_positive): Likewise.
15851         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15852         (inline_call): Likewise.
15853         * ipa-inline.c (caller_growth_limits): Likewise.
15854         (can_inline_edge_p): Likewise.
15855         (can_inline_edge_by_limits_p): Likewise.
15856         (compute_uninlined_call_time): Likewise.
15857         (compute_inlined_call_time): Likewise.
15858         (want_inline_small_function_p): Likewise.
15859         (edge_badness): Likewise.
15860         (update_caller_keys): Likewise.
15861         (update_callee_keys): Likewise.
15862         (recursive_inlining): Likewise.
15863         (inline_small_functions): Likewise.
15864         (inline_to_all_callers_1): Likewise.
15865         (dump_overall_stats): Likewise.
15866         (early_inline_small_functions): Likewise.
15867         (early_inliner): Likewise.
15868         * ipa-inline.h (estimate_edge_growth): Likewise.
15869         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15870         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15871         * ipa-prop.h (IPA_NODE_REF): Likewise.
15872         (IPA_EDGE_REF): Likewise.
15873         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15874         (propagate_malloc): Likewise.
15875         * ipa-split.c (execute_split_functions): Likewise.
15876         * symbol-summary.h: Rename get to get_create.
15877         (get): Likewise.
15878         (get_create): Likewise.
15879         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15881 2018-06-08  Martin Liska  <mliska@suse.cz>
15883         * symbol-summary.h (release): Move definition out of class
15884         declaration.
15885         (symtab_removal): Likewise.
15886         (symtab_duplication): Likewise.
15888 2018-06-08  Martin Liska  <mliska@suse.cz>
15890         * symbol-summary.h (function_summary): Move constructor
15891         implementation out of class declaration.
15892         (release): Likewise.
15893         (symtab_insertion): Likewise.
15894         (symtab_removal): Likewise.
15895         (symtab_duplication): Likewise.
15896         (get): Likewise.
15898 2018-06-08  Martin Liska  <mliska@suse.cz>
15900         * Makefile.in: Remove support for MPX (macros, related functions,
15901         fields in cgraph_node, ...).
15902         * builtin-types.def (BT_BND): Likewise.
15903         (BT_FN_BND_CONST_PTR): Likewise.
15904         (BT_FN_CONST_PTR_BND): Likewise.
15905         (BT_FN_VOID_PTR_BND): Likewise.
15906         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
15907         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
15908         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
15909         (expand_builtin_mempcpy_with_bounds): Likewise.
15910         (expand_builtin_memset_with_bounds): Likewise.
15911         (expand_builtin_memset_args): Likewise.
15912         (std_expand_builtin_va_start): Likewise.
15913         (expand_builtin): Likewise.
15914         (expand_builtin_with_bounds): Likewise.
15915         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
15916         (DEF_LIB_BUILTIN_CHKP): Likewise.
15917         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
15918         (DEF_CHKP_BUILTIN): Likewise.
15919         (BUILT_IN_MEMCPY): Likewise.
15920         (BUILT_IN_MEMMOVE): Likewise.
15921         (BUILT_IN_MEMPCPY): Likewise.
15922         (BUILT_IN_MEMSET): Likewise.
15923         (BUILT_IN_STPCPY): Likewise.
15924         (BUILT_IN_STRCAT): Likewise.
15925         (BUILT_IN_STRCHR): Likewise.
15926         (BUILT_IN_STRCPY): Likewise.
15927         (BUILT_IN_STRLEN): Likewise.
15928         (BUILT_IN_MEMCPY_CHK): Likewise.
15929         (BUILT_IN_MEMMOVE_CHK): Likewise.
15930         (BUILT_IN_MEMPCPY_CHK): Likewise.
15931         (BUILT_IN_MEMSET_CHK): Likewise.
15932         (BUILT_IN_STPCPY_CHK): Likewise.
15933         (BUILT_IN_STRCAT_CHK): Likewise.
15934         (BUILT_IN_STRCPY_CHK): Likewise.
15935         * calls.c (store_bounds): Likewise.
15936         (emit_call_1): Likewise.
15937         (special_function_p): Likewise.
15938         (maybe_warn_nonstring_arg): Likewise.
15939         (initialize_argument_information): Likewise.
15940         (finalize_must_preallocate): Likewise.
15941         (compute_argument_addresses): Likewise.
15942         (expand_call): Likewise.
15943         * cfgexpand.c (expand_call_stmt): Likewise.
15944         (expand_return): Likewise.
15945         (expand_gimple_stmt_1): Likewise.
15946         (pass_expand::execute): Likewise.
15947         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
15948         (cgraph_node::remove): Likewise.
15949         (cgraph_node::dump): Likewise.
15950         (cgraph_node::verify_node): Likewise.
15951         * cgraph.h (chkp_function_instrumented_p): Likewise.
15952         (symtab_node::get_alias_target): Likewise.
15953         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
15954         (cgraph_local_p): Likewise.
15955         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
15956         (cgraph_edge::rebuild_references): Likewise.
15957         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
15958         (walk_polymorphic_call_targets): Likewise.
15959         (cgraph_node::expand_thunk): Likewise.
15960         (symbol_table::output_weakrefs): Likewise.
15961         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
15962         (ix86_handle_option): Likewise.
15963         * config/i386/constraints.md: Likewise.
15964         * config/i386/i386-builtin-types.def (BND): Likewise.
15965         (VOID): Likewise.
15966         (PVOID): Likewise.
15967         (ULONG): Likewise.
15968         * config/i386/i386-builtin.def (BDESC_END): Likewise.
15969         (BDESC_FIRST): Likewise.
15970         (BDESC): Likewise.
15971         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
15972         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
15973         * config/i386/i386.c (enum reg_class): Likewise.
15974         (ix86_target_string): Likewise.
15975         (ix86_option_override_internal): Likewise.
15976         (ix86_conditional_register_usage): Likewise.
15977         (ix86_valid_target_attribute_inner_p): Likewise.
15978         (ix86_set_indirect_branch_type): Likewise.
15979         (ix86_set_current_function): Likewise.
15980         (ix86_function_arg_regno_p): Likewise.
15981         (init_cumulative_args): Likewise.
15982         (ix86_function_arg_advance): Likewise.
15983         (ix86_function_arg): Likewise.
15984         (ix86_pass_by_reference): Likewise.
15985         (ix86_function_value_regno_p): Likewise.
15986         (ix86_function_value_1): Likewise.
15987         (ix86_function_value_bounds): Likewise.
15988         (ix86_return_in_memory): Likewise.
15989         (ix86_setup_incoming_vararg_bounds): Likewise.
15990         (ix86_va_start): Likewise.
15991         (indirect_thunk_need_prefix): Likewise.
15992         (print_reg): Likewise.
15993         (ix86_print_operand): Likewise.
15994         (ix86_expand_call): Likewise.
15995         (ix86_output_function_return): Likewise.
15996         (reg_encoded_number): Likewise.
15997         (BDESC_VERIFYS): Likewise.
15998         (ix86_init_mpx_builtins): Likewise.
15999         (ix86_init_builtins): Likewise.
16000         (ix86_emit_cmove): Likewise.
16001         (ix86_emit_move_max): Likewise.
16002         (ix86_expand_builtin): Likewise.
16003         (ix86_builtin_mpx_function): Likewise.
16004         (ix86_get_arg_address_for_bt): Likewise.
16005         (ix86_load_bounds): Likewise.
16006         (ix86_store_bounds): Likewise.
16007         (ix86_load_returned_bounds): Likewise.
16008         (ix86_store_returned_bounds): Likewise.
16009         (ix86_class_likely_spilled_p): Likewise.
16010         (ix86_hard_regno_mode_ok): Likewise.
16011         (x86_order_regs_for_local_alloc): Likewise.
16012         (ix86_mitigate_rop): Likewise.
16013         (ix86_bnd_prefixed_insn_p): Likewise.
16014         (ix86_mpx_bound_mode): Likewise.
16015         (ix86_make_bounds_constant): Likewise.
16016         (ix86_initialize_bounds): Likewise.
16017         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16018         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16019         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16020         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16021         (TARGET_CHKP_BOUND_MODE): Likewise.
16022         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16023         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16024         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16025         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16026         * config/i386/i386.h (TARGET_MPX): Likewise.
16027         (TARGET_MPX_P): Likewise.
16028         (VALID_BND_REG_MODE): Likewise.
16029         (FIRST_BND_REG): Likewise.
16030         (LAST_BND_REG): Likewise.
16031         (enum reg_class): Likewise.
16032         (BND_REG_P): Likewise.
16033         (BND_REGNO_P): Likewise.
16034         (BNDmode): Likewise.
16035         (ADJUST_INSN_LENGTH): Likewise.
16036         * config/i386/i386.md: Likewise.
16037         * config/i386/i386.opt: Likewise.
16038         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16039         (defined): Likewise.
16040         (LINK_MPX): Likewise.
16041         (MPX_SPEC): Likewise.
16042         (LIBMPX_SPEC): Likewise.
16043         (LIBMPXWRAPPERS_SPEC): Likewise.
16044         (CHKP_SPEC): Likewise.
16045         * config/i386/predicates.md: Likewise.
16046         * dbxout.c (dbxout_type): Likewise.
16047         * doc/extend.texi: Likewise.
16048         * doc/invoke.texi: Likewise.
16049         * doc/md.texi: Likewise.
16050         * doc/tm.texi: Likewise.
16051         * doc/tm.texi.in: Likewise.
16052         * dwarf2out.c (is_base_type): Likewise.
16053         (gen_formal_types_die): Likewise.
16054         (gen_subprogram_die): Likewise.
16055         (gen_type_die_with_usage): Likewise.
16056         (gen_decl_die): Likewise.
16057         (dwarf2out_late_global_decl): Likewise.
16058         * expr.c (expand_assignment): Likewise.
16059         (emit_storent_insn): Likewise.
16060         (store_expr_with_bounds): Likewise.
16061         (store_expr): Likewise.
16062         (expand_expr_real_1): Likewise.
16063         * expr.h (store_expr_with_bounds): Likewise.
16064         * function.c (use_register_for_decl): Likewise.
16065         (struct bounds_parm_data): Likewise.
16066         (assign_parms_augmented_arg_list): Likewise.
16067         (assign_parm_find_entry_rtl): Likewise.
16068         (assign_parm_is_stack_parm): Likewise.
16069         (assign_parm_load_bounds): Likewise.
16070         (assign_bounds): Likewise.
16071         (assign_parms): Likewise.
16072         (expand_function_start): Likewise.
16073         * gcc.c (CHKP_SPEC): Likewise.
16074         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16075         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16076         (wrestrict_dom_walker::check_call): Likewise.
16077         * gimple.c (gimple_build_call_from_tree): Likewise.
16078         * gimple.h (enum gf_mask): Likewise.
16079         (gimple_call_with_bounds_p): Likewise.
16080         (gimple_call_set_with_bounds): Likewise.
16081         * gimplify.c (gimplify_init_constructor): Likewise.
16082         * ipa-cp.c (initialize_node_lattices): Likewise.
16083         (propagate_constants_across_call): Likewise.
16084         (find_more_scalar_values_for_callers_subset): Likewise.
16085         * ipa-hsa.c (process_hsa_functions): Likewise.
16086         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16087         * ipa-icf.c (sem_function::merge): Likewise.
16088         * ipa-inline.c (early_inliner): Likewise.
16089         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16090         (warn_function_cold): Likewise.
16091         (propagate_pure_const): Likewise.
16092         * ipa-ref.h (enum GTY): Likewise.
16093         * ipa-split.c (find_retbnd): Likewise.
16094         (consider_split): Likewise.
16095         (split_function): Likewise.
16096         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16097         * ipa.c (walk_polymorphic_call_targets): Likewise.
16098         (symbol_table::remove_unreachable_nodes): Likewise.
16099         (process_references): Likewise.
16100         (cgraph_build_static_cdtor_1): Likewise.
16101         * lto-cgraph.c (lto_output_node): Likewise.
16102         (output_refs): Likewise.
16103         (compute_ltrans_boundary): Likewise.
16104         (input_overwrite_node): Likewise.
16105         (input_node): Likewise.
16106         (input_cgraph_1): Likewise.
16107         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16108         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16109         (class pass_chkp_instrumentation_passes): Likewise.
16110         (make_pass_chkp_instrumentation_passes): Likewise.
16111         * passes.def: Likewise.
16112         * rtl.h (struct GTY): Likewise.
16113         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16114         * stor-layout.c (layout_type): Likewise.
16115         * symtab.c: Likewise.
16116         * target.def: Likewise.
16117         * targhooks.c (default_chkp_bound_type): Likewise.
16118         (default_chkp_bound_mode): Likewise.
16119         (default_builtin_chkp_function): Likewise.
16120         (default_chkp_function_value_bounds): Likewise.
16121         (default_chkp_make_bounds_constant): Likewise.
16122         (default_chkp_initialize_bounds): Likewise.
16123         * targhooks.h (default_chkp_bound_type): Likewise.
16124         (default_chkp_bound_mode): Likewise.
16125         (default_builtin_chkp_function): Likewise.
16126         (default_chkp_function_value_bounds): Likewise.
16127         (default_chkp_make_bounds_constant): Likewise.
16128         (default_chkp_initialize_bounds): Likewise.
16129         * toplev.c (compile_file): Likewise.
16130         (process_options): Likewise.
16131         * tree-core.h (DEF_BUILTIN): Likewise.
16132         (DEF_BUILTIN_CHKP): Likewise.
16133         * tree-inline.c (declare_return_variable): Likewise.
16134         (remap_gimple_stmt): Likewise.
16135         (copy_bb): Likewise.
16136         (initialize_inlined_parameters): Likewise.
16137         (expand_call_inline): Likewise.
16138         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16139         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16140         (make_pass_ipa_chkp_produce_thunks): Likewise.
16141         (make_pass_chkp): Likewise.
16142         (make_pass_chkp_opt): Likewise.
16143         (make_pass_chkp_instrumentation_passes): Likewise.
16144         * tree-pretty-print.c (dump_generic_node): Likewise.
16145         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16146         * tree-ssa-dce.c (propagate_necessity): Likewise.
16147         (eliminate_unnecessary_stmts): Likewise.
16148         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16149         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16150         * tree-ssa-sccvn.h: Likewise.
16151         * tree-ssa-strlen.c (get_string_length): Likewise.
16152         (valid_builtin_call): Likewise.
16153         (adjust_last_stmt): Likewise.
16154         (handle_builtin_strchr): Likewise.
16155         (handle_builtin_strcpy): Likewise.
16156         (handle_builtin_stxncpy): Likewise.
16157         (handle_builtin_memcpy): Likewise.
16158         (handle_builtin_strcat): Likewise.
16159         (strlen_check_and_optimize_stmt): Likewise.
16160         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16161         * tree-streamer-in.c: Likewise.
16162         * tree-streamer.c (record_common_node): Likewise.
16163         * tree.c (tree_code_size): Likewise.
16164         (wide_int_to_tree_1): Likewise.
16165         (type_contains_placeholder_1): Likewise.
16166         (build_common_tree_nodes): Likewise.
16167         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16168         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16169         (POINTER_BOUNDS_P): Likewise.
16170         (BOUNDED_TYPE_P): Likewise.
16171         (BOUNDED_P): Likewise.
16172         (CALL_WITH_BOUNDS_P): Likewise.
16173         (pointer_bounds_type_node): Likewise.
16174         * value-prof.c (gimple_ic): Likewise.
16175         * var-tracking.c (vt_add_function_parameters): Likewise.
16176         * varasm.c (make_decl_rtl): Likewise.
16177         (assemble_start_function): Likewise.
16178         (output_constant): Likewise.
16179         (maybe_assemble_visibility): Likewise.
16180         * varpool.c (ctor_for_folding): Likewise.
16181         * chkp-builtins.def: Remove.
16182         * ipa-chkp.c: Remove.
16183         * ipa-chkp.h: Remove.
16184         * rtl-chkp.c: Remove.
16185         * rtl-chkp.h: Remove.
16186         * tree-chkp-opt.c: Remove.
16187         * tree-chkp.c: Remove.
16188         * tree-chkp.h: Remove.
16190 2018-06-07  Carl Love  <cel@us.ibm.com>
16192         * config/rs6000/vsx.md (vextract_fp_from_shorth,
16193         vextract_fp_from_shortl): Add BE support.
16195 2018-06-07  Paul Koning  <ni1d@arrl.net>
16197         * compare-elim.c (try_merge_compare): Don't merge compare if
16198         address contains a side effect.
16199         (try_eliminate_compare): Likewise.
16201 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
16203         * config.gcc: Support "tremont".
16204         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16205         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16206         PROCESSOR_TREMONT.
16207         * config/i386/i386.c (m_TREMONT): Define.
16208         (processor_target_table): Add "tremont".
16209         (PTA_TREMONT): Define.
16210         (ix86_lea_outperforms): Add TARGET_TREMONT.
16211         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16212         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16213         and M_INTEL_GOLDMONT_PLUS.
16214         (fold_builtin_cpu): Add "tremont".
16215         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16216         (ix86_option_override_internal): Add "tremont".
16217         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16218         (processor_type): Add PROCESSOR_TREMONT.
16219         * config/i386/x86-tune.def: Add m_TREMONT.
16220         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16222 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16224         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16225         symbol defined for msp430i* devices to be lower case.
16227 2018-06-07  Richard Biener  <rguenther@suse.de>
16229         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16230         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16231         Properly wrap signed arithmetic if overflow wraps.
16233 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16235         PR tree-optimization/69615
16236         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16237         of a cast from a same precision integral SSA_NAME in a bb dominated
16238         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16239         cast to utype if rhs2 has already a compatible type.
16241 2018-06-07  Richard Biener  <rguenther@suse.de>
16243         PR tree-optimization/85935
16244         * graphite-scop-detection.c (find_params_in_bb): Analyze
16245         condition operands with respect to the correct loop.  Assert
16246         the analysis doesn't fail.
16248 2018-06-04  Carl Love  <cel@us.ibm.com>
16250         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16251         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16252         as it is slightly cheaper.
16253         (first_match_or_eos_index_<mode>):
16254         Calculate index using natural element order.
16255         (first_match_index_<mode>):
16256         Calculate index using natural element order.
16257         (first_match_or_eos_index_<mode>):
16258         Calculate index using natural order.
16259         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16260         for BE and LE modes.
16261         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16262         P9V_BUILTIN_VCLZLSBB_V16QI.
16263         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16264         specific.
16266 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16268         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16269         indentation and line wrap for many prototypes.  Add missing
16270         @smallexample directives around block of prototypes for vec_xl and
16271         vec_xst.
16273 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
16275         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16276         track if we pass or return IEEE 128-bit floating point.
16277         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16278         C++ mangling that is compatible with GCC 8.1.
16279         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16280         (init_cumulative_args): Note if we pass or return IEEE 128-bit
16281         floating point types.
16282         (rs6000_function_arg_advance_1): Likewise.
16283         (rs6000_mangle_type): Optionally generate mangled names that match
16284         what GCC 8.1 generated for IEEE 128-bit floating point types.
16285         (rs6000_globalize_decl_name): If we have an external function that
16286         passes or returns IEEE 128-bit types, generate a weak reference
16287         from the mangled name used in GCC 8.1 to the current mangled
16288         name.
16289         (rs6000_init_builtins): Make __ibm128 use the long double type if
16290         long double is IBM extended double.  Make __float128 use the long
16291         double type if long double is IEEE 128-bit.
16293         PR target/85657
16294         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16295         macro for __ibm128 built-in functions.
16296         (PACK_IF): Add __ibm128 pack/unpack functions.
16297         (UNPACK_IF): Likewise.
16298         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16299         enable long double built-in functions if long double is IEEE
16300         128-bit floating point.
16301         (rs6000_invalid_builtin): Update long double built-in function
16302         error message.
16303         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16304         functions, adjust the built-in function to use the long double
16305         built-in function if __ibm128 and long double are the same type.
16306         * doc/extend.texi (PowerPC builtins): Update documention for
16307         __builtin_{,un}pack_longdouble.  Add documentation for
16308         __builtin_{,un}pack_ibm128.
16310 2018-06-06  Jim Wilson  <jimw@sifive.com>
16312         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16313         (struct machine_function): New field interrupt_mode.
16314         (riscv_handle_type_attribute): New function.  Add forward declaration.
16315         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16316         (riscv_expand_epilogue): Check interrupt_mode field.
16317         (riscv_set_current_function): Check interrupt attribute args and
16318         set interrupt_mode field.
16319         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16320         (riscv_sret, riscv_uret): New.
16321         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16322         new arguments to interrupt attribute.
16324 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
16326         PR target/63177
16327         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16328         Don't handle -mcpu=power8 if -mpower9-vector is also used.
16330 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16332         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16333         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16334         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16335         several redundant entries.
16337 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
16339         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16340         type from "rtx" to "rtx_insn *".
16341         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16342         for local "call_insn", removing cast.
16343         (ix86_expand_call): Likewise, introducing a "call_insn" local.
16345 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
16347         PR tree-optimization/86066
16348         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16349         for BIT_INSERT_EXPR stores.
16351 2018-06-06  Richard Biener  <rguenther@suse.de>
16353         PR tree-optimization/86062
16354         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16355         component refs ontop
16356         of to be offsetted base.
16358 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16360         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16361         to be static and remove check on interrupt attribute name.
16363 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16365         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16366         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16368 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16370         PR target/79924
16371         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16372         second argument.
16373         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16374         Remove second argument, change how error is called.
16375         (aarch64_layout_arg): Remove second argument from
16376         aarch64_err_no_fpadvsimd call.
16377         (aarch64_init_cumulative_args): Ditto.
16378         (aarch64_gimplify_va_arg_expr): Ditto.
16379         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16381 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
16383         * config/i386/i386.md (simple_return_indirect_internal): New expander.
16384         (*simple_return_indirect_internal<mode>): Rename from
16385         simple_return_indirect_internal.  Use W mode iterator.
16386         (rstorssp): New expander.
16387         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
16388         (clrssbsy): New expander.
16389         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
16391 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16393         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16394         __typeof__.
16395         (cmse_check_pointed_object): Likewise.
16397 2018-06-05  Martin Liska  <mliska@suse.cz>
16399         PR gcov-profile/47618
16400         * doc/invoke.texi: Document how -fprofile-dir format
16401         is extended.
16403 2018-06-05  Richard Biener  <rguenther@suse.de>
16405         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16406         removal pretend DOM info isn't available so we do not update
16407         it and only remove edges, not dominated blocks.  Actually free
16408         DOM info in case we removed something.  Remove unreachable blocks.
16409         (mfb_keep_latches): Work with either DOM info or marked backedges.
16410         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16411         first.  Mark backedges if DOM info isn't available.
16412         (Re-)compute DOM info after cleanup_control_flow_pre.
16414 2018-06-05  Richard Biener  <rguenther@suse.de>
16416         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16417         (locus_discrim_hasher::hash): Adjust.
16418         (locus_discrim_hasher::equal): Likewise.
16419         (next_discriminator_for_locus): Work on line directly.
16420         (same_line_p): Pass in expanded locus1 as well.
16421         (assign_discriminators): Avoid redundant location expansions.
16423 2018-06-05  Richard Biener  <rguenther@suse.de>
16425         PR tree-optimization/86046
16426         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16427         if required after clearing TREE_ADDRESSABLE.
16429 2018-06-05  Richard Biener  <rguenther@suse.de>
16431         PR tree-optimization/86047
16432         * tree-ssa-loop.c (for_each_index): Glob handling of all
16433         decls and constants and really handle all of them.
16435 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16437         PR target/81497
16438         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16439         qualifier_void_pointer and qualifier_const_void_pointer.
16440         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16441         (arm_init_builtins): Handle the above.
16442         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16443         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16444         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16445         void intrinsics.
16447 2018-06-05  Martin Liska  <mliska@suse.cz>
16449         * auto-profile.c (read_autofdo_file): Do not use
16450         gcov_ctr_summary struct.
16451         (afdo_callsite_hot_enough_for_early_inline): Likewise.
16452         * coverage.c (struct counts_entry): Likewise.
16453         (read_counts_file): Read just single summary entry.
16454         (get_coverage_counts): Use gcov_summary struct.
16455         * coverage.h (get_coverage_counts): Likewise.
16456         * gcov-dump.c (dump_working_sets): Likewise.
16457         (tag_summary): Dump just single summary.
16458         * gcov-io.c (gcov_write_summary): Write just histogram
16459         summary.
16460         (gcov_read_summary): Read just single summary.
16461         (compute_working_sets): Use gcov_summary struct.
16462         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16463         of GCOV_COUNTERS_SUMMABLE.
16464         (GCOV_COUNTERS_SUMMABLE): Remove.
16465         (GCOV_FIRST_VALUE_COUNTER): Replace with
16466         GCOV_COUNTER_V_INTERVAL.
16467         (struct gcov_ctr_summary): Remove.
16468         (struct gcov_summary): Directly use fields of former
16469         gcov_ctr_summary.
16470         (compute_working_sets): Use gcov_summary struct.
16471         * gcov.c (read_count_file): Do not use ctrs fields.
16472         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16473         struct.
16474         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16475         struct.
16476         * profile.c: Likewise.
16477         * profile.h: Likewise.
16479 2018-06-05  Martin Liska  <mliska@suse.cz>
16481         PR gcov-profile/84846
16482         * gcov.c (output_lines): Print working directory only
16483         in intermediate format.
16485 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
16487         * config/s390/s390-builtin-types.def: Add void function type.
16488         * config/s390/s390-builtins.def: Use the function type for the
16489         tbeginc builtin.
16491 2018-06-04  Jim Wilson  <jimw@sifive.com>
16493         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16494         to int.
16495         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16496         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
16497         handle EH_RETURN_DATA_REGNO registers properly.
16498         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16499         (riscv_expand_epilogue): Update comment.  Change argument name and
16500         type.  Update code to use new name and type.  Pass new args to
16501         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
16502         EXCEPTION_RETURN.
16503         * config/riscv/riscv.md (NORMAL_RETURN): New.
16504         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16505         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16506         (eh_return): Call gen_eh_return_internal and emit barrier.
16507         (eh_return_internal): Call riscv_expand_epilogue.
16509 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
16511         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16512         bit_insertion field and declare can_be_merged_into method.
16513         (merged_store_group::can_be_merged_into): New method.
16514         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16515         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
16516         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16518 2018-06-04  Richard Biener  <rguenther@suse.de>
16520         PR tree-optimization/85955
16521         * builtins.c (fold_builtin_sincos): Convert pointers to
16522         destination to appropriate type before dereferencing.
16524 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
16526         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
16528 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
16530         * expr.c (expand_expr_real_1): Force the operand into memory if
16531         its TYPE_MODE is BLKmode and if there is no integer mode for
16532         the number of bits being extracted.
16534 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16536         PR target/85832
16537         PR target/86036
16538         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16539         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
16541 2018-06-04  Richard Biener  <rguenther@suse.de>
16543         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
16544         (cleanup_tree_cfg_noloop): ... single caller.  Do
16545         start_recording_case_labels later.
16547 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
16549         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
16550         to _IMMINTRIN_H_INCLUDED.
16551         * config/i386/pconfigintrin.h: Ditto.
16552         * config/i386/waitpkgintrin.h: Ditto.
16553         * config/i386/immintrin.h: Add includes for sgxintrin.h,
16554         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
16555         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16556         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
16557         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
16558         waitpkgintrin.h and cldemoteintrin.h.
16560 2018-06-04  Richard Biener  <rguenther@suse.de>
16562         PR tree-optimization/86038
16563         * tracer.c (find_best_successor): Check probability for
16564         being initialized, bail out if not.
16566 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
16568         PR target/86003
16569         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
16570         of bits to ignore when comparing architectures.
16572 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16574         PR tree-optimization/69615
16575         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
16576         maximum or minimum of the type, try to merge it also as if
16577         range1 is + [-, x - 1] or + [x + 1, -].
16579         PR c++/86025
16580         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
16582 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
16584         PR tree-optimization/86034
16585         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
16586         the unsigned bitfield type in a bit insertion sequence if it does not
16587         have a larger precision than the bitfield size.
16588         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
16590 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
16592         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
16594 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16596         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
16597         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
16598         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
16599         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
16601 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16603         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
16604         Disable -fdelete-null-pointer-checks for ELF toolchain.
16606 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16607             Kito Cheng  <kito.cheng@gmail.com>
16609         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
16610         (nds32le-*-*, nds32be-*-*): Integrate checking process.
16611         (nds32*-*-*): Add glibc and uclibc conditions.
16612         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
16613         (TARGET_EXCEPT_UNWIND_INFO): Define.
16614         * config/nds32/elf.h: New file.
16615         * config/nds32/linux.h: New file.
16616         * config/nds32/nds32-elf.opt: New file.
16617         * config/nds32/nds32-linux.opt: New file.
16618         * config/nds32/nds32-fp-as-gp.c
16619         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
16620         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
16621         TARGET_LINUX_ABI.
16622         (nds32_asm_file_end): Ditto.
16623         (nds32_print_operand): Ditto.
16624         (nds32_insert_attributes): Ditto.
16625         (nds32_init_libfuncs): New function.
16626         (TARGET_HAVE_TLS): Define.
16627         (TARGET_INIT_LIBFUNCS): Define.
16628         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
16629         spec content.
16630         (TARGET_ELF): Apply different mcmodel setting.
16631         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
16632         been migrated into elf.h and linux.h files.
16633         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
16634         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
16635         (mcmodel): The content has been migrated into nds32-elf.opt and
16636         nds32-linux.opt files.
16637         * config/nds32/t-elf: New file.
16638         * config/nds32/t-linux: New file.
16640 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16641             Shiva Chen  <shiva0217@gmail.com>
16643         * config/nds32/constants.md (unspec_volatile_element): Add
16644         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
16645         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
16646         optimization.
16647         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
16648         (make_pass_nds32_fp_as_gp): Declare.
16649         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
16650         optmization pass.
16651         (nds32_asm_function_end_prologue): Remove unused asm output.
16652         (nds32_asm_function_begin_epilogue): Remove unused asm output.
16653         (nds32_asm_file_start): Output necessary fp_as_gp information.
16654         (nds32_option_override): Adjust register usage.
16655         (nds32_expand_prologue): Consider fp_as_gp situation.
16656         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
16657         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
16658         (epilogue): Ditto.
16659         (return): Ditto.
16660         (simple_return): Ditto.
16661         (omit_fp_begin): Output special directive for fp_as_gp.
16662         (omit_fp_end): Output special directive for fp_as_gp.
16663         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
16664         mforbid-fp-as-gp): New options.
16666 2018-06-01  Mark Wielaard  <mark@klomp.org>
16668         * dwarf2out.c (dwarf2out_finish): Remove generation of
16669         DW_AT_loclists_base.
16671 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
16673         * gimple-ssa-store-merging.c: Include gimple-fold.h.
16674         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
16675         (struct merged_store_group): Add bit_insertion field.
16676         (dump_char_array): Use standard hexadecimal format.
16677         (merged_store_group::merged_store_group): Set bit_insertion to false.
16678         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
16679         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
16680         also print the mask in the dump file.
16681         (pass_store_merging::gate): Minor tweak.
16682         (imm_store_chain_info::coalesce_immediate): Fix wrong association
16683         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
16684         stores with INTEGER_CST stores.
16685         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
16686         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
16687         and use it throughout.  Generate bit insertion sequences if need be.
16688         (pass_store_merging::process_store): Remove redundant condition.
16689         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
16691 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
16693         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
16694         the 128-bit floating point types.  Fix function comment.
16696 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16698         * config/aarch64/aarch64-simd.md
16699         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
16700         mnemonics.
16701         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
16702         mnemonics.
16704 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16706         PR tree-optimization/85989
16707         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
16708         variable.
16709         (backprop::intersect_uses): Check it when deciding whether this
16710         is a backedge reference.
16711         (backprop::process_block): Add each phi to m_visited_phis
16712         after visiting it, then clear it at the end.
16714 2018-06-01  Richard Biener  <rguenther@suse.de>
16716         * tree-vectorizer.h (vect_dr_stmt): New function.
16717         (vect_get_load_cost): Adjust.
16718         (vect_get_store_cost): Likewise.
16719         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
16720         Use vect_dr_stmt instead of DR_SMTT.
16721         (vect_record_base_alignments): Likewise.
16722         (vect_calculate_target_alignment): Likewise.
16723         (vect_compute_data_ref_alignment): Likewise and make static.
16724         (vect_update_misalignment_for_peel): Likewise.
16725         (vect_verify_datarefs_alignment): Likewise.
16726         (vector_alignment_reachable_p): Likewise.
16727         (vect_get_data_access_cost): Likewise.  Pass down
16728         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
16729         (vect_get_peeling_costs_all_drs): Likewise.
16730         (vect_peeling_hash_get_lowest_cost): Likewise.
16731         (vect_enhance_data_refs_alignment): Likewise.
16732         (vect_find_same_alignment_drs): Likewise.
16733         (vect_analyze_data_refs_alignment): Likewise.
16734         (vect_analyze_group_access_1): Likewise.
16735         (vect_analyze_group_access): Likewise.
16736         (vect_analyze_data_ref_access): Likewise.
16737         (vect_analyze_data_ref_accesses): Likewise.
16738         (vect_vfa_segment_size): Likewise.
16739         (vect_small_gap_p): Likewise.
16740         (vectorizable_with_step_bound_p): Likewise.
16741         (vect_prune_runtime_alias_test_list): Likewise.
16742         (vect_analyze_data_refs): Likewise.
16743         (vect_supportable_dr_alignment): Likewise.
16744         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
16745         (vect_gen_prolog_loop_niters): Likewise.
16746         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
16747         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
16748         modify DR_STMT.
16749         (vect_recog_mask_conversion_pattern): Likewise.
16750         (vect_try_gather_scatter_pattern): Likewise.
16751         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
16752         to vect_get_store_cost.
16753         (vect_get_store_cost): Get stmt_info instead of DR.
16754         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
16755         (vect_get_load_cost): Get stmt_info instead of DR.
16757 2018-06-01  Richard Biener  <rguenther@suse.de>
16759         PR middle-end/86017
16760         * gimple-fold.c (var_decl_component_p): Also allow offsetted
16761         vars wrapped in MEM_REFs.
16763 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16765         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16766         Fix subreg tests so that we only return a choice between
16767         GENERAL_REGS and FP_REGS if the original classes included both.
16769 2018-06-01  Richard Biener  <rguenther@suse.de>
16771         PR ipa/85960
16772         * tree-ssa-structalias.c (get_function_part_constraint):
16773         Handle NULL fi->decl.
16774         (find_func_aliases_for_call): Properly handle indirect
16775         fi from direct call.
16776         (find_func_clobbers): Likewise.
16777         (ipa_pta_execute): Likewise.
16778         (create_variable_info_for): For functions that are ifunc_resolver
16779         resolve to a varinfo that contains the result of the resolver call.
16780         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
16781         aliases.
16783 2018-05-31  Michael Collison  <michael.collison@arm.com>
16785         * config/aarch64/aarch64.md:
16786         (*fix_to_zero_extenddfdi2): New pattern.
16787         * gcc.target/aarch64/fix_extend1.c: New testcase.
16789 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
16791         PR middle-end/78809
16792         PR middle-end/83026
16793         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
16794         and BUILT_IN_STRNCMP_EQ.
16795         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
16796         BUILT_IN_STRNCMP_EQ.
16797         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
16798         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16799         (gimple_fold_builtin): Likewise.
16800         * tree-ssa-strlen.c (compute_string_length): New function.
16801         (determine_min_obsize): New function.
16802         (handle_builtin_string_cmp): New function to handle calls to
16803         string compare functions.
16804         (strlen_optimize_stmt): Add handling to builtin string compare
16805         calls.
16806         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
16807         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16808         * tree.c (build_common_builtin_nodes): Add new defines of
16809         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
16811 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
16813         PR target/85984
16814         * bb-reorder.c (pass_partition_blocks::gate): Return false for
16815         functions with naked attribute.
16817 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
16819         * config/i386/sse.md (avx_vec_concat<mode>):
16820         Substitute concat_tg_mode mode attribute with xtg_mode.
16821         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
16822         (concat_tg_mode): Remove mode attribute.
16824 2018-05-31  Martin Sebor  <msebor@redhat.com>
16826         PR c/82063
16827         * calls.c (alloc_max_size): Correct a logic error/typo.
16828         Treat excessive arguments as infinite.  Warn for invalid arguments.
16829         * doc/invoke.texi (-Walloc-size-larger-than): Update.
16831 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
16833         PR target/85829
16834         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
16835         and movx for Haswell.
16837 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
16838             Cesar Philippidis  <cesar@codesourcery.com>
16840         PR middle-end/85879
16841         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
16842         when emitting error on private/firstprivate reductions.
16843         * omp-low.c (lower_omp_target): Avoid reference-type processing
16844         on pointers for firstprivate clause.
16846 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16848         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
16849         (st1x2): Likewise.
16850         (st1x3): Likewise.
16851         * config/aarch64/aarch64-simd.md
16852         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
16853         (aarch64_ld1_x3_<mode>): Likewise
16854         (aarch64_st1x2<VALLDIF:mode>): Likewise
16855         (aarch64_st1_x2_<mode>): Likewise
16856         (aarch64_st1x3<VALLDIF:mode>): Likewise
16857         (aarch64_st1_x3_<mode>): Likewise
16858         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
16859         (vld1_s8_x3): Likewise.
16860         (vld1_u16_x3): Likewise.
16861         (vld1_s16_x3): Likewise.
16862         (vld1_u32_x3): Likewise.
16863         (vld1_s32_x3): Likewise.
16864         (vld1_u64_x3): Likewise.
16865         (vld1_s64_x3): Likewise.
16866         (vld1_f16_x3): Likewise.
16867         (vld1_f32_x3): Likewise.
16868         (vld1_f64_x3): Likewise.
16869         (vld1_p8_x3): Likewise.
16870         (vld1_p16_x3): Likewise.
16871         (vld1_p64_x3): Likewise.
16872         (vld1q_u8_x3): Likewise.
16873         (vld1q_s8_x3): Likewise.
16874         (vld1q_u16_x3): Likewise.
16875         (vld1q_s16_x3): Likewise.
16876         (vld1q_u32_x3): Likewise.
16877         (vld1q_s32_x3): Likewise.
16878         (vld1q_u64_x3): Likewise.
16879         (vld1q_s64_x3): Likewise.
16880         (vld1q_f16_x3): Likewise.
16881         (vld1q_f32_x3): Likewise.
16882         (vld1q_f64_x3): Likewise.
16883         (vld1q_p8_x3): Likewise.
16884         (vld1q_p16_x3): Likewise.
16885         (vld1q_p64_x3): Likewise.
16886         (vst1_s64_x2): Likewise.
16887         (vst1_u64_x2): Likewise.
16888         (vst1_f64_x2): Likewise.
16889         (vst1_s8_x2): Likewise.
16890         (vst1_p8_x2): Likewise.
16891         (vst1_s16_x2): Likewise.
16892         (vst1_p16_x2): Likewise.
16893         (vst1_s32_x2): Likewise.
16894         (vst1_u8_x2): Likewise.
16895         (vst1_u16_x2): Likewise.
16896         (vst1_u32_x2): Likewise.
16897         (vst1_f16_x2): Likewise.
16898         (vst1_f32_x2): Likewise.
16899         (vst1_p64_x2): Likewise.
16900         (vst1q_s8_x2): Likewise.
16901         (vst1q_p8_x2): Likewise.
16902         (vst1q_s16_x2): Likewise.
16903         (vst1q_p16_x2): Likewise.
16904         (vst1q_s32_x2): Likewise.
16905         (vst1q_s64_x2): Likewise.
16906         (vst1q_u8_x2): Likewise.
16907         (vst1q_u16_x2): Likewise.
16908         (vst1q_u32_x2): Likewise.
16909         (vst1q_u64_x2): Likewise.
16910         (vst1q_f16_x2): Likewise.
16911         (vst1q_f32_x2): Likewise.
16912         (vst1q_f64_x2): Likewise.
16913         (vst1q_p64_x2): Likewise.
16914         (vst1_s64_x3): Likewise.
16915         (vst1_u64_x3): Likewise.
16916         (vst1_f64_x3): Likewise.
16917         (vst1_s8_x3): Likewise.
16918         (vst1_p8_x3): Likewise.
16919         (vst1_s16_x3): Likewise.
16920         (vst1_p16_x3): Likewise.
16921         (vst1_s32_x3): Likewise.
16922         (vst1_u8_x3): Likewise.
16923         (vst1_u16_x3): Likewise.
16924         (vst1_u32_x3): Likewise.
16925         (vst1_f16_x3): Likewise.
16926         (vst1_f32_x3): Likewise.
16927         (vst1_p64_x3): Likewise.
16928         (vst1q_s8_x3): Likewise.
16929         (vst1q_p8_x3): Likewise.
16930         (vst1q_s16_x3): Likewise.
16931         (vst1q_p16_x3): Likewise.
16932         (vst1q_s32_x3): Likewise.
16933         (vst1q_s64_x3): Likewise.
16934         (vst1q_u8_x3): Likewise.
16935         (vst1q_u16_x3): Likewise.
16936         (vst1q_u32_x3): Likewise.
16937         (vst1q_u64_x3): Likewise.
16938         (vst1q_f16_x3): Likewise.
16939         (vst1q_f32_x3): Likewise.
16940         (vst1q_f64_x3): Likewise.
16941         (vst1q_p64_x3): Likewise.
16943 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16945         * config/msp430/msp430.c (msp430_output_labelref): Prepend
16946         user_label_prefix to name.
16948         * tree-core.h: Update comment about the format of NAME string
16949         passed to handler in attribute_spec.
16951         * config/msp430/msp430.md: Remove erroneous subreg expression from
16952         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
16953         zero_extend{q,h}isi2.
16955 2018-05-30  Borislav Petkov  <bp@suse.de>
16957         * doc/extend.texi: Document some architecture specific
16958         constraints and sort entries.
16960 2018-05-30  Martin Sebor  <msebor@redhat.com>
16962         PR middle-end/85369
16963         * builtins.c (expand_builtin_stpcpy_1): New function.
16964         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
16965         only if the former succeeds.
16967 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16969         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
16970         in saphira.
16972 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16974         * doc/invoke.texi (-flinker-output): Document
16976 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16978         * passes.c (ipa_write_summaries): Only modify statements if body
16979         is in memory.
16980         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
16981         incrementally linking.
16982         (ipa_passes): Likewise.
16983         * lto-cgraph.c (lto_output_node): When incrementally linking do not
16984         pass down resolution info.
16985         * common.opt (flag_incremental_link): Update info.
16986         * gcc.c (plugin specs): Turn flinker-output=* to
16987         -plugin-opt=-linker-output-known
16988         * toplev.c (compile_file): Also cut compilation when doing incremental
16989         link.
16990         * flag-types. (enum lto_partition_model): Add
16991         LTO_LINKER_OUTPUT_NOLTOREL.
16992         (invoke.texi): Add -flinker-output docs.
16993         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
16994         link same way as WPA; do not stream in dead initializers.
16996         * dwarf2out.c (dwarf2out_die_ref_for_decl,
16997         darf2out_register_external_decl): Support incremental link.
16999 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17001         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17003 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17005         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17006         it down to simple_object_copy_lto_debug_sections.
17007         (run_gcc): Determine incremental LTO link time and configure
17008         lto1 into non-wpa mode, disable renaming of debug sections.
17010 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17012         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17013         descriptions of various incorrectly documented functions.
17015 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17017         Revert:
17018         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17019         address check not strict.
17021 2018-05-30  Richard Biener  <rguenther@suse.de>
17023         PR tree-optimization/85964
17024         * tracer.c (better_p): Drop initialized count check, we only
17025         call the function with initialized counts now.
17026         (find_best_successor): Do find a best edge if one
17027         has uninitialized count.
17028         (find_best_predecessor): Likewise.  Do BB frequency check only
17029         if count is initialized.
17031 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17033         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17034         (aarch64_ldrstr_offset_compare): New.
17035         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17036         load/store orderings.
17037         (aarch64_gen_adjusted_ldpstp): Likewise.
17039 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17041         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17042         Check for subset of GENERAL_REGS and FP_REGS.
17043         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17044         r=w alternative.
17046 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17048         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17049         and wi::to_poly_offset.  Add the current offset and then check
17050         whether the sum fits, rather than using an unchecked addition of
17051         a checked term.  Check for a shwi rather than a uhwi.
17052         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17053         (store_constructor): Use poly_int_tree_p.
17054         (expand_expr_real_1): Likewise.
17055         * function.c (assign_temp): Likewise.
17056         * fold-const.c (const_binop): Use poly_int_tree_p and
17057         wi::to_poly_offset.
17058         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17059         division.
17060         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17061         to_poly_offset for MEM offsets.
17062         * ipa-icf.c (sem_variable::equals): Likewise.
17063         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17064         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17065         wi::to_poly_offset for BIT_FIELD_REF offsets.
17066         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17067         wi::to_poly_offset.
17068         * var-tracking.c (emit_note_insn_var_location): Use
17069         tree_to_poly_uint64.
17071 2018-05-29  Jim Wilson  <jimw@sifive.com>
17073         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17075 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17077         PR target/85950
17078         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17079         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17080         sequence.
17081         (sse4_1_round<mode>2): Use nonimmediate_operand
17082         for operand 1 predicate.
17084 2018-05-29  Martin Sebor  <msebor@redhat.com>
17085             Richard Biener  <rguenther@suse.de>
17087         PR testsuite/85888
17088         * calls.c (get_size_range): Call determine_value_range instead
17089         of get_value_range..
17090         * tree-vrp.h (determine_value_range): Declared new function.
17091         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17093 2018-05-29  Richard Biener  <rguenther@suse.de>
17095         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17096         sure to use non-pattern stmts for get_earlier_stmt arguments.
17097         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17098         called on pattern stmts.
17099         (get_later_stmt): Likewise.
17101 2018-05-29  Martin Liska  <mliska@suse.cz>
17103         PR gcov-profile/85759
17104         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17105         env variables.
17107 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17109         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17110         VEC_UNPACK_*_EXPR.
17111         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17112         VEC_PACK_*_EXPR.
17114         PR target/85918
17115         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17116         VEC_PACK_FLOAT_EXPR): New tree codes.
17117         * tree-pretty-print.c (op_code_prio): Handle
17118         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17119         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17120         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17121         * tree-inline.c (estimate_operator_cost): Likewise.
17122         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17123         * fold-const.c (const_binop): Likewise.
17124         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17125         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17126         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17127         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17128         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17129         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17130         * expr.c (expand_expr_real_2): Likewise.
17131         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17132         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17133         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17134         optabs.
17135         * optabs.c (expand_widen_pattern_expr): For
17136         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17137         sign from result type rather than operand's type.
17138         (expand_binop_directly): For vec_packu_float_optab and
17139         vec_packs_float_optab allow result type to be different from operand's
17140         type.
17141         * optabs-tree.c (optab_for_tree_code): Handle
17142         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17143         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
17144         * tree-vect-generic.c (expand_vector_operations_1):  Handle
17145         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17146         VEC_PACK_FLOAT_EXPR.
17147         * tree-vect-stmts.c (supportable_widening_operation): Handle
17148         FIX_TRUNC_EXPR.
17149         (supportable_narrowing_operation): Handle FLOAT_EXPR.
17150         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17151         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17152         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
17153         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17154         mode attributes.
17155         (vec_pack<floatprefix>_float_<mode>): New expander.
17156         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17157         attributes.
17158         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17159         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17160         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17161         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17162         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17163         Document.
17164         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17165         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17166         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17167         VEC_PACK_FLOAT_EXPR): Document.
17169 2018-05-29  Richard Biener  <rguenther@suse.de>
17171         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17172         member.
17173         (stmt_vec_info_vec): Make pointer.
17174         (init_stmt_vec_info_vec): Remove.
17175         (free_stmt_vec_info_vec): Likewise.
17176         (set_stmt_vec_info_vec): New function.
17177         (free_stmt_vec_infos): Likewise.
17178         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17179         (set_vinfo_for_stmt): Likewise.
17180         (get_earlier_stmt): Likewise.
17181         (get_later_stmt): Likewise.
17182         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17183         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17184         (vec_info::~vec_info): Free stmt_vec_infos.
17185         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17186         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17187         (pass_slp_vectorize::execute): Likewise.
17188         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17189         (free_stmt_vec_info_vec): Likewise.
17190         (set_stmt_vec_info_vec): New function.
17191         (free_stmt_vec_infos): Likewise.
17192         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17193         the global stmt_vec_info_vec.
17194         * tree-parloops.c (gather_scalar_reductions): Use
17195         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17196         vector.
17198 2018-05-29  Richard Biener  <rguenther@suse.de>
17200         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17202 2018-05-29  Martin Liska  <mliska@suse.cz>
17203             David Malcolm  <dmalcolm@redhat.com>
17205         * vec.c (test_reverse): New.
17206         (vec_c_tests): Add new test.
17207         * vec.h (vl_ptr>::reverse): New function.
17209 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
17211         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17213         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17214         and later.
17216 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17218         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17220 2018-05-28  Richard Biener  <rguenther@suse.de>
17222         PR tree-optimization/85933
17223         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17224         look at stmts marked as vectorizable.
17226 2018-05-28  Richard Biener  <rguenther@suse.de>
17228         PR tree-optimization/85934
17229         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17230         vector boolean check before scalar optimization.
17232 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17234         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17235         for armv5te.
17237 2018-05-28  Mark Wielaard  <mark@klomp.org>
17239         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17240         if it is an expression containing a minus sign.
17242 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17244         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17246 2018-05-27  Paul Koning  <ni1d@arrl.net>
17248         * config/pdp11/pdp11.md (truncsihi2): Remove.
17250 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17251             Chung-Ju Wu  <jasonwucj@gmail.com>
17253         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17254         implementation.
17255         (unaligned_store_dw): Ditto.
17256         * config/nds32/nds32-memory-manipulation.c
17257         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17258         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17259         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17260         (emit_setmem_word_loop): Rename to ...
17261         (emit_setmem_doubleword_loop): ... this.
17262         (nds32_gen_dup_4_byte_to_word_value): New function.
17263         (nds32_gen_dup_8_byte_to_double_word_value): New function.
17264         (nds32_expand_setmem_loop): Refine implementation.
17265         (nds32_expand_setmem_loop_v3m): Ditto.
17266         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17267         pattern.
17269 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17271         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17273 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17275         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17276         (nds32_init_machine_status): Initialize machine->attr_naked_p and
17277         machine->attr_no_prologue_p.
17278         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17279         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17280         (nds32_expand_epilogue): Consider attr_naked_p.
17281         (nds32_expand_epilogue_v3pop): Likewise.
17282         (nds32_can_use_return_insn): Likewise.
17283         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17284         attr_no_prologue_p fields.
17285         * config/nds32/nds32.opt (mret-in-naked-func): New option.
17287 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
17289         PR target/85918
17290         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17291         attributes.
17292         * config/i386/sse.md
17293         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17294         Rename to ...
17295         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17296         ... this.
17297         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17298         Rename to ...
17299         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17300         ... this.
17301         (*<floatsuffix>floatv2div2sf2): Rename to ...
17302         (*float<floatunssuffix>v2div2sf2): ... this.
17303         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17304         (float<floatunssuffix>v2div2sf2_mask): ... this.
17305         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17306         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17307         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17308         to ...
17309         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17310         ... this.
17311         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17312         Rename to ...
17313         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17314         ... this.
17315         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17316         Rename to ...
17317         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17318         ... this.
17319         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17320         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17321         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17322         gen_ufix_truncv8dfv8si2.
17323         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17324         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17325         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17326         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17327         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17328         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17329         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17330         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17332 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17334         PR target/85900
17335         PR target/85345
17336         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17338 2018-05-25  Jim Wilson  <jimw@sifive.com>
17340         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17341         * config/riscv/riscv.c (struct machine_function): Add
17342         interrupt_handler_p and attribute_checked_p fields.
17343         (riscv_attribute_table): Add interrupt.
17344         (riscv_interrupt_type_p): New.
17345         (riscv_save_reg_p): Save extra regs for interrupt handler.
17346         (riscv_use_save_libcall): Return false  for interrupt handler.
17347         (riscv_first_stack_step): Add forward declaration.
17348         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
17349         for interrupt handler with large frame.  Use it for saved reg list.
17350         (riscv_expand_prologue): Move flag_stack_usage_info support to
17351         eliminate duplication.
17352         (riscv_expand_epilogue): Generate mret for interrupt handler.
17353         (riscv_epilogue_uses): New.
17354         (riscv_can_use_return_insn): Return false for interrupt handler.
17355         (riscv_function_ok_for_sibcall): Likewise.
17356         (riscv_set_current_function): Add interrupt handler support.
17357         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17358         * config/riscv/riscv.md (UNSPECV_MRET): New.
17359         (GP_REGNUM): New.
17360         (riscv_frflags, riscv_fsflags): Use tab after opcode.
17361         (riscv_mret): New.
17362         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17364 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
17366         PR tree-optimization/85712
17367         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17368         this candidate has already been replaced in-situ by a copy.
17370 2018-05-25  Jason Merrill  <jason@redhat.com>
17372         PR c++/80485 - inline function non-zero address.
17373         * symtab.c (nonzero_address): Check DECL_COMDAT.
17375 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17377         PR target/83628
17378         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17379         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17380         extension of SImode operation.  Use const123_operand predicate.
17381         (*saddsi_1): Remove.
17382         (*saddl_se_1): Ditto.
17383         (*ssubsi_1): Ditto.
17384         (*ssubl_se_1): Ditto.
17385         * config/alpha/predicates.md (const123_operand): New predicate.
17386         * config/alpha/constraints.md (P): Use IN_RANGE.
17388 2018-05-25  Richard Biener  <rguenther@suse.de>
17390         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17391         defaulted to true.
17392         (ref_maybe_used_by_stmt_p): Likewise.
17393         (stmt_may_clobber_ref_p): Likewise.
17394         (stmt_may_clobber_ref_p_1): Likewise.
17395         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17396         and pass it along.
17397         (ref_maybe_used_by_stmt_p): Likewise.
17398         (stmt_may_clobber_ref_p): Likewise.
17399         (stmt_may_clobber_ref_p_1): Likewise.
17400         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17401         the alias oracle to disambiguate DRs with stmts DR analysis
17402         couldn't handle.
17403         (vect_analyze_data_refs): Do not give up on not analyzable
17404         DRs for BB vectorization.  Remove code truncating the dataref
17405         vector.
17407 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
17409         PR target/85832
17410         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17411         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
17412         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17414 2018-05-25  Richard Biener  <rguenther@suse.de>
17416         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17417         function, combining stmt data ref gathering and fatal analysis
17418         parts.
17419         (vect_analyze_data_refs): Remove now redudnant code and simplify.
17420         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17421         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17422         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17423         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17425 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17427         PR tree-optimization/85720
17428         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17429         SCC if all partitions are builtins.
17430         (version_loop_by_alias_check): New parameter.  Generate cancelable
17431         runtime alias check if all partitions are builtins.
17432         (distribute_loop): Update call to above function.
17434 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17436         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17437         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17438         (parm_default_def_partition_arg): Ditto.
17439         (set_parm_default_def_partition): Ditto.
17440         (get_parm_default_def_partitions): Ditto and make it static.
17441         (get_undefined_value_partitions): Ditto and make it static.
17442         (remove_ssa_form): Refactor call to init_var_map here.
17443         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17444         computation for loop region.
17445         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17446         (register_default_def): Delete.
17447         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17448         (parm_default_def_partition_arg): Ditto.
17449         (set_parm_default_def_partition): Ditto.
17450         (get_parm_default_def_partitions): Ditto and make it static.
17451         (get_undefined_value_partitions): Ditto and make it static.
17452         (coalesce_with_default, coalesce_with_default): Update comment.
17453         (create_coalesce_list_for_region): New func factored out from
17454         create_outofssa_var_map.
17455         (populate_coalesce_list_for_outofssa): New func factored out from
17456         create_outofssa_var_map and coalesce_ssa_name.
17457         (create_outofssa_var_map): Delete.
17458         (coalesce_ssa_name): Refactor to support live range computation.
17459         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17460         (get_parm_default_def_partitions): Delete.
17461         (get_undefined_value_partitions): Ditto.
17462         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17463         computation for loop region.
17464         (new_tree_live_info, loe_visit_block): Ditto.
17465         (live_worklist, set_var_live_on_entry): Ditto.
17466         (calculate_live_on_exit, verify_live_on_entry): Ditto.
17467         * tree-ssa-live.h (struct _var_map): New fields.
17468         (init_var_map): Change decl.
17469         (region_contains_p): New.
17471 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17473         * tree-ssa-live.h (live_merge_and_clear): Delete.
17475 2018-05-25  Richard Biener  <rguenther@suse.de>
17477         PR c++/85912
17478         * tree-dump.c (dequeue_and_dump): Remove access to removed
17479         operand 2 of a SWITCH_EXPR.
17481 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17483         * doc/sourcebuild.texi (vect_double_cond_arith): Include
17484         multiplication and division.
17485         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17486         (cond_udiv@var{m}, cond_umod@var{m}): Document.
17487         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17488         (cond_udiv_optab, cond_umod_optab): New optabs.
17489         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17490         (IFN_COND_RDIV): New internal functions.
17491         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17492         TRUNC_MOD_EXPR and RDIV_EXPR.
17493         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17494         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17495         New unspecs.
17496         (SVE_INT_BINARY): Include mult.
17497         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17498         (optab, sve_int_op): Handle mult.
17499         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17500         UNSPEC_COND_DIV.
17501         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17502         for SVE_INT_BINARY_SD.
17504 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17506         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17507         (optab, sve_int_op): Handle div and udiv.
17508         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17509         for SVE_INT_BINARY_SD.
17510         (*<optab><mode>3): New insn for the same.
17512 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17514         * tree-vect-patterns.c: Include predict.h.
17515         (vect_recog_divmod_pattern): Restrict check for division support
17516         to when optimizing for size.
17518 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17520         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17521         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17522         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17523         (gimple_match_op::set_op): Likewise.
17524         (gimple_resimplify4): Declare.
17525         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
17526         (expr::gen_transform): Likewise.
17527         (decision_tree::gen): Generate a simplification routine for 4 operands.
17528         * gimple-match-head.c (gimple_simplify): Add an overload for
17529         4 operands.  In the top-level function, handle up to 4 call
17530         arguments and call gimple_resimplify4.
17531         (gimple_resimplify4): New function.
17532         (build_call_internal): Pass a fourth operand.
17533         (maybe_push_to_seq): Likewise.
17534         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
17535         Fold VEC_COND_EXPRs of an operation and a default value into
17536         an IFN_COND_* function if possible.
17537         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
17538         New unspecs.
17539         (SVE_COND_FP_BINARY): Include them.
17540         (optab, sve_fp_op): Handle them.
17541         (SVE_INT_BINARY_REV): New code iterator.
17542         (SVE_COND_FP_BINARY_REV): New int iterator.
17543         (commutative): New int attribute.
17544         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
17545         Declare.
17546         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
17547         function.
17548         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
17549         (*cond_<optab><mode>): New patterns for reversed operands.
17551 2018-05-25  Richard Biener  <rguenther@suse.de>
17553         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
17554         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
17555         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
17556         (STMT_VINFO_GROUPED_ACCESS): Adjust.
17557         * tree-vect-data-refs.c (everywhere): Adjust users.
17558         * tree-vect-loop.c (everywhere): Likewise.
17559         * tree-vect-slp.c (everywhere): Likewise.
17560         * tree-vect-stmts.c (everywhere): Likewise.
17561         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
17563 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17565         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
17566         Rename to...
17567         (gcc_cv_as_section_exclude): ... this.
17568         Try Solaris as #exclude syntax.
17569         * configure: Regenerate.
17570         * config.in: Regenerate.
17571         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
17572         SECTION_EXCLUDE.
17573         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
17574         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
17576         * varasm.c (default_elf_asm_named_section): Don't check if
17577         HAVE_GAS_SECTION_EXCLUDE is defined.
17579 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17581         * doc/md.texi: Update the documentation of the cond_* optabs
17582         to mention the new final operand.  Fix GET_MODE_NUNITS call.
17583         Describe the scalar case too.
17584         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
17585         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
17586         instead of 2.
17587         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
17588         (get_conditional_internal_fn): Update comment.
17589         * tree-vect-loop.c (vectorizable_reduction): Pass the original
17590         accumulator value as a final argument to conditional functions.
17591         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
17592         a define_expand and add an "else" operand.  Assert for now that
17593         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
17594         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
17595         (*cond_<optab><mode>): New patterns.
17596         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
17597         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17598         (UNSPEC_COND_EOR): Delete.
17599         (optab): Remove associated mappings.
17600         (SVE_INT_BINARY): New code iterator.
17601         (sve_int_op): Remove int attribute and add "minus" to the code
17602         attribute.
17603         (SVE_COND_INT_OP): Delete.
17604         (SVE_COND_FP_OP): Rename to...
17605         (SVE_COND_FP_BINARY): ...this.
17607 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17609         * optabs.c (can_reuse_operands_p): New function.
17610         (maybe_legitimize_operands): Try to reuse the results for
17611         earlier operands.
17613 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17615         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
17616         Add {q} suffix to insn mnemonic.
17618 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17620         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
17621         (msp430_warn_func_return): New.
17623 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
17625         * fold-const.c (tree_nonzero_bits): New function.
17626         * fold-const.h (tree_nonzero_bits): Likewise.
17627         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
17628         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
17630 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17632         PR target/85900
17633         PR target/85345
17634         * varasm.c (assemble_alias): Check ifunc_resolver only on
17635         FUNCTION_DECL.
17637 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17639         PR target/85903
17640         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
17641         when memory input operand is handled.
17643 2018-05-24  Luis Machado  <luis.machado@linaro.org>
17645         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
17646         global.
17647         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
17649 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17651         * match.pd: Delay FMA folds until after vectorization.
17653 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17655         PR target/83009
17656         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17657         address check not strict.
17659 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17661         * gimple-match.h (gimple_match_op): New class.
17662         (mprts_hook): Replace parameters with a gimple_match_op *.
17663         (maybe_build_generic_op): Likewise.
17664         (gimple_simplified_result_is_gimple_val): Replace parameters with
17665         a const gimple_match_op *.
17666         (gimple_simplify): Replace code_helper * and tree * parameters with
17667         a gimple_match_op * parameter.
17668         (gimple_resimplify1): Replace code_helper *, tree and tree *
17669         parameters with a gimple_match_op * parameter.
17670         (gimple_resimplify2): Likewise.
17671         (gimple_resimplify3): Likewise.
17672         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
17673         parameters with a gimple_match_op * parameter.
17674         * gimple-match-head.c (gimple_simplify): Change prototypes of
17675         auto-generated functions to take a gimple_match_op * instead of
17676         separate code_helper * and tree * parameters.  Make the same
17677         change in the top-level overload and update calls to the
17678         gimple_resimplify routines.  Update calls to the auto-generated
17679         functions and to maybe_push_res_to_seq in the publicly-facing
17680         operation-specific gimple_simplify overloads.
17681         (gimple_match_op::MAX_NUM_OPS): Define.
17682         (gimple_resimplify1): Replace rcode and ops with a single res_op
17683         parameter.  Update call to gimple_simplify.
17684         (gimple_resimplify2): Likewise.
17685         (gimple_resimplify3): Likewise.
17686         (mprts_hook): Replace parameters with a gimple_match_op *.
17687         (maybe_build_generic_op): Likewise.
17688         (build_call_internal): Replace type, nargs and ops with
17689         a gimple_match_op *.
17690         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
17691         with a single gimple_match_op *.  Update calls to mprts_hook,
17692         build_call_internal and gimple_simplified_result_is_gimple_val.
17693         Factor out code that is common to the tree_code and combined_fn cases.
17694         * genmatch.c (expr::gen_transform): Replace tem_code and
17695         tem_ops with a gimple_match_op called tem_op.  Update calls
17696         to the gimple_resimplify functions and maybe_push_res_to_seq.
17697         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
17698         res_ops.  Update call to the gimple_resimplify functions.
17699         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
17700         (decision_tree::gen): Make the functions take a gimple_match_op *
17701         called res_op instead of separate res_code and res_ops parameters.
17702         Update call accordingly.
17703         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
17704         and ops with a single res_op parameter.  Update calls to
17705         maybe_build_generic_op and maybe_push_res_to_seq.
17706         (fold_stmt_1): Update calls to gimple_simplify and
17707         replace_stmt_with_simplification.
17708         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
17709         and gimple_simplified_result_is_gimple_val.
17710         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
17711         gimple_simplify.
17712         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
17713         with a gimple_match_op *.
17714         (vn_nary_build_or_lookup): Likewise.  Update call to
17715         vn_nary_build_or_lookup_1.
17716         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
17717         gimple_match_op *.  Update calls to the gimple_resimplify routines
17718         and to gimple_simplified_result_is_gimple_val.
17719         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
17720         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
17721         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
17722         (visit_nary_op): Likewise.
17723         (visit_reference_op_load): Likewise.
17725 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17727         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
17728         modifier for printing the step amount.
17730 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
17732         PR target/78849
17733         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
17734         types.
17736 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
17738         * doc/sourcebuild.texi (Endianness): New subsubsection.
17740 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17742         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17743         <prefetch_dynamic_strides>: New const bool field.
17744         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17745         prefetch_dynamic_strides.
17746         (exynosm1_prefetch_tune): Likewise.
17747         (thunderxt88_prefetch_tune): Likewise.
17748         (thunderx_prefetch_tune): Likewise.
17749         (thunderx2t99_prefetch_tune): Likewise.
17750         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
17751         false.
17752         (aarch64_override_options_internal): Update to set
17753         PARAM_PREFETCH_DYNAMIC_STRIDES.
17754         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17755         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17756         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17757         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17758         prefetch-dynamic-strides setting.
17760 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17762         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17763         <minimum_stride>: New const int field.
17764         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17765         minimum_stride field defaulting to -1.
17766         (exynosm1_prefetch_tune): Likewise.
17767         (thunderxt88_prefetch_tune): Likewise.
17768         (thunderx_prefetch_tune): Likewise.
17769         (thunderx2t99_prefetch_tune): Likewise.
17770         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
17771         <default_opt_level>: Set to 3.
17772         (aarch64_override_options_internal): Update to set
17773         PARAM_PREFETCH_MINIMUM_STRIDE.
17774         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17775         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17776         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17777         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17778         stride is constant and is below the minimum stride threshold.
17780 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17782         * config/arm/arm-cpus.in (mode26): Delete.
17783         (armv4): Delete mode26 reference.
17784         * config/arm/arm.c (arm_configure_build_target): Delete use of
17785         isa_bit_mode26.
17787 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17789         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
17790         New insn pattern.
17791         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
17792         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
17793         for non-SSE modes.
17794         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
17795         (floatunsdidf2): Ditto.
17797 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17799         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
17800         (fixuns_trunc<mode>si2_avx512f): Ditto.
17801         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
17802         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
17803         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
17805 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17807         PR rtl-optimization/79985
17808         * df-scan.c (df_insn_refs_collect): Remove special case for
17809         global registers and asm statements.
17811 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17813         * extend.texi (Global Register Variables): Rewrite the bullet list.
17814         Note that the register is available for allocation. Note that access
17815         via inline asm must use constraints. Add note about async-signal
17816         handlers. Remove paragraph about automagic register selection.
17818 2018-05-23  Richard Biener  <rguenther@suse.de>
17820         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
17821         of fixed offset from memset VN.
17823 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
17825         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
17826         first_interp field.
17827         (alloc_cand_and_find_basis): Initialize first_interp field.
17828         (slsr_process_mul): Modify first_interp field.
17829         (slsr_process_add): Likewise.
17830         (slsr_process_cast): Modify first_interp field for each new
17831         interpretation.
17832         (slsr_process_copy): Likewise.
17833         (dump_candidate): Dump first_interp field.
17834         (replace_mult_candidate): Process all interpretations, not just
17835         subsequent ones.
17836         (replace_rhs_if_not_dup): Likewise.
17837         (replace_one_candidate): Likewise.
17839 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
17841         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
17842         Add new boolean.
17843         (aarch64_needs_frame_chain): New function.
17844         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
17846 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
17848         PR target/84882
17849         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
17850         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
17851         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
17852         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
17853         as true for strict-align.
17854         (aarch64_can_inline_p): Perform checks even when callee has no
17855         attributes to check for strict alignment.
17856         * doc/extend.texi (AArch64 Function Attributes): Document
17857         no-strict-align.
17858         * doc/invoke.texi: (AArch64 Options): Likewise.
17860 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
17862         PR tree-optimization/85853
17863         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
17864         the handling of the root of the node to...
17865         (vect_slp_analyze_node_operations_1): ...this new function,
17866         and run the whole thing with the child nodes' def types
17867         set according to their SLP node's def type.
17869 2018-05-23  Richard Biener  <rguenther@suse.de>
17871         PR middle-end/85874
17872         * tree-data-ref.c (create_runtime_alias_checks): Defer
17873         and ignore overflow warnings.
17875 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
17877         PR tree-optimization/85822
17878         * tree-vrp.c (is_masked_range_test): Fix handling of negative
17879         constants.
17881 2018-05-23  Richard Biener  <rguenther@suse.de>
17883         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17884         memset constants via native_interpret_expr.
17886 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
17888         PR target/85345
17889         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
17890         attribute.
17891         (cgraph_node::create_alias): Likewise.
17892         (cgraph_node::get_availability): Check ifunc_resolver instead
17893         of looking up ifunc attribute.
17894         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
17895         * varasm.c (do_assemble_alias): Likewise.
17896         (assemble_alias): Likewise.
17897         (default_binds_local_p_3): Likewise.
17898         * cgraph.h (cgraph_node): Add ifunc_resolver.
17899         (cgraph_node::only_called_directly_or_aliased_p): Return false
17900         for IFUNC resolver.
17901         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
17902         attribute.
17903         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
17904         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
17905         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
17906         instead of looking up ifunc attribute.
17908 2018-05-22  Luis Machado  <luis.machado@linaro.org>
17910         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
17912 2018-05-22  Martin Sebor  <msebor@redhat.com>
17914         PR middle-end/85359
17915         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
17916         only when expasion succeeds.
17917         (expand_builtin_strcmp): Same.
17918         (expand_builtin_strncmp): Same.
17920 2018-05-22  Martin Sebor  <msebor@redhat.com>
17922         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
17924 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17925             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17927         * config/aarch64/aarch64-ldpstp.md: Replace uses of
17928         aarch64_mem_pair_operand with memory_operand and delete operand swapping
17929         code.
17930         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17931         Add check for legitimate_address.
17932         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
17933         (aarch64_swap_ldrstr_operands): New.
17934         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
17935         Define prototype.
17937 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17938             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17940         * config/aarch64/aarch64.md: New patterns to generate stp
17941         and ldp.
17942         (store_pair_sw, store_pair_dw): New patterns to generate stp for
17943         single words and double words.
17944         (load_pair_sw, load_pair_dw): Likewise.
17945         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
17946         Delete.
17947         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
17948         Delete.
17949         * config/aarch64/aarch64-ldpstp.md: Modify peephole
17950         for different mode ldpstp and add peephole for merged zero stores.
17951         Likewise for loads.
17952         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17953         Add size check.
17954         (aarch64_gen_store_pair): Rename calls to match new patterns.
17955         (aarch64_gen_load_pair): Rename calls to match new patterns.
17956         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
17957         (load_pair<DREG:mode><DREG2:mode>): ... This.
17958         (store_pair<mode>): Rename to...
17959         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
17960         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
17961         New mode iterators.
17962         (V_INT_EQUIV): Handle SImode.
17963         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
17964         New predicate.
17966 2018-05-22  Martin Sebor  <msebor@redhat.com>
17968         PR c/85623
17969         * calls.c (maybe_warn_nonstring_arg): Use string length to set
17970         or ajust the presumed bound on an operation to avoid unnecessary
17971         warnings.
17973 2018-05-22  Martin Sebor  <msebor@redhat.com>
17975         PR tree-optimization/85826
17976         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
17977         assuming that a DECL necesarily has a constant size.
17979 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
17981         PR middle-end/85862
17982         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
17984 2018-05-22  Richard Biener  <rguenther@suse.de>
17986         PR tree-optimization/85834
17987         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
17988         non-constant and non-zero memset arguments.
17990 2018-05-22  Martin Liska  <mliska@suse.cz>
17992         PR ipa/85607
17993         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
17995 2018-05-22  Richard Biener  <rguenther@suse.de>
17997         PR tree-optimization/85863
17998         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
17999         comparisons when vectype is specified.
18000         (vectorizable_condition): Do not specify vectype for
18001         vect_is_simple_cond when SLP vectorizing.
18003 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18005         PR target/85657
18006         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18007         define __ibm128 as long double.
18008         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18009         as a distinct type when IEEE 128-bit support is enabled.
18010         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18011         128-bit types to use the correct functions.
18012         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18013         convert between 128-bit floating point types that have different
18014         modes but the same representation, instead of using gen_lowpart to
18015         makean alias.
18016         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18017         KFmode.
18018         (IFKF_reg): New attributes to give the register constraints for
18019         IFmode and KFmode.
18020         (extend<mode>tf2_internal): New insns to mark an explicit
18021         conversion between 128-bit floating point types that have a
18022         different mode but share the same representation.
18024 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18026         PR tree-optimization/85814
18027         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18028         a null return from get_strinfo when unsharing the next
18029         strinfo in the chain.
18031 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18033         PR gcc/84923
18034         * varasm.c (weak_finish): Clean up weak_decls.
18036 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18038         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18039         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18040         UNSPEC_UADALP values.
18041         * config/aarch64/iterators.md (ABAL): New int iterator.
18042         (ABDL2): Likewise.
18043         (ADALP): Likewise.
18044         (sur): Add mappings for the above.
18045         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18046         New define_insn.
18047         (aarch64_<sur>abal<mode>_4): Likewise.
18048         (aarch64_<sur>adalp<mode>_3): Likewise.
18049         (<sur>sadv16qi): New define_expand.
18051 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18053         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18054         (*movdf_internal): Ditto.
18055         (*rcpsf2_sse): Ditto.
18056         (*rsqrtsf2_sse): Ditto.
18057         (*sqrt<mode>2_sse): Ditto.
18059 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18061         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18062         eor3q<mode>4.
18063         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18064         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18065         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18066         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18067         vbcaxq_s64): New.
18068         * config/aarch64/arm_neon.h: Likewise.
18069         * config/aarch64/iterators.md (VQ_I): New.
18071 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18073         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18074         arc*-*-linux*.
18075         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18076         MULTILIB_DIRNAMES
18078 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18080         * config/nds32/constraints.md (S): New constraint.
18081         * config/nds32/nds32.md (call_internal): Use constraint S.
18082         (call_value_internal): Likewise.
18083         (sibcall_internal): Likewise.
18084         (sibcall_value_internal): Likewise.
18086 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18087             Chung-Ju Wu  <jasonwucj@gmail.com>
18089         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18090         into consideration.
18092 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18093             Chung-Ju Wu  <jasonwucj@gmail.com>
18095         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18096         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18097         (nds32_rtx_costs_impl): Simplify.
18098         (nds32_address_cost_impl): Simplify.
18099         (nds32_init_rtx_costs): New function.
18100         (nds32_rtx_costs_speed_prefer): Likewise.
18101         (nds32_rtx_costs_size_prefer): Likewise.
18102         (nds32_address_cost_speed_prefer): Likewise.
18103         (nds32_address_cost_speed_fwprop): Likewise.
18104         (nds32_address_cost_size_prefer): Likewise.
18105         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18106         * config/nds32/nds32.c (nds32_option_override): Use
18107         nds32_init_rtx_costs function.
18109 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18111         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18112         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18113         (TARGET_PIPELINE_N8): Likewise.
18114         (TARGET_PIPELINE_N10): Likewise.
18115         (TARGET_PIPELINE_N13): Likewise.
18116         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18118 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18120         * config/nds32/nds32-fpu.md: Update copyright year.
18122 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18124         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18126 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18128         * config/nds32/nds32.c
18129         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18130         * config/nds32/nds32.opt (minline-asm-r15): New option.
18132 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18134         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18135         MASK_HW_ABS.
18136         * config/nds32/nds32.md (abssi2): New pattern.
18138 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
18140         * config/i386/i386.md (rex64namesuffix): New mode attribute.
18141         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18142         Merge insn pattern from sse_cvtsi2ss<round_name> and
18143         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18144         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18145         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18146         using SWI48 mode iterator.
18147         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18148         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18149         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18150         pattern from sse_cvttss2si<round_saeonly_name>
18151         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18152         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18153         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18154         using SWI48 mode iterator.
18155         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18156         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18157         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18158         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18159         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18160         using SWI48 mode iterator.
18161         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18162         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18163         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18164         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18165         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18166         SWI48 mode iterator.
18167         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18168         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18169         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18170         pattern from sse_cvttsd2si<round_saeonly_name>
18171         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18173 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18175         * config/nds32/nds32-md-auxiliary.c
18176         (nds32_valid_smw_lwm_base_p): Refine.
18177         (nds32_output_smw_single_word): Refine.
18178         (nds32_output_smw_double_word): New.
18179         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18181 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18183         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18184         (nds32_output_stack_pop): Refine.
18185         (nds32_expand_unaligned_load): Refine.
18186         (nds32_expand_unaligned_store): Refine.
18188 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18189             Chung-Ju Wu  <jasonwucj@gmail.com>
18191         * config/nds32/constants.md: Add TP_REGNUM constant.
18192         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18193         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18194         UNSPEC_ADD32.
18195         * config/nds32/nds32-doubleword.md: Consider flag_pic.
18196         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18197         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18198         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18199         and PIC code generation.
18200         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18201         code generation.
18202         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18203         optimization.
18204         * config/nds32/nds32.md: Support TLS and PIC.
18205         * config/nds32/nds32.c: Support TLS and PIC.
18206         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18207         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18208         predicate.
18210 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18212         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18213         mode with E_ prefix.
18215 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18216             Chung-Ju Wu  <jasonwucj@gmail.com>
18218         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18219         * config/nds32/nds32-md-auxiliary.c
18220         (symbolic_reference_mentioned_p): New.
18221         (nds32_legitimize_ict_address): New.
18222         (nds32_expand_ict_move): New.
18223         (nds32_indirect_call_referenced_p): New.
18224         (nds32_symbol_binds_local_p): Delete.
18225         (nds32_long_call_p): Modify.
18226         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18227         * config/nds32/nds32-protos.h
18228         (symbolic_reference_mentioned_p): Declare.
18229         (nds32_legitimize_ict_address): Declare.
18230         (nds32_expand_ict_move): Declare.
18231         (nds32_indirect_call_referenced_p): Declare.
18232         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18233         (nds32_relax_group): Use nds32_ict_const_p as condition.
18234         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18235         (nds32_asm_file_start): Output ict_model directive in asm code.
18236         (nds32_legitimate_address_p): Consider indirect call.
18237         (nds32_print_operand): Consider indirect call.
18238         (nds32_print_operand_address): Consider indirect call.
18239         (nds32_insert_attributes): Handle "indirect_call" attribute.
18240         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18241         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18242         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18243         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18244         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18245         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18246         (TARGET_ICT_MODEL_SMALL): Define.
18247         (TARGET_ICT_MODEL_LARGE): Define.
18248         * config/nds32/nds32.md (movsi): Consider ict model.
18249         (call, call_value): Consider ict model.
18250         (sibcall, sibcall_value): Consider ict model.
18251         * config/nds32/nds32.opt (mict-model): New option.
18252         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18253         model.
18255 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18256             Monk Chiang  <sh.chiang04@gmail.com>
18257             Jim Wilson <jimw@sifive.com>
18259         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18260         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18261         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18262         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18263         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
18264         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18265         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18266         compute save_libcall_adjustment properly.
18267         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18268         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18269         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18270         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18271         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18272         (ABI_SPEC): Handle mabi=ilp32e.
18273         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18274         (RVE): Add RVE mask.
18275         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18276         <-march>: Add rv32e as an example.
18278 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
18280         PR c++/82899
18281         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18282         (intra_create_variable_infos): Handle C++ constructors.
18284 2018-05-18  Martin Liska  <mliska@suse.cz>
18286         * passes.def: Remove a redundant pass.
18288 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
18290         PR bootstrap/85838
18291         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18293 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18295         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18296         (ARMv4): Update.
18297         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18298         (ARMv6m): Update.
18299         (armv2, armv2a, armv3, armv3m): Delete architectures.
18300         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18301         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18302         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18303         Delete cpus.
18304         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18305         (*mulsidi3adddi): Likewise.
18306         (mulsidi3): Likewise.
18307         (*mulsidi3_nov6): Likewise.
18308         (umulsidi3): Likewise.
18309         (umulsidi3_nov6): Likewise.
18310         (umaddsidi4): Likewise.
18311         (*umulsidi3adddi): Likewise.
18312         (smulsi3_highpart): Likewise.
18313         (*smulsi3_highpart_nov6): Likewise.
18314         (umulsi3_highpart): Likewise.
18315         (*umulsi3_highpart_nov6): Likewise.
18316         * config/arm/arm.h (arm_arch3m): Delete.
18317         * config/arm/arm.c (arm_arch3m): Delete.
18318         (arm_option_override_internal): Update armv3-related comment.
18319         (arm_configure_build_target): Delete use of isa_bit_mode32.
18320         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18321         (arm_rtx_costs_internal): Delete check of arm_arch3m.
18322         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18323         (mulsa3): Likewise.
18324         (mulusa3): Likewise.
18325         * config/arm/arm-protos.h (arm_arch3m): Delete.
18326         * config/arm/arm-tables.opt: Regenerate.
18327         * config/arm/arm-tune.md: Likewise.
18328         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18329         deleted architectures.
18331 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18333         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18334         (armv5t, armv5te): New features.
18335         (ARMv5, ARMv5e): Delete fgroups.
18336         (ARMv5t, ARMv5te): Adjust for above changes.
18337         (ARMv6m): Likewise.
18338         (armv5, armv5e): Delete arches.
18339         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18340         arm_arch5.
18341         (*call_reg_arm): Likewise.
18342         (*call_value_reg_armv5): Likewise.
18343         (*call_value_reg_arm): Likewise.
18344         (*call_symbol): Likewise.
18345         (*call_value_symbol): Likewise.
18346         (*sibcall_insn): Likewise.
18347         (*sibcall_value_insn): Likewise.
18348         (clzsi2): Likewise.
18349         (prefetch): Likewise.
18350         (define_split and define_peephole2 dependent on arm_arch5):
18351         Likewise.
18352         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18353         arm_arch5e.
18354         (TARGET_ARM_QBIT): Likewise.
18355         (TARGET_DSP_MULTIPLY): Likewise.
18356         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18357         (arm_arch5, arm_arch5e): Delete.
18358         (arm_arch5t, arm_arch5te): Declare.
18359         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18360         (arm_arch5t): Declare.
18361         (arm_option_reconfigure_globals): Update for the above.
18362         (arm_options_perform_arch_sanity_checks): Update comment, replace
18363         use of arm_arch5 with arm_arch5t.
18364         (use_return_insn): Likewise.
18365         (arm_emit_call_insn): Likewise.
18366         (output_return_instruction): Likewise.
18367         (arm_final_prescan_insn): Likewise.
18368         (arm_coproc_builtin_available): Likewise.
18369         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18370         arm_arch5e with arm_arch5t and arm_arch5te.
18371         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18372         (arm_arch5t, arm_arch5te): Declare.
18373         * config/arm/arm-tables.opt: Regenerate.
18374         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18375         * config/arm/t-multilib: Likewise.
18376         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18377         instead of arm_arch5.
18378         (*call_reg_thumb1): Likewise.
18379         (*call_value_reg_thumb1_v5): Likewise.
18380         (*call_value_reg_thumb1): Likewise.
18381         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18382         unreachable path.
18383         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18385 2018-05-18  Martin Liska  <mliska@suse.cz>
18387         PR gcov-profile/84846
18388         * doc/gcov.texi: Document -t option of gcov tool.
18390 2018-05-18  Martin Liska  <mliska@suse.cz>
18392         PR gcov-profile/84846
18393         * gcov.c (print_usage): Add new -t option.
18394         (process_args): Handle the option.
18395         (generate_results): Use stdout as output when requested by
18396         the option.
18398 2018-05-18  Martin Liska  <mliska@suse.cz>
18400         PR gcov-profile/84846
18401         * coverage.c (coverage_init): Write PWD to .gcno file.
18402         * doc/gcov.texi: Document how working directory is printed.
18403         * gcov-dump.c (dump_gcov_file): Print PWD.
18404         * gcov.c (output_intermediate_file): Likewise.
18405         (read_graph_file): Read PWD string.
18406         (output_lines): Print PWD.
18408 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18410         PR middle-end/85817
18411         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18412         for retval and return false if all args to phi are zero.
18414 2018-05-18  Richard Biener  <rguenther@suse.de>
18416         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18417         method.
18418         (evrp_dom_walker::before_dom_children): Call it.
18420 2018-05-18  Richard Biener  <rguenther@suse.de>
18422         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18423         results when processing array refs with variable index.
18425 2018-05-18  Toon Moene  <toon@moene.org>
18427         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18428         directly after that of -floop-interchange. Indicate that both
18429         options are enabled by default when specifying -O3.
18431 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18433         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18434         iterator.  Delete separate integer-mode vec_set<mode> expander.
18435         (aarch64_simd_vec_setv2di): Delete.
18436         (vec_setv2di): Delete.
18437         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18438         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
18439         the "w, r" alternative.
18441 2018-05-18  Martin Liska  <mliska@suse.cz>
18443         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18444         * tree-pass.h (make_pass_lower_switch_O0): New function.
18445         * tree-switch-conversion.c (node_has_low_bound): Remove.
18446         (node_has_high_bound): Likewise.
18447         (node_is_bounded): Likewise.
18448         (class pass_lower_switch): Make it a template type and create
18449         two instances.
18450         (pass_lower_switch::execute): Add template argument.
18451         (make_pass_lower_switch): New function.
18452         (make_pass_lower_switch_O0): New function.
18453         (do_jump_if_equal): Remove.
18454         (emit_case_nodes): Simplify to just handle all 3 cases and leave
18455         all the hard work to tree optimization passes.
18457 2018-05-18  Martin Liska  <mliska@suse.cz>
18459         * dbgcnt.c (limit_low): Renamed from limit.
18460         (limit_high): New variable.
18461         (dbg_cnt_is_enabled): Check for upper limit.
18462         (dbg_cnt): Adjust dumping.
18463         (dbg_cnt_set_limit_by_index): Add new argument for high
18464         value.
18465         (dbg_cnt_set_limit_by_name): Likewise.
18466         (dbg_cnt_process_single_pair): Parse new format.
18467         (dbg_cnt_process_opt): Use strtok.
18468         (dbg_cnt_list_all_counters): Remove 'value' and add
18469         'limit_high'.
18470         * doc/invoke.texi: Document changes.
18472 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
18474         * doc/sourcebuild.texi (scalar_all_fma): Document.
18475         * tree.def (FMA_EXPR): Delete.
18476         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18477         * internal-fn.c (ternary_direct): New macro.
18478         (expand_ternary_optab_fn): Likewise.
18479         (direct_ternary_optab_supported_p): Likewise.
18480         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18481         * builtins.c (fold_builtin_fma): Delete.
18482         (fold_builtin_3): Don't call it.
18483         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18484         * expr.c (expand_expr_real_2): Likewise.
18485         * fold-const.c (operand_equal_p): Likewise.
18486         (fold_ternary_loc): Likewise.
18487         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18488         * gimple.c (DEFTREECODE): Likewise.
18489         * gimplify.c (gimplify_expr): Likewise.
18490         * optabs-tree.c (optab_for_tree_code): Likewise.
18491         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18492         * tree-eh.c (operation_could_trap_p): Likewise.
18493         (stmt_could_throw_1_p): Likewise.
18494         * tree-inline.c (estimate_operator_cost): Likewise.
18495         * tree-pretty-print.c (dump_generic_node): Likewise.
18496         (op_code_prio): Likewise.
18497         * tree-ssa-loop-im.c (stmt_cost): Likewise.
18498         * tree-ssa-operands.c (get_expr_operands): Likewise.
18499         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18500         * fold-const-call.h (fold_fma): Delete.
18501         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18502         CFN_FNMA and CFN_FNMS.
18503         (fold_fma): Delete.
18504         * genmatch.c (combined_fn): New enum.
18505         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18506         (commutative_op): New function.
18507         (commutate): Use it.  Handle more than 2 operands.
18508         (dt_operand::gen_gimple_expr): Use commutative_op.
18509         (parser::parse_expr): Allow :c to be used with non-binary
18510         operators if the commutative operand is known.
18511         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18512         CFN_FMS, CFN_FNMA and CFN_FNMS.
18513         (backprop::process_assign_use): Remove FMA_EXPR handling.
18514         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18515         (gen_hsa_fma): New function.
18516         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18517         IFN_FNMA and IFN_FNMS.
18518         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18519         * gimple-fold.h (follow_all_ssa_edges): Declare.
18520         * gimple-fold.c (follow_all_ssa_edges): New function.
18521         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18522         gimple_build interface and use follow_all_ssa_edges to fold the result.
18523         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18524         instead of checking for optabs directly.
18525         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
18526         rather than FMA_EXPRs.
18527         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
18528         call to IFN_FMA instead of an FMA_EXPR.
18530 2018-05-17  Jim Wilson  <jimw@sifive.com>
18532         * expr.c (do_tablejump): When converting index to Pmode, if we have a
18533         sign extended promoted subreg, and the range does not have the sign bit
18534         set, then do a sign extend.
18536         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
18537         test, check for sign extended subreg and/or constant operands, and
18538         do a sign extend in that case.
18540 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
18542         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
18543         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
18544         Add untyped.
18545         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
18546         Change logics_shift_reg to logics_shift_imm.
18547         (thunderx2t99_fp_loadpair_basic): Delete.
18548         (thunderx2t99_fp_storepair_basic): Delete.
18549         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
18550         (thunderx2t99_asimd_polynomial): Delete.
18551         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
18552         and neon_fp_mul_d_scalar_q.
18553         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
18554         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
18555         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
18556         (thunderx2t99_asimd_lut): Add missing tbl types.
18557         (thunderx2t99_asimd_ext): Delete.
18558         (thunderx2t99_asimd_load1_1_mult): Delete.
18559         (thunderx2t99_asimd_load1_2_mult): Delete.
18560         (thunderx2t99_asimd_load1_ldp): New.
18561         (thunderx2t99_asimd_load1): New.
18562         (thunderx2t99_asimd_load2): Add missing *load2* types.
18563         (thunderx2t99_asimd_load3): New.
18564         (thunderx2t99_asimd_load4): New.
18565         (thunderx2t99_asimd_store1_1_mult): Delete.
18566         (thunderx2t99_asimd_store1_2_mult): Delete.
18567         (thunderx2t99_asimd_store2_mult): Delete.
18568         (thunderx2t99_asimd_store2_onelane): Delete.
18569         (thunderx2t99_asimd_store_stp): New.
18570         (thunderx2t99_asimd_store1): New.
18571         (thunderx2t99_asimd_store2): New.
18572         (thunderx2t99_asimd_store3): New.
18573         (thunderx2t99_asimd_store4): New.
18575 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
18577         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
18578         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
18580 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
18581             Segher Boessenkool  <segher@kernel.crashing.org>
18583         PR target/85698
18584         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
18585         operand.
18587 2018-05-17  Richard Biener  <rguenther@suse.de>
18589         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
18590         for pruning loop and prune defs feeding only already visited PHIs.
18592 2018-05-17  Richard Biener  <rguenther@suse.de>
18594         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
18596 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
18597             Richard Biener  <rguenther@suse.de>
18599         PR tree-optimization/85793
18600         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
18601         for VMAT_ELEMENTWISE.
18603 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18605         * internal-fn.h (lookup_internal_fn): Declare
18606         * internal-fn.c (lookup_internal_fn): New function.
18607         * gimple.c (gimple_build_call_from_tree): Handle calls to
18608         internal functions.
18609         * gimple-pretty-print.c (dump_gimple_call): Print "." before
18610         internal function names.
18611         * tree-pretty-print.c (dump_generic_node): Likewise.
18612         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
18614 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18616         * gimple-fold.h (gimple_build): Make the function forms take
18617         combined_fn rather than built_in_function.
18618         (gimple_simplify): Likewise.
18619         * gimple-match-head.c (gimple_simplify): Likewise.
18620         * gimple-fold.c (gimple_build): Likewise.
18621         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
18622         rather than gimple_build_call_internal.
18623         (get_initial_defs_for_reduction): Likewise.
18624         (vect_create_epilog_for_reduction): Likewise.
18625         (vectorizable_live_operation): Likewise.
18627 2018-05-17  Martin Liska  <mliska@suse.cz>
18629         * gimple-ssa-sprintf.c (format_directive): Do not use
18630         space in between 'G_' and '('.
18632 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
18634         PR target/85323
18635         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
18636         even if the mask is not all ones.
18638         PR target/85323
18639         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
18640         vector.
18641         (ix86_gimple_fold_builtin): Likewise.
18643         PR target/85323
18644         * config/i386/i386.c: Include tree-vector-builder.h.
18645         (ix86_vector_shift_count): New function.
18646         (ix86_fold_builtin): Fold shift builtins by scalar count.
18647         (ix86_gimple_fold_builtin): Likewise.
18649         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
18650         _mm512_setzero): New intrinsics.
18652 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
18653             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18655         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
18656         code generation for cases where splatting a value is not useful.
18657         * simplify-rtx.c (simplify_ternary_operation): Simplify
18658         vec_merge across a vec_duplicate and a paradoxical subreg forming
18659         a vector mode to a vec_concat.
18661 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
18663         * config.gcc: Support "goldmont-plus".
18664         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18665         "goldmont-plus".
18666         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18667         PROCESSOR_GOLDMONT_PLUS.
18668         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
18669         (processor_target_table): Add "goldmont-plus".
18670         (PTA_GOLDMONT_PLUS): Define.
18671         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
18672         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
18673         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
18674         (fold_builtin_cpu): Add "goldmont-plus".
18675         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
18676         (ix86_option_override_internal): Add "goldmont-plus".
18677         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
18678         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
18679         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
18680         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
18682 2018-05-17  Richard Biener  <rguenther@suse.de>
18684         PR tree-optimization/85757
18685         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
18686         remove defs that only feed that PHI from further processing.
18688 2018-05-16  Jim Wilson  <jimw@sifive.com>
18690         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
18691         asterisk to name.
18692         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
18694 2018-05-16  Mark Wielaard  <mark@klomp.org>
18696         * dwarf2out.c (count_index_strings): New function.
18697         (output_indirect_strings): Call count_index_strings and generate
18698         header for dwarf_version >= 5.
18700 2018-05-16  Mark Wielaard  <mark@klomp.org>
18702         * dwarf2out.c (dwarf_FORM): New function.
18703         (set_indirect_string): Use dwarf_FORM.
18704         (reset_indirect_string): Likewise.
18705         (size_of_die): Likewise.
18706         (value_format): Likewise.
18707         (output_die): Likewise.
18708         (add_skeleton_AT_string): Likewise.
18709         (output_macinfo_op): Likewise.
18710         (index_string): Likewise.
18711         (output_index_string_offset): Likewise.
18712         (output_index_string): Likewise.
18713         (count_index_strings): Likewise.
18715 2018-05-16  Carl Love  <cel@us.ibm.com>
18717         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18718         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
18720 2018-05-16  Martin Jambor  <mjambor@suse.cz>
18722         * ipa-prop.c (ipa_free_all_edge_args): Remove.
18723         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
18725 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
18727         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
18728         (fnma<mode>4): Likewise.
18729         (fms<mode>4): Likewise.
18730         (fnms<mode>4): Likewise.
18731         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
18732         (aarch64_fnma<mode>4): Likewise.
18733         (aarch64_fms<mode>4): Likewise.
18734         (aarch64_fnms<mode>4): Likewise.
18735         (aarch64_fnmadd<mode>4): Likewise.
18737 2018-05-16  Jason Merrill  <jason@redhat.com>
18739         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
18741 2018-05-16  Richard Biener  <rguenther@suse.de>
18743         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
18744         (dump_stmt_cost): Declare.
18745         (add_stmt_cost): Dump cost we add.
18746         (add_stmt_costs): New function.
18747         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
18748         No longer exported.
18749         (vect_analyze_stmt): Adjust prototype.
18750         (vectorizable_condition): Likewise.
18751         (vectorizable_live_operation): Likewise.
18752         (vectorizable_reduction): Likewise.
18753         (vectorizable_induction): Likewise.
18754         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
18755         cost vector to pass to vectorizable_ and record afterwards.
18756         (vect_model_reduction_cost): Take cost vector argument and adjust.
18757         (vect_model_induction_cost): Likewise.
18758         (vectorizable_reduction): Likewise.
18759         (vectorizable_induction): Likewise.
18760         (vectorizable_live_operation): Likewise.
18761         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
18762         SLP_TREE_NUMBER_OF_VEC_STMTS.
18763         (vect_analyze_slp_cost_1): Remove.
18764         (vect_analyze_slp_cost): Likewise.
18765         (vect_slp_analyze_node_operations): Take visited args and
18766         a target cost vector.  Avoid processing already visited stmt sets.
18767         (vect_slp_analyze_operations): Use a local cost vector to gather
18768         costs and register those of non-discarded instances.
18769         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
18770         (vect_schedule_slp_instance): Remove copying of
18771         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
18772         zero.
18773         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
18774         adding cost.  Record cost entry location.
18775         (vect_prologue_cost_for_slp_op): Function to compute cost of
18776         a constant or invariant generated for SLP vect in the prologue,
18777         split out from vect_analyze_slp_cost_1.
18778         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
18779         (vect_model_promotion_demotion_cost): Likewise.
18780         (vect_model_store_cost): Likewise, make static.
18781         (vect_model_load_cost): Likewise.
18782         (vectorizable_bswap): Add cost vector arg and adjust.
18783         (vectorizable_call): Likewise.
18784         (vectorizable_simd_clone_call): Likewise.
18785         (vectorizable_conversion): Likewise.
18786         (vectorizable_assignment): Likewise.
18787         (vectorizable_shift): Likewise.
18788         (vectorizable_operation): Likewise.
18789         (vectorizable_store): Likewise.
18790         (vectorizable_load): Likewise.
18791         (vectorizable_condition): Likewise.
18792         (vectorizable_comparison): Likewise.
18793         (can_vectorize_live_stmts): Likewise.
18794         (vect_analyze_stmt): Likewise.
18795         (vect_transform_stmt): Adjust calls to vectorizable_*.
18796         * tree-vectorizer.c: Include gimple-pretty-print.h.
18797         (dump_stmt_cost): New function.
18799 2018-05-16  Richard Biener  <rguenther@suse.de>
18801         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
18802         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
18803         * tree-ssa-dse.c: Include tree-ssa-loop.h.
18804         (check_name): New callback.
18805         (dse_classify_store): Track cycles via a visited bitmap of PHI
18806         defs and simplify handling of in-loop and across loop dead stores
18807         and properly fail for loop-variant refs.  Handle byte-tracking with
18808         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
18809         limiting the walk.
18811 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
18813         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
18814         (vect_get_mask_type_for_stmt): Likewise.
18815         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
18816         split out from...
18817         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
18818         to determine the statement's vector type and the vector type that
18819         should be used for calculating nunits.  Deal with cases in which
18820         the type has to be deferred.
18821         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
18822         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
18823         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
18824         (vect_determine_vf_for_stmt): New functions, split out from...
18825         (vect_determine_vectorization_factor): ...here.
18826         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
18827         (vect_get_mask_type_for_stmt): New functions, split out from
18828         vect_determine_vectorization_factor.
18830 2018-05-16  Richard Biener  <rguenther@suse.de>
18832         * tree-cfg.c (verify_gimple_assign_ternary): Properly
18833         verify the [VEC_]COND_EXPR embedded comparison.
18835 2018-05-15  Martin Sebor  <msebor@redhat.com>
18837         PR tree-optimization/85753
18838         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
18839         RECORD_TYPE in addition to ARRAY_TYPE.
18841 2018-05-15  Martin Sebor  <msebor@redhat.com>
18843         PR middle-end/85643
18844         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
18846 2018-05-15  Richard Biener  <rguenther@suse.de>
18848         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
18849         add by_clobber_p one.  Change algorithm to collect all defs
18850         representing uses we need to walk and try reducing them to
18851         a single one before failing.
18852         (dse_dom_walker::dse_optimize_stmt): Adjust.
18854 2018-05-13  Mark Wielaard  <mark@klomp.org>
18856         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
18857         (size_of_loc_descr): Likewise.
18858         (output_loc_operands): Likewise.
18859         (output_loc_operands_raw): Likewise.
18860         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
18861         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
18862         (hash_loc_operands): Likewise.
18863         (compare_loc_operands): Likewise.
18865 2018-05-14  Mark Wielaard  <mark@klomp.org>
18867         * dwarf2out.c (count_index_addrs): New function.
18868         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
18870 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18872         PR tree-optimization/83648
18873         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
18874         return value as malloc candidate.
18876 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18878         PR ipa/85734
18879         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
18880         param as true in call to suggest_attribute.
18882 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
18884         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
18885         -mreadonly-in-sdata.
18887 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18889         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
18890         New pattern.
18891         (aarch64_crypto_aesd_fused): Likewise.
18893 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
18895         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
18896         (movsi_aarch64): Likewise.
18897         (load_pairsi): Likewise.
18898         (load_pairdi): Likewise.
18899         (store_pairsi): Likewise.
18900         (store_pairdi): Likewise.
18901         (load_pairsf): Likewise.
18902         (load_pairdf): Likewise.
18903         (store_pairsf): Likewise.
18904         (store_pairdf): Likewise.
18905         (zero_extend): Likewise.
18906         (trunc): Swap alternatives.
18907         (fcvt_target): Add '?' to prefer w over r.
18909 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
18911         PR target/85756
18912         * config/i386/i386.md: Disallow non-commutative arithmetics in
18913         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
18914         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
18915         in the peephole2 before it.
18917 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
18919         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
18920         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
18921         (ix86_handle_option): Handle -mcldemote.
18922         * config.gcc: New header.
18923         * config/i386/cldemoteintrin.h: New file.
18924         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
18925         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18926         -mcldemote.
18927         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18928         OPTION_MASK_ISA_CLDEMOTE.
18929         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
18930         (ix86_valid_target_attribute_inner_p): Ditto.
18931         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
18932         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
18933         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
18934         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
18935         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
18936         (cldemote): New.
18937         * config/i386/i386.opt: Add -mcldemote.
18938         * config/i386/x86intrin.h: New header.
18939         * doc/invoke.texi: Add -mcldemote.
18941 2018-05-14  Richard Biener  <rguenther@suse.de>
18943         * doc/match-and-simplify.texi: Adjust :s documentation.
18945 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
18947         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
18948         intended memcpy size.
18949         (REORDER_45): Likewise.
18951 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
18953         * sort.cc: New file.
18954         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
18955         * vec.c (qsort_chk): Use gcc_qsort.
18956         * Makefile.in (OBJS-libcommon): Add sort.o.
18957         (build/sort.o): New target.  Use it...
18958         (BUILD_RTL): ... here, and...
18959         (build/gencfn-macros): ... here, and...
18960         (build/genmatch): ... here.
18962 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18963             Chung-Ju Wu  <jasonwucj@gmail.com>
18965         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
18966         * config/nds32/nds32-graywolf.md: New file.
18967         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
18968         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
18969         pipeline.
18970         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
18971         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
18972         * config/nds32/nds32.md (pipeline_model): Add graywolf.
18973         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
18974         * config/nds32/pipelines.md: Include n15 settings.
18976 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18977             Chung-Ju Wu  <jasonwucj@gmail.com>
18979         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
18980         * config/nds32/nds32-n13.md: New file.
18981         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
18982         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
18983         pipeline.
18984         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
18985         * config/nds32/nds32.md (pipeline_model): Add n13.
18986         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
18987         * config/nds32/pipelines.md: Include n13 settings.
18989 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18990             Chung-Ju Wu  <jasonwucj@gmail.com>
18992         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
18993         * config/nds32/nds32-n10.md: New file.
18994         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
18995         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
18996         pipeline.
18997         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
18998         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
18999         * config/nds32/nds32.md (pipeline_model): Add n10.
19000         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19001         * config/nds32/pipelines.md: Include n10 settings.
19003 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19004             Kito Cheng  <kito.cheng@gmail.com>
19005             Chung-Ju Wu  <jasonwucj@gmail.com>
19007         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19008         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19009         Add enum values for DSP extension instructions.
19010         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19011         New constraints.
19012         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19013         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19014         New code iterators.
19015         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19016         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19017         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19018         * config/nds32/nds32-intrinsic.md: Likewise.
19019         * config/nds32/nds32_intrinsic.h: Likewise.
19020         * config/nds32/nds32-md-auxiliary.c: Likewise.
19021         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19022         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19023         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19024         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19025         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19026         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19027         TYPE_DMAC in switch statement.
19028         * config/nds32/nds32.c: New checking and implementation for DSP
19029         extension instructions.
19030         * config/nds32/nds32.h: Likewise.
19031         * config/nds32/nds32.md: Likewise.
19032         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19033         * config/nds32/predicates.md: Implement new predicates for DSP
19034         extension.
19036 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19038         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19039         Reformat alternatives and attributes so it is easier to identify
19040         which constraints/attributes go with which instruction.
19041         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19042         (mov<mode>_softfloat32, FMOVE64): Likewise.
19043         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19044         (mov<mode>_softfloat64, FMOVE64): Likewise.
19046 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19048         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19049         subsection.
19050         (Basic PowerPC Built-in Functions): The new name of the
19051         subsection previously known as "PowerPC Built-in Functions".
19052         (Basic PowerPC Built-in Functions Available on all Configurations):
19053         New subsubsection.
19054         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19055         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19056         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19057         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19059 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19061         PR ipa/85655
19062         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19063         single const.
19065 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19067         PR target/85733
19068         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19070 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19072         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19073         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19074         (ix86_handle_option): Handle -mwaitpkg.
19075         * config.gcc: New header.
19076         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19077         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19078         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19079         function type.
19080         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19081         OPTION_MASK_ISA_WAITPKG.
19082         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19083         (ix86_option_override_internal): Add PTA_WAITPKG.
19084         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19085         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19086         IX86_BUILTIN_TPAUSE.
19087         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19088         __builtin_ia32_umwait and __builtin_ia32_tpause.
19089         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19090         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19091         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19092         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19093         UNSPECV_TPAUSE): New.
19094         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19095         * config/i386/i386.opt: Add -mwaitpkg.
19096         * config/i386/waitpkgintrin.h: New file.
19097         * config/i386/x86intrin.h: New header.
19098         * doc/invoke.texi: Add -mwaitpkg.
19100 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19102         PR target/85606
19103         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19104         equivalent.
19105         (cortex-m0): Use armv6s-m isa.
19106         (cortex-m0plus): Likewise.
19107         (cortex-m1): Likewise.
19108         (cortex-m0.small-multiply): Likewise.
19109         (cortex-m0plus.small-multiply): Likewise.
19110         (cortex-m1.small-multiply): Likewise.
19112 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19113             Jakub Jelinek  <jakub@redhat.com>
19115         PR tree-optimization/85692
19116         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19117         source permute as well.
19119 2018-05-11  Martin Liska  <mliska@suse.cz>
19121         PR sanitizer/85556
19122         * doc/extend.texi: Document LLVM style format for no_sanitize
19123         attribute.
19125 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19127         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19128         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19129         (mode_supports_vsx_dform_quad): Likewise.
19130         (mode_supports_vmx_dform): Move these functions to be next to the
19131         other mode_supports functions.
19132         (mode_supports_dq_form): Likewise.
19133         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19134         mode_supports_dq_form.
19135         (reg_offset_addressing_ok_p): Likewise.
19136         (offsettable_ok_by_alignment): Likewise.
19137         (rs6000_legitimate_offset_address_p): Likewise.
19138         (legitimate_lo_sum_address_p): Likewise.
19139         (rs6000_legitimize_address): Likewise.
19140         (rs6000_legitimize_reload_address): Likewise.
19141         (rs6000_secondary_reload_inner): Likewise.
19142         (rs6000_preferred_reload_class): Likewise.
19143         (rs6000_output_move_128bit): Likewise.
19145 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19147         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19148         Generate SImode target register for null target.
19149         <case IX86_BUILTIN_XGETBV>: Ditto.
19150         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19151         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19153 2018-05-10  Carl Love  <cel@us.ibm.com>
19155         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19156         dcbtt and dcbtstt if operands[2] is 0.
19158 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19160         PR target/85693
19161         * config/i386/sse.md (usadv64qi): New expander.
19163 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
19165         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19166         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19167         -maltivec=be support.
19168         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19169         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19170         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19171         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19172         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19173         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19174         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19175         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19176         altivec_vsumsws): Adjust.
19177         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19178         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19179         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19180         support.
19181         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19182         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19183         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19184         (altivec_lve<VI_char>x): Delete expand.
19185         (*altivec_lve<VI_char>x_internal): Rename to...
19186         (altivec_lve<VI_char>x): ... this.
19187         (altivec_lvxl_<mode>): Delete expand.
19188         (*altivec_lvxl_<mode>_internal): Rename to ...
19189         (altivec_lvxl_<mode>): ... this.
19190         (altivec_stvxl_<mode>): Delete expand.
19191         (*altivec_stvxl_<mode>_internal): Rename to ...
19192         (altivec_stvxl_<mode>): ... this.
19193         (altivec_stve<VI_char>x): Delete expand.
19194         (*altivec_stve<VI_char>x_internal): Rename to ...
19195         (altivec_stve<VI_char>x): ... this.
19196         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19197         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19198         reduc_plus_scal_<mode>): Adjust.
19199         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19200         comment.
19201         (rs6000_cpu_cpp_builtins): Adjust.
19202         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19203         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19204         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19205         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19206         -maltivec=be support.
19207         (rs6000_split_vec_extract_var): Adjust.
19208         (rs6000_split_v4si_init): Adjust.
19209         (swap_selector_for_mode): Delete.
19210         (altivec_expand_lvx_be, altivec_expand_stvx_be,
19211         altivec_expand_stvex_be): Delete.
19212         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19213         -maltivec=be support.
19214         (rs6000_gimple_fold_builtin): Ditto.
19215         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19216         Adjust.
19217         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19218         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19219         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19220         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19221         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19222         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19223         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19224         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19225         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19226         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19227         anonymous split): Adjust.
19228         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19229         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19231 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19233         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19234         when --with-gxx-include-dir is also specified.
19235         * configure: Regenerate.
19237 2018-05-09  Jim Wilson  <jimw@sifive.com>
19239         PR target/84797
19240         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19241         * config/riscv/t-withmultilib: New.
19242         * config/riscv/withmultilib.h: New.
19243         * doc/install.texi: Document RISC-V --with-multilib-list support.
19245 2018-05-09  Richard Biener  <rguenther@suse.de>
19247         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19248         vector.
19249         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19250         actual scalar cost using the cost vector and the add_stmt_cost
19251         machinery.
19253 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19255         PR rtl-optimization/85645
19256         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19257         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19259 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19261         PR rtl-optimization/85645
19262         * shrink-wrap.c (spread_components): Return a boolean saying if
19263         anything was changed.
19264         (try_shrink_wrapping_separate): Iterate spread_components until
19265         nothing changes anymore.
19267 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19269         PR rtl-optimization/85645
19270         * regrename.c (build_def_use): Also kill the chains that include the
19271         destination of a REG_CFA_REGISTER note.
19273 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19275         PR rtl-optimization/85645
19276         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19277         insn that has a REG_CFA_REGISTER note.
19279 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
19281         * cfgexpand.c (expand_clobber): New function.
19282         (expand_gimple_stmt_1): Use it.
19283         * tree-vect-stmts.c (vect_clobber_variable): New function,
19284         split out from...
19285         (vectorizable_simd_clone_call): ...here.
19286         (vectorizable_store): Emit a clobber either side of an
19287         IFN_STORE_LANES sequence.
19288         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19290 2018-05-09  Tom de Vries  <tom@codesourcery.com>
19292         PR target/85626
19293         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19294         (define_insn "trap_if_false"): Add exit after trap.
19296 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
19298         PR rtl-optimization/85638
19299         * bb-reorder.c: Include common/common-target.h.
19300         (create_forwarder_block): New function extracted from...
19301         (fix_up_crossing_landing_pad): ...here.  Rename into...
19302         (dw2_fix_up_crossing_landing_pad): ...this.
19303         (sjlj_fix_up_crossing_landing_pad): New function.
19304         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19305         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19306         from both partitions and exit the loop after one iteration.
19308 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19310         Revert:
19311         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19312         subsection.
19313         (Basic PowerPC Built-in Functions): The new name of the
19314         subsection previously known as "PowerPC Built-in Functions".
19315         (Basic PowerPC Built-in Functions Available on all Configurations):
19316         New subsubsection.
19317         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19318         subsubsection.
19319         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19320         subsubsection.
19321         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19322         subsubsection.
19323         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19324         subsubsection.
19326 2018-05-08  Jim Wilson  <jimw@sifive.com>
19328         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19329         (LD_EMUL_SUFFIX): New.
19330         (LINK_SPEC): Use it.
19332 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19334         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19335         subsection.
19336         (Basic PowerPC Built-in Functions): The new name of the
19337         subsection previously known as "PowerPC Built-in Functions".
19338         (Basic PowerPC Built-in Functions Available on all Configurations):
19339         New subsubsection.
19340         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19341         subsubsection.
19342         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19343         subsubsection.
19344         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19345         subsubsection.
19346         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19347         subsubsection.
19349 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19351         PR target/85683
19352         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19353         after cmpelim optimization.
19355 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
19357         * config.gcc: Support "goldmont".
19358         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19359         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19360         PROCESSOR_GOLDMONT.
19361         * config/i386/i386.c (m_GOLDMONT): Define.
19362         (processor_target_table): Add "goldmont".
19363         (PTA_GOLDMONT): Define.
19364         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19365         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19366         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19367         (fold_builtin_cpu): Add "goldmont".
19368         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19369         (ix86_option_override_internal): Add "goldmont".
19370         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19371         (processor_type): Add PROCESSOR_GOLDMONT.
19372         * config/i386/i386.md: Add CPU "glm".
19373         * config/i386/glm.md: New file.
19374         * config/i386/x86-tune.def: Add m_GOLDMONT.
19375         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19377 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19379         PR target/85572
19380         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19381         E_V4DImode.
19382         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19383         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
19384         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
19386         PR target/85317
19387         * config/i386/i386.c (ix86_fold_builtin): Handle
19388         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19390         PR target/85480
19391         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19392         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19394 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
19396         PR target/85658
19397         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19398         (check_arch): Likewise.
19399         (check_fpu): Return the result rather than printing it.
19400         (end arch): Fix operator precedence.
19401         (end cpu): Likewise.
19402         (END): Print the result from check_fpu.
19404 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19405             Alan Hayward  <alan.hayward@arm.com>
19406             David Sherwood  <david.sherwood@arm.com>
19408         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19409         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19410         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19411         (*fcmuo<mode>_and): New patterns.
19413 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19415         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19416         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19417         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19418         (cmp_op, sve_imm_con): New code attributes.
19419         (SVE_COND_INT_CMP, imm_con): Delete.
19420         (cmp_op): Remove above unspecs from int attribute.
19421         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19422         to...
19423         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
19424         comparison-specific unspecs.
19425         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19426         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19427         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19428         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19429         (*vec_fcm<cmp_op><mode>): Rename to...
19430         (*fcm<cmp_op><mode>): ...this and adjust likewise.
19431         (*vec_fcmuo<mode>): Rename to...
19432         (*fcmuo<mode>): ...this and adjust likewise.
19433         (*pred_fcm<cmp_op><mode>): New pattern.
19434         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19435         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19436         functions.
19437         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19438         and UNORDERED.
19439         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19440         (aarch64_emit_sve_predicated_cond): New function.
19441         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19442         (aarch64_emit_unspec_cond_or): Replace with...
19443         (aarch64_emit_sve_or_conds): ...this new function.  Use
19444         aarch64_emit_sve_ptrue_op for the individual comparisons and
19445         aarch64_emit_binop to OR them together.
19446         (aarch64_emit_inverted_unspec_cond): Replace with...
19447         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
19448         aarch64_emit_sve_ptrue_op for the comparison and
19449         aarch64_emit_unop to invert the result.
19450         (aarch64_expand_sve_vec_cmp_float): Update after the above
19451         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
19453 2018-05-07  Nathan Sidwell  <nathan@acm.org>
19455         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19456         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19457         (Backwards Compatibility): Likewise.
19459 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19461         PR bootstrap/85681
19462         Revert:
19463         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19465         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19466         <prefetch_dynamic_strides>: New const bool field.
19467         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19468         prefetch_dynamic_strides.
19469         (exynosm1_prefetch_tune): Likewise.
19470         (thunderxt88_prefetch_tune): Likewise.
19471         (thunderx_prefetch_tune): Likewise.
19472         (thunderx2t99_prefetch_tune): Likewise.
19473         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19474         to false.
19475         (aarch64_override_options_internal): Update to set
19476         PARAM_PREFETCH_DYNAMIC_STRIDES.
19477         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19478         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19479         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19480         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19481         prefetch-dynamic-strides setting.
19483         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19485         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19486         <minimum_stride>: New const int field.
19487         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19488         minimum_stride field.
19489         (exynosm1_prefetch_tune): Likewise.
19490         (thunderxt88_prefetch_tune): Likewise.
19491         (thunderx_prefetch_tune): Likewise.
19492         (thunderx2t99_prefetch_tune): Likewise.
19493         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19494         (aarch64_override_options_internal): Update to set
19495         PARAM_PREFETCH_MINIMUM_STRIDE.
19496         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19497         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19498         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19499         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19500         stride is constant and is below the minimum stride threshold.
19502 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19504         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19505         to 512.
19507 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19509         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19510         <prefetch_dynamic_strides>: New const bool field.
19511         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19512         prefetch_dynamic_strides.
19513         (exynosm1_prefetch_tune): Likewise.
19514         (thunderxt88_prefetch_tune): Likewise.
19515         (thunderx_prefetch_tune): Likewise.
19516         (thunderx2t99_prefetch_tune): Likewise.
19517         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19518         to false.
19519         (aarch64_override_options_internal): Update to set
19520         PARAM_PREFETCH_DYNAMIC_STRIDES.
19521         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19522         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19523         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19524         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19525         prefetch-dynamic-strides setting.
19527 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19529         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19530         <minimum_stride>: New const int field.
19531         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19532         minimum_stride field.
19533         (exynosm1_prefetch_tune): Likewise.
19534         (thunderxt88_prefetch_tune): Likewise.
19535         (thunderx_prefetch_tune): Likewise.
19536         (thunderx2t99_prefetch_tune): Likewise.
19537         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19538         (aarch64_override_options_internal): Update to set
19539         PARAM_PREFETCH_MINIMUM_STRIDE.
19540         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19541         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19542         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19543         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19544         stride is constant and is below the minimum stride threshold.
19546 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
19548         PR c++/85659
19549         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
19550         the type is addressable.  Don't force op into register if it has
19551         BLKmode.
19553 2018-05-05  Roland McGrath  <mcgrathr@google.com>
19555         PR other/77609
19556         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
19557         any section for which we don't know a specific type it should have,
19558         regardless of name.  Previously this was done only for the exact
19559         names ".init_array", ".fini_array", and ".preinit_array".
19560         (default_elf_asm_named_section): Add comment about
19561         relationship with default_section_type_flags and SECTION_NOTYPE.
19562         (get_section): Don't consider it a type conflict if one side has
19563         SECTION_NOTYPE and the other doesn't, as long as neither has the
19564         SECTION_BSS et al used in the default_section_type_flags logic.
19566 2018-05-05  Tom de Vries  <tom@codesourcery.com>
19568         PR target/85653
19569         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
19570         (workaround_barsyncs): New function.
19571         (nvptx_reorg): Use workaround_barsyncs.
19572         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19573         (define_expand "nvptx_membar_cta"): New define_expand.
19574         (define_insn "*nvptx_membar_cta"): New insn.
19576 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
19578         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
19579         To improve optimization opportunities.
19580         * builtin-types.def: The new needed builtin types for the above.
19582 2018-05-04  Richard Biener  <rguenther@suse.de>
19584         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
19585         * gimple-ssa-store-merging.c
19586         (imm_store_chain_info::output_merged_store): Remove redundant create,
19587         release split_store vector contents on failure.
19588         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
19589         scalar stmt vector on cache hit.
19591 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
19593         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19594         Xilinx FP support.
19595         * config.gcc (powerpc-xilinx-eabi*): Remove.
19596         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
19597         support.
19598         (fusion_addis_mem_combo_load): Ditto.
19599         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
19600         FP support.
19601         (rs6000_cpu_cpp_builtins): Ditto.
19602         * config/rs6000/rs6000-linux.c
19603         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
19604         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
19605         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
19606         support.
19607         (rs6000_setup_reg_addr_masks): Ditto.
19608         (rs6000_init_hard_regno_mode_ok): Ditto.
19609         (rs6000_option_override_internal): Ditto.
19610         (legitimate_lo_sum_address_p): Ditto.
19611         (rs6000_legitimize_address): Ditto.
19612         (rs6000_legitimize_reload_address): Ditto.
19613         (rs6000_legitimate_address_p): Ditto.
19614         (abi_v4_pass_in_fpr): Ditto.
19615         (setup_incoming_varargs): Ditto.
19616         (rs6000_gimplify_va_arg): Ditto.
19617         (rs6000_split_multireg_move): Ditto.
19618         (rs6000_savres_strategy): Ditto.
19619         (rs6000_emit_prologue_components): Ditto.
19620         (rs6000_emit_epilogue_components): Ditto.
19621         (rs6000_emit_prologue): Ditto.
19622         (rs6000_emit_epilogue): Ditto.
19623         (rs6000_elf_file_end): Ditto.
19624         (rs6000_function_value): Ditto.
19625         (rs6000_libcall_value): Ditto.
19626         * config/rs6000/rs6000.h: Ditto.
19627         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
19628         (TARGET_MINMAX): ... this.  New.
19629         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
19630         * config/rs6000/rs6000.md: Remove Xilinx FP support.
19631         (*movsi_internal1_single): Delete.
19632         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
19633         mfpu=, mxilinx-fpu): Delete.
19634         * config/rs6000/singlefp.h: Delete.
19635         * config/rs6000/sysv4.h: Remove Xilinx FP support.
19636         * config/rs6000/t-rs6000: Ditto.
19637         * config/rs6000/t-xilinx: Delete.
19638         * config/rs6000/titan.md: Adjust for fp_type removal.
19639         * config/rs6000/vsx.md: Remove Xilinx FP support.
19640         (VStype_simple): Delete.
19641         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
19642         * config/rs6000/xfpu.h: Delete.
19643         * config/rs6000/xfpu.md: Delete.
19644         * config/rs6000/xilinx.h: Delete.
19645         * config/rs6000/xilinx.opt: Delete.
19646         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
19647         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
19649 2018-05-04  Tom de Vries  <tom@codesourcery.com>
19651         PR libgomp/85639
19652         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
19653         if ignore == 0.
19655 2018-05-04  Richard Biener  <rguenther@suse.de>
19657         PR middle-end/85627
19658         * tree-complex.c (update_complex_assignment): We are always in SSA form.
19659         (expand_complex_div_wide): Likewise.
19660         (expand_complex_operations_1): Likewise.
19661         (expand_complex_libcall): Preserve EH info of the original stmt.
19662         (tree_lower_complex): Handle removed blocks.
19663         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
19664         on complex multiplication and division libcall builtins.
19666 2018-05-04  Richard Biener  <rguenther@suse.de>
19668         PR middle-end/85574
19669         * fold-const.c (negate_expr_p): Restrict negation of operand
19670         zero of a division to when we know that can happen without
19671         overflow.
19672         (fold_negate_expr_1): Likewise.
19674 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
19676         PR libstdc++/85466
19677         * real.h (real_nextafter): Declare.
19678         * real.c (real_nextafter): New function.
19679         * fold-const-call.c (fold_const_nextafter): New function.
19680         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
19681         CASE_CFN_NEXTTOWARD.
19682         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
19683         even when arg1_mode is different from arg0_mode.
19685 2018-05-03  Nathan Sidwell  <nathan@acm.org>
19687         * doc/extend.texi (Deprecated Features): Remove
19688         -ffriend-injection.
19689         (Backwards Compatibility): Likewise.
19690         * doc/invoke.texi (C++ Language Options): Likewise.
19691         (C++ Dialect Options): Likewise.
19693 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
19695         PR target/85530
19696         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
19697         _mm512_mask_mullox_epi64): New intrinsics.
19699 2018-05-03  Tom de Vries  <tom@codesourcery.com>
19701         PR testsuite/85106
19702         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19703         dump files): Add offload-tree.
19705 2018-05-03  Richard Biener  <rguenther@suse.de>
19707         PR tree-optimization/85615
19708         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
19709         to loops not nested in BBs loop father to avoid creating multi-entry
19710         loops.
19712 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19714         PR tree-optimization/70291
19715         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
19716         arguments.  Change return type to tree.  Emit libcall as a new
19717         statement rather than replacing existing one when inplace_p is true.
19718         (expand_complex_multiplication_components): New function.
19719         (expand_complex_multiplication): Expand floating-point complex
19720         multiplication using the above.
19721         (expand_complex_division): Rename inner_type parameter to type.
19722         Update expand_complex_libcall call-site.
19723         (expand_complex_operations_1): Update expand_complex_multiplication
19724         and expand_complex_division call-sites.
19726 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19728         PR target/85582
19729         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19730         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19731         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
19732         the highest significant bit of the shift count mask is clear.  In
19733         check whether and[sq]i3 is needed verify that all significant bits
19734         of the shift count other than the highest are set.
19736 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19738         PR libgomp/82428
19739         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
19740         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
19741         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
19742         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
19743         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
19744         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
19745         __builtin_goacc_parlevel_size.
19747 2018-05-02  Richard Biener  <rguenther@suse.de>
19749         PR tree-optimization/85597
19750         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
19751         do not use split vect_get_vec_defs call but call vect_get_slp_defs
19752         directly.
19754 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19756         PR testsuite/85106
19757         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19758         dump files): Add ltrans-tree.
19760 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19762         PR testsuite/85106
19763         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19764         dump files): Add wpa-ipa.
19766 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
19768         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
19769         powerpc*-*-linux*paired* target.
19770         * config/rs6000/750cl.h: Delete.
19771         * config/rs6000/paired.h: Delete.
19772         * config/rs6000/paired.md: Delete.
19773         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
19774         float support.
19775         * config/rs6000/rs6000-builtin.def: Remove paired float support.
19776         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
19777         comment.  Remove paired float support.
19778         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
19779         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
19780         VECTOR_PAIRED.
19781         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
19782         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
19783         declarations.
19784         * config/rs6000/rs6000.c: Remove paired float support.
19785         (paired_expand_vector_init, paired_expand_vector_move,
19786         paired_emit_vector_compare, paired_emit_vector_cond_expr,
19787         (paired_expand_lv_builtin, paired_expand_stv_builtin,
19788         paired_expand_builtin, paired_expand_predicate_builtin,
19789         paired_init_builtins): Delete.
19790         * config/rs6000/rs6000.h: Remove paired float support.
19791         * config/rs6000/rs6000.md: Remove paired float support.
19792         (move_from_CR_ov_bit): Delete.
19793         * config/rs6000/rs6000.opt (mpaired): Delete.
19794         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
19795         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
19797 2018-05-02  Richard Biener  <rguenther@suse.de>
19799         PR middle-end/85567
19800         * gimplify.c (gimplify_save_expr): When in SSA form allow
19801         SAVE_EXPRs to compute to SSA vars.
19803 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19805         PR target/85582
19806         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19807         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19808         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
19809         clobber operands[2], instead use a new pseudo.  Formatting fixes.
19811 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
19813         PR tree-optimization/85586
19814         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
19815         exit early for statements in the same group if the accesses are
19816         not strided.
19818 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19820         PR lto/85451
19821         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
19822         error message.
19824 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
19826         PR tree-optimization/85143
19827         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
19829 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19831         PR lto/85451
19832         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
19833         not found" error message.
19835 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19837         PR other/83786
19838         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
19839         * vec.c (test_ordered_remove_if): New function.
19840         (vec_c_tests): Call test_ordered_remove_if.
19841         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
19842         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
19843         * tree-vect-patterns.c (vect_pattern_recog_1): Use
19844         VEC_ORDERED_REMOVE_IF.
19846 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19848         PR tree-optimization/82665
19849         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
19850         pointer subtraction where arguments come from a memchr call.
19852 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
19854         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
19855         --push-state --as-needed and --pop-state instead of --as-needed and
19856         --no-as-needed if ld supports it.
19857         * configure: Regenerated.
19859         PR web/85578
19860         * doc/install.texi2html: Replace _002d with - and _002a with * in
19861         generated html files using sed.
19863 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19865         PR c++/85523
19866         * gcc-rich-location.c (blank_line_before_p): New function.
19867         (use_new_line): New function.
19868         (gcc_rich_location::add_fixit_insert_formatted): New function.
19869         * gcc-rich-location.h
19870         (gcc_rich_location::add_fixit_insert_formatted): New function.
19872 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19874         * selftest.c (assert_streq): Rename "expected" and "actual" to
19875         "val1" and "val2".  Extend NULL-handling to cover both inputs
19876         symmetrically, while still requiring both to be non-NULL for a pass.
19877         * selftest.h (assert_streq): Rename "expected" and "actual" to
19878         "val1" and "val2".
19879         (ASSERT_EQ): Likewise.
19880         (ASSERT_EQ_AT): Likewise.
19881         (ASSERT_KNOWN_EQ): Likewise.
19882         (ASSERT_KNOWN_EQ_AT): Likewise.
19883         (ASSERT_NE): Likewise.
19884         (ASSERT_MAYBE_NE): Likewise.
19885         (ASSERT_MAYBE_NE_AT): Likewise.
19886         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
19887         the assertion to pass.
19888         (ASSERT_STREQ_AT): Likewise.
19890 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
19892         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
19893         interaction with -pie.
19895 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19897         * selftest.h: Fix alphabetization of per-source-file selftest
19898         declarations.
19900 2018-04-30  Jason Merrill  <jason@redhat.com>
19902         PR c++/61982 - dead stores to destroyed objects.
19903         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
19904         of clobber.
19906 2018-04-30  Jason Merrill  <jason@redhat.com>
19908         * tree.c (build_clobber): New.
19909         * tree.h: Declare it.
19910         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
19912 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19914         * diagnostic-show-locus.c (layout::layout): Update for
19915         location_get_source_line returning a char_span.
19916         (struct char_span): Move to input.h.
19917         (struct correction): Update for fields in char_span becoming
19918         private.
19919         (struct source_line): Update for location_get_source_line
19920         returning a char_span.
19921         (layout::print_line): Likewise.
19922         * edit-context.c (edited_file::print_content): Likewise.
19923         (edited_file::print_diff_hunk): Likewise.
19924         (edited_file::print_run_of_changed_lines): Likewise.
19925         (edited_file::get_num_lines): Likewise.
19926         (edited_line::edited_line): Likewise.
19927         * final.c (asm_show_source): Likewise.
19928         * input.c (location_get_source_line): Convert return type
19929         from const char * to char_span, losing the final "line_len"
19930         param.
19931         (dump_location_info): Update for the above.
19932         (get_substring_ranges_for_loc): Likewise.  Use a char_span
19933         when handling the literal within the line.
19934         (test_reading_source_line): Update for location_get_source_line
19935         returning a char_span.
19936         * input.h (class char_span): Move here from
19937         diagnostic-show-locus.c, converting from a struct to a class.
19938         Make data members private.
19939         (char_span::operator bool): New.
19940         (char_span::length): New.
19941         (char_span::get_buffer): New.
19942         (char_span::operator[]): New.
19943         (char_span::subspan): Make const.
19944         (char_span::xstrdup): New.
19945         (location_get_source_line): Convert return type from const char *
19946         to char_span, losing the final "line_size" param.
19948 2018-04-30  Jan Hubicka  <jh@suse.cz>
19950         * lto-wrapper.c (ltrans_priorities): New static var.
19951         (cmp_priority): New.
19952         (run_gcc): Read priorities and if doing parallel build order
19953         the Makefile by them.
19955 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19957         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
19959 2018-04-30  Richard Biener  <rguenther@suse.de>
19961         * tree-cfg.c (verify_address): Remove base argument, add
19962         flag whether to check TREE_ADDRESSABLE and do that.
19963         (verify_expr): Remove.
19964         (verify_types_in_gimple_reference): Add pieces from verify_expr.
19965         (verify_gimple_assign_single): Likewise.
19966         (verify_gimple_switch): Likewise.
19967         (verify_expr_location_1): Dereference tp once.  Add (disabled)
19968         piece from verify_expr.
19969         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
19971 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19973         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
19975 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19977         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
19978         (small_data_pattern): Likewise.
19979         (arc_rewrite_small_data): Likewise.
19980         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
19981         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
19982         (get_symbol_alignment): New function.
19983         (legitimate_small_data_address_p): Likewise.
19984         (legitimate_scaled_address): Update, call
19985         legitimate_small_data_address_p.
19986         (output_sdata): New static variable.
19987         (arc_print_operand): Update how we handle small data operands.
19988         (arc_print_operand_address): Likewise.
19989         (arc_legitimate_address_p): Update, use
19990         legitimate_small_data_address_p.
19991         (arc_rewrite_small_data_p): Remove.
19992         (arc_rewrite_small_data_1): Likewise.
19993         (arc_rewrite_small_data): Likewise.
19994         (small_data_pattern): Likewise.
19995         (compact_sda_memory_operand): Update to use
19996         legitimate_small_data_address_p and get_symbol_alignment.
19997         (prepare_move_operands): Don't rewite sdata pattern.
19998         (prepare_extend_operands): Remove.
19999         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20000         pattern.
20001         (zero_extendqisi2): Likewise.
20002         (zero_extendhisi2): Likewise.
20003         (extendqihi2): Likewise.
20004         (extendqisi2): Likewise.
20005         (extendhisi2): Likewise.
20006         (addsi3): Likewise.
20007         (subsi3): Likewise.
20008         (andsi3): Likewise.
20009         * config/arc/constraints.md (Usd): Change it to memory constraint.
20011 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20013         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20014         as source of std instructions.
20015         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20016         allow 6-bit constants as source for store instructions.
20017         (movdi_insn): Update instruction pattern to allow 6-bit constants
20018         as source for store instructions.
20020 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20022         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20024 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20025             Sandra Loosemore <sandra@codesourcery.com>
20027         * dumpfile.c (dump_open): Allow '-' for stdout.
20028         * doc/invoke.texi (Developer Options): Document dump filename
20029         determination early.  Document stdin/stdout selection.
20031 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20033         Microblaze Target: PIC data text relative
20035         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20036         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20037         Add declaration.
20038         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20039         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20040         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20041         New addressing mode for data-text relative position indepenedent code.
20042         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20043         'ADDRESS_SYMBOLIC_TXT_REL'.
20044         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20045         (microblaze_legitimate_pic_operand): Exclude function calls from
20046         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20047         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20048         addresses cases.
20049         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20050         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20051         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20052         for 'address + offset'.
20053         (microblaze_expand_prologue): Add new function prologue call for
20054         'r20' assignation.
20055         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20056         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20057         table in case of TARGET_PIC_DATA_TEXT_REL.
20058         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20059         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20060         Add new macros 'UNSPEC_TEXT',
20061         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20062         + exclude function calls from 'UNSPEC_PLT' in case of data text
20063         relative mode.
20064         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20065         new target hook for generating address diff vector tables in case of
20066         flag_pic.
20067         * doc/tm.texi : Regenerate.
20068         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20069         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20070         of addr diff vector generation.
20071         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20072         target hook definition.
20073         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20074         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20075         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20076         option for data text relative.
20078 2018-04-30  Richard Biener  <rguenther@suse.de>
20080         * tree-chrec.h (evolution_function_is_constant_p): Remove
20081         redundant check.
20082         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20084 2018-04-30  Richard Biener  <rguenther@suse.de>
20086         PR bootstrap/85571
20087         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20089 2018-04-30  Richard Biener  <rguenther@suse.de>
20091         PR tree-optimization/28364
20092         PR tree-optimization/85275
20093         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20094         copying first exit test.
20096 2018-04-28  Mark Wielaard  <mark@klomp.org>
20098         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20099         dwarf_version >= 5.
20100         (dwarf_AT): Handle DW_AT_addr_base.
20101         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20103 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20105         PR target/84431
20106         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20107         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20108         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20109         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20111 2018-04-28  Richard Biener  <rguenther@suse.de>
20113         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20114         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20115         to reflect use.  Only add interesting stmts.
20117 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20119         PR ipa/85549
20120         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20121         the jump function allows for passing through aggregate values.
20123 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20125         * input.h (in_system_header_at): Convert from macro to inline
20126         function.
20127         (from_macro_expansion_at): Likewise.
20128         (from_macro_definition_at): Likewise.
20130 2018-04-27  Jeff Law  <law@redhat.com>
20132         * config.gcc: Mark tile* targets as deprecated/obsolete.
20134 2018-04-27  Richard Biener  <rguenther@suse.de>
20136         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20137         fix for ILP32.
20139 2018-04-27  Richard Biener  <rguenther@suse.de>
20141         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20143 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
20145         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20146         with Yd constraint. Set "preferred_for_speed" attribute from
20147         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20148         with Yd constraint.
20149         (*movdi_internal): Ditto.
20150         (movti_interunit splitters): Remove
20151         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20152         (movdi_interunit splitters): Ditto.
20153         * config/i386/constraints.md (Ye): Remove.
20154         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20156 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20158         PR target/85512
20159         * config/aarch64/constraints.md (Usg): Limit to 31.
20160         (Usj): Limit to 63.
20162 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
20164         PR tree-optimization/85529
20165         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20166         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20167         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20168         zero extension or masking of the MSB bit.
20169         (optimize_range_tests): Add FIRST_BB argument, pass it through
20170         to optimize_range_tests_var_bound.
20171         (maybe_optimize_range_tests, reassociate_bb): Adjust
20172         optimize_range_tests callers.
20174 2018-04-26  Richard Biener  <rguenther@suse.de>
20175             Jakub Jelinek  <jakub@redhat.com>
20177         * cgraph.h (symbol_table): Just declare debug method here.
20178         * symtab.c (symbol_table::debug): Define.
20180 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
20182         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20184 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
20186         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20187         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20188         (*movdi_internal): Substitute Yi and Yj constraint with x
20189         and Ym and Yn constraint with y constraint.  Update "isa"
20190         attribute and set "preferred_for_speed" attribute from
20191         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20192         (*movsi_internal): Ditto.
20193         (*movdf_internal): Ditto.
20194         (*movsf_internal): Ditto.
20195         (*zero_extendsidi2): Ditto.
20196         * config/i386/sse.md (vec_set<mode>_0): Ditto.
20197         (sse2_loadld): Ditto.
20198         (*vec_extract<ssevecmodelower>_0): Ditto.
20199         (*vec_extractv4si_0_zext_sse4): Ditto.
20200         (vec_concatv2di): Ditto.
20201         (*vec_dup<mode>): Ditto.
20202         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20203         * config/i386/constraints.md (Yi): Remove.
20204         (Yj): Remove.
20205         (Ym): Remove.
20206         (Yn): Remove.
20208 2018-04-26  Nathan Sidwell  <nathan@acm.org>
20210         * dumpfile.c (dump_open): New.
20211         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20212         (dump_finish): Detect stdio/stderr by value not name.
20214 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20216         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20218 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20220         PR target/84952
20221         * config/nvptx/nvptx.c (verify_neutering_jumps)
20222         (verify_neutering_labels): New function
20223         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20225 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20227         PR target/84025
20228         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20229         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20230         need neutering.
20232 2018-04-26  Richard Biener <rguenther@suse.de>
20233             Tom de Vries  <tom@codesourcery.com>
20235         PR lto/85422
20236         * lto-streamer-out.c (output_function): Fixup loops if required to match
20237         discovery done in the reader.
20239 2018-04-26  Richard Biener  <rguenther@suse.de>
20241         PR tree-optimization/85116
20242         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20243         have a loop exit from the single latch predecessor.  Remove
20244         case of header with just condition.
20245         (ch_base::copy_headers): Exclude infinite loops from any
20246         processing.
20247         (pass_ch::execute): Record exits.
20249 2018-04-26  Richard Biener  <rguenther@suse.de>
20251         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20252         prologue cost vector and pass it to vect_get_load_cost.
20253         (vect_get_peeling_costs_all_drs): Likewise.
20254         (vect_peeling_hash_get_lowest_cost): Likewise.
20255         (vect_enhance_data_refs_alignment): Likewise.
20257 2018-04-26  Richard Biener  <rguenther@suse.de>
20259         PR middle-end/85450
20260         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20261         checking of integer<->pointer conversions.
20262         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20263         sign-/zero-extending pointer types.
20264         (expand_omp_for_static_chunk): Likewise.
20266 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
20267             Jean Lee  <xiaoyur347@gmail.com>
20269         * config/mips/mips.c (mips_asan_shadow_offset): New function.
20270         (TARGET_ASAN_SHADOW_OFFSET): Define.
20271         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20272         true for -fsanitize=address.
20274 2018-04-25  Mark Wielaard  <mark@klomp.org>
20276         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20277         shorter ones.
20279 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20281         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20282         than "alu", remove explicit "memory" and "imm_disp" attributes.
20283         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20285         PR middle-end/85414
20286         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20287         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20288         gen_lowpart_no_emit.
20290 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
20292         PR target/85473
20293         * config/i386/i386.c (ix86_expand_builtin): Change memory
20294         operand to XI, extend p0 to Pmode.
20295         * config/i386/i386.md: Change unspec volatile and operand
20296         1 mode to XI, change operand 0 mode to P.
20298 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20300         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20301         GET_MODE_MASK before any checking.
20302         (nds32_can_use_bset_p): Likewise.
20303         (nds32_can_use_btgl_p): Likewise.
20305 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20307         * config/nds32/nds32-doubleword.md: New define_split pattern for
20308         illegal register number.
20310 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20312         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20314 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20316         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20318 2018-04-25  Richard Biener  <rguenther@suse.de>
20320         * lto-streamer.h (LTO_major_version): Bump to 8.
20322 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20324         * BASE-VER: Set to 9.0.0.
20326 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
20328         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20329         in __abskf2 and __powikf2.
20331 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20333         PR target/85512
20334         * config/aarch64/constraints.md (Usg, Usj): New constraints.
20335         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20336         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20337         Use the above on operand 2.  Reindent.
20338         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20340 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20342         PR target/85485
20343         * common/config/i386/i386-common.c (ix86_handle_option): Don't
20344         handle OPT_mcet.
20345         * config/i386/i386.opt (mcet): Removed.
20346         * doc/install.texi: Remove -mcet documentation.
20347         * doc/invoke.texi: Likewise.
20349 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20351         PR target/85485
20352         * doc/install.texi: Remove -mcet from bootstrap-cet.
20354 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20356         PR target/85511
20357         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20358         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20359         if TARGET_64BIT.
20361         PR target/85503
20362         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20363         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20364         containing a CONST_VECTOR.
20366 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20368         * doc/install.texi: Update newlib dependency for nvptx.
20370 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20372         PR target/85508
20373         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20374         instead of INTVAL when shifting x left.
20376 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20378         PR tree-optimization/85478
20379         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20380         vect_grouped_store_supported for single element vectors.
20382 2018-04-24  Richard Biener  <rguenther@suse.de>
20384         PR target/85491
20385         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20386         load cost increase to the case of non-constant step.
20388 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20390         PR target/84828
20391         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20392         destination if any_malformed_asm.
20394 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
20396         PR middle-end/85496
20397         * expr.c (store_field): In the bitfield case, if the value comes from
20398         a function call and is returned in registers by means of a PARALLEL,
20399         do not change the mode of the temporary unless BLKmode and VOIDmode.
20401 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
20403         PR rtl-optimization/85423
20404         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20405         dependencies to debug insns when the previous insn is non-debug.
20407 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20409         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20410         enums into a single definition.
20411         (fls): Fix predicates and printing.
20412         (seti): Likewise.
20414 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20416         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20417         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20418         and short u6 immediate.
20419         (check_if_valid_sleep_operand): Remove.
20420         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20422 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20424         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20425         flag_always_save_lp condition.
20426         * config/nds32/nds32.opt (malways-save-lp): New option.
20428 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20430         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20431         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20432         * config/nds32/nds32.h
20433         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20434         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20436 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20438         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20439         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20441 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20442             Chung-Ju Wu  <jasonwucj@gmail.com>
20444         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20445         Declare.
20446         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20447         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20449 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20451         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20453 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20455         * config/nds32/nds32-protos.h (nds32_data_alignment,
20456         nds32_local_alignment): Declare.
20457         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20458         nds32_local_alignment): New functions.
20459         (TARGET_CONSTANT_ALIGNMENT): Define.
20460         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20462 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20464         * config/nds32/nds32.c
20465         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20466         (TARGET_MODES_TIEABLE_P): Likewise.
20468 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20470         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20471         level Ofast and Og.
20473 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
20474             Chung-Ju Wu  <jasonwucj@gmail.com>
20476         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20477         for unaligned access.
20478         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20479         unaligned access.
20480         * config/nds32/nds32-intrinsic.md: Likewise.
20481         * config/nds32/nds32_intrinsic.h: Likewise.
20482         * config/nds32/nds32.h (nds32_builtins): Likewise.
20483         * config/nds32/nds32.opt (munaligned-access): New option.
20484         * config/nds32/nds32.c (nds32_asm_file_start): Display
20485         flag_unaligned_access status.
20487 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20489         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20490         -mno-relax is present.
20491         * config/riscv/linux.h (LINK_SPEC): Ditto.
20493 2018-04-20  Martin Sebor  <msebor@redhat.com>
20495         PR c/85365
20496         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20497         for null pointers.
20498         (gimple_fold_builtin_stxcpy_chk): Same.
20499         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20501 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
20503         PR target/85456
20504         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20505         __powikf2 when long double is IEEE 128-bit.
20507 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20509         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20510         step to make sure stack always aligned.
20512 2018-04-20  Carl Love  <cel@us.ibm.com>
20514         PR target/83402
20515         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20516         size check for arg0.
20518 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
20519             Tom de Vries  <tom@codesourcery.com>
20521         PR target/85445
20522         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20523         Emit insns for calls too.
20524         (nvptx_find_par): Always look for worker-level predecessor insn.
20525         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
20526         calls.
20527         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
20528         (nvptx_process_pars): Propagate frames for calls.
20530 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
20532         PR target/85469
20533         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
20534         Removed.
20535         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
20536         (ix86_handle_option): Don't handle OPT_mibt.
20537         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
20538         __SHSTK__.
20539         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
20540         has_ibt and ibt.
20541         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
20542         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
20543         (ix86_target_macros): Define __CET__ with flag_cf_protection
20544         for -fcf-protection.
20545         * config/i386/i386.c (isa2_opts): Remove -mibt.
20546         * config/i386/i386.h (TARGET_IBT): Removed.
20547         (TARGET_IBT_P): Likewise.
20548         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
20549         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
20550         * config/i386/i386.opt (mcet): Update help message.
20551         (mshstk): Likewise.
20552         (mibt): Removed.
20553         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
20554         -mcet as an alias for -mshstk.
20556 2018-04-20  Richard Biener <rguenther@suse.de>
20558         PR middle-end/85475
20559         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
20560         complexity by forcing a single use of the multiply operand.
20562 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20564         ipa/85449
20565         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
20566         recursion dependency to only apply to non-clones.
20568 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20570         ipa/85447
20571         * ipa-cp.c (create_specialized_node): Check that clones of
20572         self-recursive edges exist during IPA-CP.
20574 2018-04-19  Toon Moene  <toon@moene.org>
20576         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
20577         by -O3.
20579 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20581         PR tree-optimization/85467
20582         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
20583         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
20584         VECTOR_CST element to type.
20586 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20588         PR target/85397
20589         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
20590         * config/i386/i386.md (builtin_setjmp_setup): Removed.
20591         (builtin_longjmp): Likewise.
20592         (save_stack_nonlocal): New pattern.
20593         (restore_stack_nonlocal): Likewise.
20595 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20597         PR target/85404
20598         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20599         Replace ASM_OUTPUT_LABEL with fprintf.
20601 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20603         PR target/85417
20604         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20605         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
20606         * config/i386/i386-c.c (ix86_target_macros_internal): Also
20607         define __IBT__ and __SHSTK__ for -fcf-protection.
20608         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
20609         TARGET_IBT.
20610         (ix86_trampoline_init): Likewise.
20611         (x86_output_mi_thunk): Likewise.
20612         (ix86_notrack_prefixed_insn_p): Likewise.
20613         (ix86_option_override_internal): Don't disallow -fcf-protection.
20614         * config/i386/i386.md (rdssp<mode>): Also enable for
20615         -fcf-protection.
20616         (incssp<mode>): Likewise.
20617         (nop_endbr): Likewise.
20618         * config/i386/i386.opt (mcet): Change help message to built-in
20619         functions only.
20620         (mibt): Likewise.
20621         (mshstk): Likewise.
20622         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
20623         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
20624         enable CET built-in functions.
20626 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
20628         * common/config/i386/i386-common.c
20629         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
20630         OPTION_MASK_ISA_MOVDIRI_UNSET,
20631         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
20632         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
20633         * config.gcc (movdirintrin.h): New header.
20634         * config/i386/cpuid.h (bit_MOVDIRI,
20635         bit_MOVDIR64B): New bits.
20636         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
20637         and -mmvodir64b.
20638         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
20639         (VOID, PVOID, PCVOID)): New function types.
20640         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
20641         __builtin_ia32_directstoreu_u64,
20642         __builtin_ia32_movdir64b): New builtins.
20643         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
20644         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
20645         and -mmovdiri.
20646         (ix86_valid_target_attribute_inner_p): Ditto.
20647         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
20648         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
20649         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
20650         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
20651         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
20652         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
20653         (movdiri<mode>, movdir64b_<mode>): New.
20654         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
20655         * config/i386/immintrin.h: Include movdirintrin.h.
20656         * config/i386/movdirintrin.h: New file.
20657         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
20659 2018-04-19  Richard Biener  <rguenther@suse.de>
20661         PR middle-end/85455
20662         * cfg.c (clear_bb_flags): When loop state says we have
20663         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
20665 2018-04-19  Richard Biener  <rguenther@suse.de>
20667         PR tree-optimization/84737
20668         * tree-vect-data-refs.c (vect_copy_ref_info): New function
20669         copying restrict info.
20670         (vect_setup_realignment): Use it.
20671         * tree-vectorizer.h (vect_copy_ref_info): Declare.
20672         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
20673         the first DR to all generated stores.
20674         (vectorizable_load): Likewise for loads.
20676 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20678         PR tree-optimization/85446
20679         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
20680         the integral and pointer types to have the same precision.
20682         * doc/install.texi: Document --disable-cet being the default and
20683         --enable-cet=auto.
20685 2018-04-18  Martin Liska  <mliska@suse.cz>
20687         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
20688         style.
20690 2018-04-18  Martin Liska  <mliska@suse.cz>
20692         Revert
20693         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20695         PR ipa/83983
20696         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20697         arguments if they are comparable.
20699 2018-04-18  Martin Liska  <mliska@suse.cz>
20701         Revert
20702         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20704         PR lto/84805
20705         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20706         incomplete types.
20708 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
20710         PR target/85388
20711         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
20712         ENDBR after calling __morestack.
20714 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
20716         PR jit/85384
20717         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
20718         by using gcc_base_ver to generate a gcc_driver_version, and use
20719         it when generating GCC_DRIVER_NAME.
20720         * configure: Regenerate.
20722 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
20724         PR target/81084
20725         * config.gcc: Obsolete powerpc*-*-*spe*.
20727 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20729         PR debug/84637
20730         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
20731         (stabstr_D): Change type of unum from unsigned int to
20732         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
20733         type.
20735 2018-04-17  Jim Wilson  <jimw@sifive.com>
20737         PR 84856
20738         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
20739         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
20740         Set arg_pointer_offset after using pretend_args_size.
20742 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20744         PR rtl-optimization/85431
20745         * dse.c (record_store): Ignore zero width stores.
20747         PR sanitizer/85230
20748         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
20749         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
20750         __builtin_stack_restore rather than after it.
20751         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
20752         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
20753         argument instead of virtual_dynamic_stack_rtx.
20755 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20757         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
20758         New prototype.
20759         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20760         Add note to error message to explain internal mapping of overloaded
20761         built-in function name to non-overloaded built-in function name.
20762         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
20763         function.
20765 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
20767         PR target/85424
20768         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
20769         where the inputs overlap with the output.
20771 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20773         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
20774         (=v, v) alternative and explicit "memory" attribute.
20775         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
20776         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20777         attributes.
20778         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
20779         "sselog1" type instead of "sselog".
20780         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
20781         "sselog".  Remove explicit "memory" attribute.
20782         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
20783         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20784         attributes.
20785         (vec_extract_hi_v32hi): Merge all alternatives into one, use
20786         "sselog1" type instead of "sselog".  Remove explicit "memory"
20787         attribute.
20788         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
20789         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20790         attribute.
20791         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
20792         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20793         attributes.
20794         (vec_extract_hi_v64qi): Merge all alternatives into one, use
20795         "sselog1" type instead of "sselog".  Remove explicit "memory"
20796         attribute.
20797         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
20798         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20799         attribute.
20801         PR target/85430
20802         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
20804         PR middle-end/85414
20805         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
20806         on a SUBREG.
20808 2018-04-17  Martin Jambor  <mjambor@suse.cz>
20810         PR ipa/85421
20811         * ipa-cp.c (create_specialized_node): Call
20812         expand_all_artificial_thunks if necessary.
20814 2018-04-17  Martin Liska  <mliska@suse.cz>
20816         PR lto/85405
20817         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
20818         in message, remote space in between '_G' and '('.
20820 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20822         PR target/85281
20823         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
20824         avx512f_vmcmp<mode>3<round_saeonly_name>,
20825         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
20826         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
20827         avx512f_rndscale<mode><round_saeonly_name>,
20828         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
20829         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
20830         Use %<iptr>2 instead of %2 for -masm=intel.
20831         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
20832         avx512f_vcvttss2usi<round_saeonly_name>,
20833         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
20834         -masm=intel.
20835         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
20836         avx512f_vcvttsd2usi<round_saeonly_name>,
20837         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
20838         Use %q1 instead of %1 for -masm=intel.
20839         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
20840         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
20841         of %3 for -masm=intel.
20842         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
20843         -masm=intel.
20844         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
20845         -masm=intel.
20846         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
20847         -masm=intel.
20848         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
20849         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
20850         %g1.
20851         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
20852         -masm=intel.
20853         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
20854         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
20855         %g1 and one with %0 and %1.
20856         (avx512er_vmrcp28<mode><round_saeonly_name>,
20857         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
20858         %1 for -masm=intel.
20859         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
20860         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
20861         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
20862         of %0 and %{%4%} for -masm=intel.
20863         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
20864         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
20865         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
20866         order of %0 and %{%5%}%{z%} for -masm=intel.
20868 2018-04-17  Jan Hubicka  <jh@suse.cz>
20870         PR lto/85405
20871         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
20873 2018-04-17  Martin Liska  <mliska@suse.cz>
20875         PR ipa/85329
20876         * multiple_target.c (create_dispatcher_calls): Set apostrophes
20877         for target_clone error message.  Make default implementation
20878         clone to be a local declaration.
20879         (separate_attrs): Add new argument and check for an empty
20880         string.
20881         (expand_target_clones): Handle it.
20882         (ipa_target_clone): Make redirection just for target_clones
20883         functions.
20885 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
20886             Tom de Vries  <tom@codesourcery.com>
20888         PR middle-end/84955
20889         * omp-expand.c (expand_oacc_for): Add dummy false branch for
20890         tiled basic blocks without omp continue statements.
20892 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
20894         PR target/83660
20895         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
20896         vec_extract expression as having side effects to make sure it gets
20897         a cleanup point.
20899 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
20901         PR target/85403
20902         * config/i386/i386.c (get_builtin_code_for_version): Check
20903         error_mark_node.
20905 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
20907         PR target/84331
20908         * config.gcc: Support "skylake".
20909         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20910         PROCESSOR_SKYLAKE.
20911         * config/i386/i386.c (m_SKYLAKE): Define.
20912         (processor_target_table): Add "skylake".
20913         (ix86_option_override_internal): Add "skylake".
20914         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
20915         PROCESSOR_CANNONLAKE.
20916         (get_builtin_code_for_version): Fix priority for
20917         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
20918         PROCESSOR_SKYLAKE-AVX512.
20919         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
20920         (processor_type): Add PROCESSOR_SKYLAKE.
20922 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
20923             Jason Merrill  <jason@redhat.com>
20925         PR c++/85112
20926         * convert.c (convert_to_integer_1): Use direct recursion for
20927         enumeral types and types with a precision less than the number
20928         of bits in their mode.
20930 2018-04-16  Julia Koval  <julia.koval@intel.com>
20932         PR target/84413
20933         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
20934         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
20936 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
20938         PR target/85293
20939         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
20940         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
20941         and -mno-direct-move.
20943 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
20945         PR target/83402
20946         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
20947         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
20948         Ensure negative shifts result in {0}.
20950 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
20952         PR rtl-optimization/79916
20953         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
20954         regs (if any) to define how to gnerate SD moves when LRA is in
20955         progress.
20957 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
20959         PR rtl-optimization/85393
20960         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
20961         * except.c (expand_dw2_landing_pad_for_region): Make static.
20962         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
20963         a label and unconditional jump to old_bb, rather than
20964         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
20965         basic block.
20967         PR rtl-optimization/85376
20968         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
20969         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
20970         instead of a specific value.
20972 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20973             Bin Cheng  <bin.cheng@arm.com>
20975         PR tree-optimization/82965
20976         PR tree-optimization/83991
20977         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
20978         by_profile_only parameter.
20979         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
20980         information if the loop was predicted to iterate too many times.
20981         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
20983 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20985         PR lto/71991
20986         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
20987         always inline.
20989 2018-04-13  Martin Liska  <mliska@suse.cz>
20990             Jakub Jelinek  <jakub@redhat.com>
20992         PR middle-end/81657
20993         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
20994         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
20995         * builtins.c (expand_builtin_memory_copy_args): Use
20996         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
20997         handle dest_addr == pc_rtx.
20999 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21001         PR target/85291
21002         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21003         asked to not generate direct moves.
21004         (fix_trunc<mode>si2_stfiwx): Similar.
21005         (fix_trunc<mode>si2_internal): Similar.
21007 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21009         PR debug/83157
21010         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21011         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21012         lookup if dest in some wider mode is known to be const0_rtx and
21013         if so, record permanent equivalence for it to be ZERO_EXTEND of
21014         the narrower mode destination.
21016 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21018         * lto-streamer-out.c (output_function): Revert 259346.
21019         * omp-expand.c (expand_oacc_for): Likewise.
21021 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21023         PR rtl-optimization/85354
21024         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21025         * sel-sched.c (sel_global_init): ... here.
21027 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21029         PR target/85238
21030         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21031         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21032         mode for PE-COFF targets.
21033         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21034         (i386_pe_asm_lto_end): Likewise.
21035         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21036         (TARGET_ASM_LTO_END): Likewise.
21037         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21038         (i386_pe_asm_lto_start): New function.
21039         (i386_pe_asm_lto_end): Likewise.
21041 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21042             Richard Biener  <rguenther@suse.de>
21044         PR middle-end/84955
21045         * lto-streamer-out.c (output_function): Fix CFG loop state before
21046         streaming out.
21047         * omp-expand.c (expand_oacc_for): Handle calls to internal
21048         functions like regular functions.
21050 2018-04-12  Richard Biener  <rguenther@suse.de>
21052         PR lto/85371
21053         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21054         for the early LTO debug to properly generate references to it
21055         during DIE emission.  Do not re-use that for the skeleton for
21056         split-dwarf.
21057         (dwarf2out_early_finish): Likewise.
21059 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21061         PR target/85328
21062         * config/i386/sse.md
21063         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21064         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21065         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21066         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21067         and output is a reg, avoid creating invalid lowpart subreg, but
21068         instead split into a 512-bit move.  Don't split if not AVX512VL,
21069         input is xmm16+ reg and output is a mem.
21070         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21071         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21072         xmm16+ reg and output is a mem.
21074 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21076         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21077         also for flag_dwarf2_cfi_asm.
21079 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21081         PR rtl-optimization/85342
21082         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21083         a bool scalar var inside of the loop instead.  Don't try to update
21084         recog_data.operand after failed apply_change_group.
21086 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21088         PR target/85296
21089         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21090         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21091         array with flexible array member as array without given dimension.
21092         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21093         argument for undefined param to true.
21095 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21097         PR target/85321
21098         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21099         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21100         from PowerPC section.
21101         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21102         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21103         help text that is too long.
21104         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21105         help text that is too long.
21106         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21107         help text that is too long.
21109 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21111         * config/alpha/alpha.md (stack_probe_internal): Rename
21112         from "probe_stack".  Update all callers.
21114 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21116         PR rtl-optimization/84566
21117         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21118         sched_macro_fuse_insns.
21120 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21122         PR target/84301
21123         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21124         (compute_block_dependences): ... from here.
21126 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21128         PR tree-optimization/85331
21129         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21130         from int to HOST_WIDE_INT.
21132 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21134         PR ipa/84149
21135         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21136         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21137         not the same as the source val.
21138         (cgraph_edge_brings_value_p): New parameter.
21139         (gather_edges_for_value): Pass destination value to
21140         cgraph_edge_brings_value_p.
21141         (perhaps_add_new_callers): Likewise.
21142         (get_info_about_necessary_edges): Likewise and exclude values brought
21143         only by self-recursive edges.
21144         (create_specialized_node): Redirect only clones of self-calling edges.
21145         (+self_recursive_pass_through_p): New function.
21146         (find_more_scalar_values_for_callers_subset): Use it.
21147         (find_aggregate_values_for_callers_subset): Likewise.
21148         (known_aggs_to_agg_replacement_list): Removed.
21149         (decide_whether_version_node): Re-calculate known constants for all
21150         remaining context clones.
21152 2018-04-11  Richard Biener  <rguenther@suse.de>
21154         PR lto/85339
21155         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21156         from early DWARF output.
21157         (dwarf2out_early_finish): Output line info unconditionally into
21158         early DWARF and add reference to it.
21160 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21162         PR target/85281
21163         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21164         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21165         other than V2DFmode using iptr mode attribute.
21166         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21168 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21170         PR rtl-optimization/84659
21171         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21173 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21175         PR debug/85302
21176         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21177         SIZEP is NULL.
21178         (output_loc_list): Pass address of a dummy size variable even in the
21179         locview handling loop.
21180         (index_location_lists): Add comment on why skip_loc_list_entry can't
21181         call size_of_locs.
21183 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21185         PR target/85261
21186         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21187         into register.
21189 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
21191         PR target/85321
21192         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21193         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21194         and -mstring-compare-inline-limit.
21196 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21198         PR target/85287
21199         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21200         for stack clash protection in a register whenever we need it to be in
21201         a register.
21203 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21205         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21206         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21208 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21210         PR target/85321
21211         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21212         the help text.
21213         (mlong-double-): Ditto.
21214         * config/rs6000/sysv4.opt (msdata=): Ditto.
21215         (mtls-size=): Ditto.
21217 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21219         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21220         erroneous entries for
21221         "vector int vec_ldl (int, long int *)", and
21222         "vector unsigned int vec_ldl (int, unsigned long int *)".
21223         Add comments and entries for
21224         "vector bool char vec_ldl (int, bool char *)",
21225         "vector bool short vec_ldl (int, bool short *)",
21226         "vector bool int vec_ldl (int, bool int *)",
21227         "vector bool long long vec_ldl (int, bool long long *)",
21228         "vector pixel vec_ldl (int, pixel *)",
21229         "vector long long vec_ldl (int, long long *)",
21230         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21231         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21232         type tree bool_long_long_type_node and correct definition of
21233         bool_V2DI_type_node to make reference to this new type tree.
21234         (rs6000_mangle_type): Replace erroneous reference to
21235         bool_long_type_node with bool_long_long_type_node.
21236         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21237         comments to emphasize sign distinctions for char and int types and
21238         replace RS6000_BTI_bool_long constant with
21239         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21240         use of RS6000_BTI_pixel.
21241         (bool_long_type_node): Remove this macro definition.
21242         (bool_long_long_type_node): New macro definition
21244 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21246         PR rtl-optimization/85300
21247         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21248         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21249         simplify_unary_operation fails.
21251 2018-04-10  Martin Liska  <mliska@suse.cz>
21253         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21254         cgraph_edge and ipa_ref.
21256 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21258         PR target/85177
21259         PR target/85255
21260         * config/i386/sse.md
21261         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21262         computation of the VEC_MERGE selector from mask.
21263         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21264         Fix decoding of the VEC_MERGE selector into mask.
21266 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21268         PR tree-optimization/85286
21269         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21271 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21273         * final.c (final_1): Set insn_last_address as well as
21274         insn_current_address.
21276 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21278         PR target/85173
21279         * explow.c (emit_stack_probe): Call validize_mem on memory location
21280         before passing it to gen_probe_stack.  Create address operand and
21281         legitimize it for the probe_stack_address case.
21283 2018-04-09  Jan Hubicka  <jh@suse.cz>
21285         PR lto/85078
21286         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21287         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21288         * tree.c (free_lang_data_in_type): Fix handling of binfos;
21289         walk basetypes.
21290         (free_lang_data): Rebuild type inheritance graph.
21292 2018-04-09  Martin Sebor  <msebor@redhat.com>
21294         * invoke.texi (-finline-small-functions): Mention other optimization
21295         options.
21296         (-findirect-inlining, -fpartial-inlining): Same.
21297         (-finline-functions-called-once): Same.
21298         (-freorder-blocks-and-partition): Same.
21300 2018-04-09  Jan Hubicka  <jh@suse.cz>
21302         PR rtl/84058
21303         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21304         jumps; choose last target that matches the criteria (i.e.
21305         no partition changes for non-crossing jumps).
21306         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21307         support for redirecting crossing jumps to non-crossing.
21309 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
21311         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21312         also for naked functions.
21314 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
21316         * config/arc/arc.md (add_shift): New pattern.
21317         (add_shift2): Likewise.
21318         (sub_shift): Likewise.
21319         (sub_shift_cmp0_noout): Likewise.
21320         (compare_si_ashiftsi): Likewise.
21321         (xbfu_cmp0_noout): New combine pattern.
21322         (xbfu_cmp0"): Likewise.
21323         (movsi_set_cc_insn): Place the predicable variant first.
21324         (commutative_binary_cmp0_noout): Remove clobber.
21325         (commutative_binary_cmp0): New pattern.
21326         (noncommutative_binary_cmp0): Likewise.
21327         (noncommutative_binary_cmp0_noout): Likewise.
21328         (noncommutative_binary_comparison_result_used): Removed.
21329         (rsub_cmp0): New pattern.
21330         (rsub_cmp0_noout): Likewise.
21331         (extzvsi): Changed, keep only meaningful variants.
21332         (SQH, SEZ): New iterators.
21333         (SQH_postfix): New mode attribute.
21334         (SEZ_prefix): New code attribute.
21335         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21336         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21337         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21338         of numerical value.
21339         (noncommutative_operator): Check the availability of barrel
21340         shifter option.
21342 2018-04-09  Richard Biener  <rguenther@suse.de>
21344         PR tree-optimization/85284
21345         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21346         Only use the niter constraining form of simple_iv when the exit
21347         is always executed.
21349 2018-04-09  Tom de Vries  <tom@codesourcery.com>
21351         PR target/84041
21352         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21353         (define_expand "*memory_barrier"): New define_expand.
21354         (define_insn "memory_barrier"): New insn.
21356 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21358         PR rtl-optimization/80463
21359         PR rtl-optimization/83972
21360         PR rtl-optimization/83480
21362         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21363         correct producer for the insn.
21364         (tidy_control_flow): Fixup seqnos in case of debug insns.
21366 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21368         PR rtl-optimization/83913
21370         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21371         different sched-times when merging exprs.
21373 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21375         PR rtl-optimization/83962
21377         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21378         tidy_fallthru_edge and tidy_control_flow.
21380 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21382         PR rtl-optimization/83530
21384         * sel-sched.c (force_next_insn): New global variable.
21385         (remove_insn_for_debug): When force_next_insn is true, also leave only
21386         next insn in the ready list.
21387         (sel_sched_region): When the region wasn't scheduled, make another pass
21388         over it with force_next_insn set to 1.
21390 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21392         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21393         into tm_file.
21394         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21395         for interrupt control.
21396         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21397         functions for interrupt control.
21398         * config/nds32/nds32-intrinsic.md: Likewise.
21399         * config/nds32/nds32_intrinsic.h: Likewise.
21400         * config/nds32/nds32.h (nds32_builtins): Likewise.
21402 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21404         * config/nds32/nds32.c (nds32_init_machine_status,
21405         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21406         strict_aligned_p field.
21407         (nds32_expand_to_rtl_hook): New function.
21408         (TARGET_EXPAND_TO_RTL_HOOK): Define.
21409         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21411 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21412             Chung-Ju Wu  <jasonwucj@gmail.com>
21414         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21415         * config/nds32/nds32-n7.md: New file.
21416         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21417         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21418         pipeline.
21419         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21420         * config/nds32/nds32.md (pipeline_model): Add n7.
21421         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21422         * config/nds32/pipelines.md: Include n7 settings.
21424 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21425             Chung-Ju Wu  <jasonwucj@gmail.com>
21427         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21428         * config/nds32/nds32-e8.md: New file.
21429         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21430         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21431         pipeline.
21432         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21433         * config/nds32/nds32.md (pipeline_model): Add e8.
21434         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21435         * config/nds32/pipelines.md: Include e8 settings.
21437 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21438             Chung-Ju Wu  <jasonwucj@gmail.com>
21440         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21441         * config/nds32/nds32-n8.md: New file.
21442         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21443         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21444         pipeline.
21445         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21446         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21447         * config/nds32/nds32.md (pipeline_model): Add n8.
21448         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21449         * config/nds32/pipelines.md: Include n8 settings.
21451 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21452             Chung-Ju Wu  <jasonwucj@gmail.com>
21454         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21455         * config/nds32/nds32-n9-2r1w.md: New file.
21456         * config/nds32/nds32-n9-3r2w.md: New file.
21457         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21458         nds32_register_ports): New or modify for cpu n9.
21459         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21460         pipeline.
21461         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21462         * config/nds32/nds32-utils.c: New file.
21463         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21464         TARGET_MUL_SLOW): Define.
21465         * config/nds32/nds32.md (pipeline_model): New attribute.
21466         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21467         New options that support cpu n9.
21468         * config/nds32/pipelines.md: Include n9 settings.
21469         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21471 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21473         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21474         information if necessary.
21475         (output_cond_branch_compare_zero): Likewise.
21476         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21477         (nds32_target_alignment): Refine for alignment.
21478         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21479         (FUNCTION_BOUNDARY): Modify.
21480         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21481         align case.
21482         * config/nds32/nds32.opt (malways-align, malign-functions): New.
21484 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21486         * config/nds32/constants.md (unspec_volatile_element): Add values for
21487         TLB operation and data prefetch.
21488         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21489         functions for TLB operation and data prefetch.
21490         * config/nds32/nds32-intrinsic.md: Likewise.
21491         * config/nds32/nds32_intrinsic.h: Likewise.
21492         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21493         (nds32_print_operand): Likewise.
21494         * config/nds32/nds32.h (nds32_builtins): Likewise.
21496 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21497         Andrew Pinski <pinsika@gcc.gnu.org>
21499         PR middle-end/82976
21500         * match.pd: Use constant_boolean_node of correct type instead of
21501         boolean_true_node or boolean_false_node for simplifying
21502         pointer comparisons to zero.
21504 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21506         PR tree-optimization/80021
21507         * tree.c (verify_type_variant): Make error call in verify_variant_match
21508         translatable and remove final full stop.
21510 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21512         * config/nds32/constants.md (unspec_volatile_element): Add
21513         UNSPEC_VOLATILE_EH_RETURN.
21514         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21515         nds32_output_stack_pop): Support dwarf exception handling process.
21516         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21517         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21518         exception handling process.
21519         (nds32_compute_stack_frame): Likewise.
21520         (nds32_return_addr_rtx): Likewise.
21521         (nds32_initial_elimination_offset): Likewise.
21522         (nds32_expand_prologue): Likewise.
21523         (nds32_expand_epilogue): Likewise.
21524         (nds32_dynamic_chain_address): New function.
21525         * config/nds32/nds32.h (machine_function): Add fields for dwarf
21526         exception handling.
21527         (DYNAMIC_CHAIN_ADDRESS): Define.
21528         (EH_RETURN_DATA_REGNO): Define.
21529         (EH_RETURN_STACKADJ_RTX): Define.
21530         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
21531         patterns for dwarf exception handling.
21533 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21535         * config/nds32/nds32.h: Clean up obsolete macros.
21537 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21539         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21540         Add enum values for particular instructions.
21541         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21542         particular intrinsic functions.
21543         * config/nds32/nds32-intrinsic.md: Likewise.
21544         * config/nds32/nds32_intrinsic.h: Likewise.
21545         * config/nds32/nds32.h (nds32_builtins): Likewise.
21546         * config/nds32/nds32.md (type): Add pbsad and pbsada.
21547         (btst, ave): New patterns for particular instructions.
21549 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21551         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21552         Add enum values for atomic load/store and memory sync.
21553         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
21554         and memory sync.
21555         * config/nds32/nds32-intrinsic.md: Likewise.
21556         * config/nds32/nds32_intrinsic.h: Likewise.
21557         * config/nds32/nds32.h (nds32_builtins): Likewise.
21559 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21561         PR tree-optimization/85257
21562         * fold-const.c (native_encode_vector): If not all elts could fit
21563         and off is -1, return 0 rather than offset.
21564         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
21565         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
21566         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
21567         adjust buffer in native_interpret_expr call.
21569 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21571         * config/nds32/constants.md (unspec_volatile_element): Add cache
21572         control enum values.
21573         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
21574         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
21575         * config/nds32/nds32.c (nds32_cctl_names): New.
21576         (nds32_print_operand): Handle cache control register names.
21577         * config/nds32/nds32.h (nds32_builtins): New enum values.
21578         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
21579         macros.
21580         * config/nds32/nds32.md (type): Add mmu.
21581         * config/nds32/pipelines.md (simple_insn): Add mmu.
21583 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21585         * config/nds32/nds32.md (type): Remove call.
21586         * config/nds32/pipelines.md (simple_insn): Likewise.
21588 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21590         * config/nds32/constants.md (unspec_volatile_element): Add
21591         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
21592         UNSPEC_VOLATILE_FMFCFG.
21593         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
21594         description for fmfcfg and fmfcsr.
21595         (bdesc_1arg): Add fmtcsr.
21596         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
21597         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
21598         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
21599         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
21600         unspec_fmfcfg): New patterns.
21601         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
21602         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
21603         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
21604         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
21605         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
21606         __nds32__fmfcfg): Define.
21608 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21610         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
21611         intrinsic register names.
21612         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
21613         intrinsic register enum values and macros.
21615 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21617         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
21618         for load/store addressing form.
21619         (nds32_print_operand_address): Likewise.
21621 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
21623         PR target/85196
21624         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
21625         based on LABEL_REF.  Remove useless assertion.
21626         (pic_address_needs_scratch): Fix formatting.
21627         (sparc_legitimize_pic_address): Minor tweaks.
21628         (sparc_delegitimize_address): Adjust assertion accordingly.
21629         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
21630         into symbolic_operand.
21631         (movsi_high_pic_label_ref): Likewise.
21632         (movsi_lo_sum_pic_label_ref): Likewise.
21633         (movdi_pic_label_ref): Likewise.
21634         (movdi_high_pic_label_ref): Likewise.
21635         (movdi_lo_sum_pic_label_ref): Likewise.
21637 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
21639         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
21640         custom LIB_SPEC setup.
21642 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
21643             Kito Cheng  <kito.cheng@gmail.com>
21645         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
21646         * config/riscv/freebsd.h: New.
21648 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21650         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
21651         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
21652         file.
21654 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21655             Kito Cheng  <kito.cheng@gmail.com>
21657         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
21658         nds32_output_call, nds32_symbol_binds_local_p): New functions.
21659         * config/nds32/nds32-protos.h (nds32_output_call,
21660         nds32_output_return): Declare.
21661         * config/nds32/nds32.md: Refine all the call and return patterns.
21663 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21665         PR debug/85252
21666         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
21667         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
21669         PR rtl-optimization/84872
21670         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
21671         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
21672         EDGE_CROSSING edge.
21674 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
21676         * expr.c (copy_blkmode_to_reg): Revert 254862.
21677         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
21679 2018-04-06  Richard Biener  <rguenther@suse.de>
21681         PR middle-end/85244
21682         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
21683         after seeing a component reference with an adjacent field.  Treat
21684         refs to arrays at struct end of external decls similar to
21685         refs to unconstrained commons.
21687 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21689         PR sanitizer/85213
21690         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
21691         look through SAVE_EXPRs with non-side-effects argument.  Adjust
21692         recursive calls.
21693         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
21694         save_p here.
21696 2018-04-06  Richard Biener  <rguenther@suse.de>
21698         PR middle-end/85180
21699         * alias.c (find_base_term): New wrapper around find_base_term
21700         unwinding CSELIB_VAL_PTR changes.
21701         (find_base_term): Do not restore CSELIB_VAL_PTR during the
21702         recursion.
21704 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21706         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
21707         instructions.
21708         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
21709         constant definitions.
21710         ("nop"): lr 0,0 -> nopr r0
21711         ("nop_lr0", "nop_lr1"): New insn definitions.
21713 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21715         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
21716         NDS32_V3PUSH_AVAILABLE_P macro.
21718 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
21719             Chung-Ju Wu  <jasonwucj@gmail.com>
21721         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
21722         (nds32*-*-*): Add float and fpu_config into supported_defaults.
21723         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
21724         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
21725         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
21726         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
21727         * config/nds32/constraints.md: New constraints and checking for hard
21728         float configuration.
21729         * config/nds32/iterators.md: New mode iterator and attribute for hard
21730         float configuration.
21731         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
21732         patterns.
21733         * config/nds32/nds32-fpu.md: New file.
21734         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
21735         deal with hard float code generation.
21736         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
21737         ARCH_V3S.
21738         (abi_type, float_reg_number): New enum type.
21739         * config/nds32/nds32-predicates.c: New predicates for hard float.
21740         * config/nds32/nds32-protos.h: Declare functions for hard float.
21741         * config/nds32/nds32.c: Implementation for hard float configuration.
21742         * config/nds32/nds32.h: Definitions for hard float configuration.
21743         * config/nds32/nds32.md: Include hard float machine description and
21744         modify patterns for hard float configuration.
21745         * config/nds32/nds32.opt: New options for hard float configuration.
21746         * config/nds32/predicates.md: New predicates for hard float
21747         configuration.
21749 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
21751         * common/config/nds32/nds32-common.c
21752         (nds32_option_optimization_table): Enable -mreleax-hint by default.
21754 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
21756         PR middle-end/85195
21757         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
21758         CONSTRUCTOR_ELT (ctor, ...)->value.
21760 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
21762         PR target/85193
21763         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
21765 2018-04-05  Tom de Vries  <tom@codesourcery.com>
21767         PR target/85204
21768         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
21769         cond jump.
21771 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21772             Kito Cheng  <kito.cheng@gmail.com>
21774         * config/nds32/constraints.md (U33): Fine-tune checking condition.
21775         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
21776         * config/nds32/nds32.h (nds32_16bit_address_type): Add
21777         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
21779 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21780             Kito Cheng  <kito.cheng@gmail.com>
21782         * config/nds32/constraints.md (Ufe): New memory constraint.
21783         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
21784         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
21785         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
21786         operands.
21787         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
21788         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
21790 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21792         * config/nds32/nds32.md: Use optimize_size in the condition for
21793         alu-shift instructions.
21795 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21797         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
21799 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21801         * config/nds32/nds32.md (negsi2): Refine pattern.
21803 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21804             Chung-Ju Wu  <jasonwucj@gmail.com>
21806         * config/nds32/iterators.md (shift_rotate): New code iterator.
21807         (shift): New code attribute.
21808         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
21809         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
21810         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
21811         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
21812         bit-wise operations.
21813         (andsi3, *andsi3): Ditto.
21814         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
21815         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
21816         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
21817         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
21818         nds32_ior_operand, nds32_xor_operand): New predicates.
21820 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21822         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
21823         (addsi3, subsi3): ... this.
21825 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21827         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
21829 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21831         * config/nds32/nds32.md: Adjust indention.
21833 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21835         * config/nds32/nds32.md (feature): New attribute.
21837 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21839         * config/nds32/nds32.md (subtype): New attribute.
21841 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21843         PR target/85203
21844         * config/arm/arm-builtins.c (arm_expand_builtin): Change
21845         expansion to perform a bitwise AND of the argument followed by a
21846         boolean negation of the result.
21848 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
21850         PR rtl-optimization/84878
21851         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
21852         the basic block.  Assert the use reference is not artificial and that
21853         it has an associated insn.
21855 2018-04-04  Michael Matz  <matz@suse.de>
21857         * builtins.c (compute_objsize): Pass correct operand
21858         to array_at_struct_end_p.
21860 2018-04-04  Richard Biener  <rguenther@suse.de>
21862         PR lto/85176
21863         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
21864         from contexts for DINFO_LEVEL_TERSE and below.
21866 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21868         * config/nds32/nds32-doubleword.md (move_<mode>): Require
21869         resiter_operand condition.
21870         * config/nds32/nds32.md (*move<mode>): Ditto.
21872 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21873             Monk Chiang  <sh.chiang04@gmail.com>
21875         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
21877 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21879         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
21881 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21882             Kito Cheng  <kito.cheng@gmail.com>
21884         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
21885         nds32_cond_code_str, output_cond_branch,
21886         output_cond_branch_compare_zero, nds32_expand_cbranch,
21887         nds32_expand_cstore, nds32_expand_movcc,
21888         nds32_output_cbranchsi4_equality_zero,
21889         nds32_output_cbranchsi4_equality_reg,
21890         nds32_output_cbranchsi4_equality_reg_or_const_int,
21891         nds32_output_cbranchsi4_greater_less_zero: New functions.
21892         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
21893         nds32_expand_cstore, nds32_expand_movcc,
21894         nds32_output_cbranchsi4_equality_zero,
21895         nds32_output_cbranchsi4_equality_reg,
21896         nds32_output_cbranchsi4_equality_reg_or_const_int,
21897         nds32_output_cbranchsi4_greater_less_zero): Declare.
21898         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
21899         nds32_rimm11s_operand): New predicates.
21900         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
21901         * config/nds32/nds32.md: Rewrite all the branch and conditional move
21902         patterns.
21904 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21906         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
21907         * config/nds32/nds32.md: Ditto.
21908         * config/nds32/pipelines.md: Ditto.
21910 2018-04-04  Richard Biener  <rguenther@suse.de>
21912         PR tree-optimization/85168
21913         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
21914         propagating abnormals.
21916 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21918         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
21920 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21921             Kito Cheng  <kito.cheng@gmail.com>
21923         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
21924         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
21925         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
21926         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
21927         * config/nds32/nds32.md (sibcall_internal): New.
21928         (sibcall_register): Remove.
21929         (sibcall_immediate): Remove.
21930         (sibcall_value_internal): New.
21931         (sibcall_value_register): Remove.
21932         (sibcall_value_immediate): Remove.
21933         * config/nds32/predicates.md (nds32_general_register_operand): New.
21934         (nds32_call_address_operand): New.
21936 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21938         PR rtl-optimization/85167
21939         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
21940         bb_defs if *split_p, instead preinitialize it to NULL.
21942         PR tree-optimization/85156
21943         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
21944         evaluating the argument multiple times.
21946 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
21948         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
21949         than vector.
21950         (_mm_cvtpd_ps): Likewise.
21951         (_mm_cvttpd_epi32): Likewise.
21952         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
21953         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
21954         vector, pixel, and bool following altivec.h include.
21956 2018-04-03  Martin Sebor  <msebor@redhat.com>
21958         * doc/extend.texi (Common Function Attributes): Clarify.
21959         (const attribute): Likewise.
21960         (pure attribute): Likewise.
21962 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21964         PR target/85169
21965         * config/i386/i386.c (ix86_expand_vector_set): Use
21966         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
21968 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
21970         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
21971         instructions when changing rounding bits to preserve precision bits
21972         in the x87 control word.
21974 2018-04-03  Martin Liska  <mliska@suse.cz>
21976         PR tree-optimization/82491
21977         * rtl.h (strip_offset_and_add): Replace += suboffset with
21978         poly_uint64 () + suboffset.
21980 2018-03-29  Martin Liska  <mliska@suse.cz>
21981             Martin Jambor  <mjambor@suse.cz>
21983         PR ipa/84947
21984         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
21985         param_type is not an integral or pointer type.
21987 2018-04-03  Richard Biener  <rguenther@suse.de>
21989         * sese.h (recompute_all_dominators): Remove.
21991 2018-04-02  Martin Sebor  <msebor@redhat.com>
21993         * doc/invoke.texi (-Wrestrict): Fix typos.
21995 2018-04-02  Jim Wilson  <jimw@sifive.com>
21997         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
21998         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
21999         (<optab>di3, <optab>si3_extend): Likewise.
22000         (<optab>si3_mask, <optab>si3_mask_1): New.
22001         (<optab>di3_mask, <optab>di3_mask_1): New.
22002         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22003         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22004         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22006 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22008         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22009         example.
22011 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22013         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22014         (nds32_canonicalize_comparison): New function.
22016 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22017             Kito Cheng  <kito.cheng@gmail.com>
22018             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22020         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22021         * config/nds32/constants.md (unspec_volatile_element): Add
22022         UNSPEC_VOLATILE_RELAX_GROUP.
22023         * config/nds32/nds32-relax-opt.c: New file.
22024         * config/nds32/nds32-predicates.c
22025         (nds32_symbol_load_store_p): New function.
22026         * config/nds32/nds32-protos.h
22027         (nds32_symbol_load_store_p): Declare function.
22028         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22029         * config/nds32/nds32.c
22030         (nds32_register_pass): New function to register pass.
22031         (nds32_register_passes): New function to register passes.
22032         * config/nds32/nds32.md (relax_group): New pattern.
22033         * config/nds32/nds32.opt (mrelax-hint): New option.
22034         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22036 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22038         * config/nds32/t-nds32: Modify files dependency.
22040 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22042         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22043         (PROFILE_HOOK): Define its implementation.
22045 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22047         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22048         type and 32-bit size.
22050 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22052         PR middle-end/85090
22053         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22054         (V_128_256): New mode iterator.
22055         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22056         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22057         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22058         of V.
22059         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22060         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22062 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22064         PR target/83315
22065         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22066         NaN inputs correctly.
22068 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22070         PR target/80546
22071         * config/rs6000/vsx.md (??r): New mode attribute.
22072         (*vsx_mov<mode>_64bit): Use it.
22073         (*vsx_mov<mode>_32bit): Likewise.
22075 2018-03-30  Martin Sebor  <msebor@redhat.com>
22077         PR tree-optimization/84818
22078         * builtins.c (check_access): Use warning_n.
22080 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22082         PR target/83822
22083         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22084         condition.
22085         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22086         condition.
22088 2018-03-30  Julia Koval  <julia.koval@intel.com>
22090         PR target/84413
22091         * x86-tune.def (movx, partial_reg_dependency): Enable for
22092         m_SKYLAKE_AVX512.
22094 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22096         PR inline-asm/84985
22097         * lra-constraints.c (process_alt_operands): Move setting
22098         this_alternative_matches below.
22100 2018-03-29  Martin Liska  <mliska@suse.cz>
22102         PR lto/84995.
22103         * doc/invoke.texi: Document how LTO works with debug info.
22104         Describe auto-load support of binutils.  Mention 'x86-64'
22105         as valid option value of -march option.
22107 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22109         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22111         PR c/85094
22112         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22113         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22114         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22115         checking.
22117 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22119         PR target/84912
22120         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22121         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22122         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22123         for RS6000_BTM_POWERPC64.
22124         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22125         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22126         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22127         definition.
22128         (DIVDE): Use it.
22129         (DIVDEU): Likewise.
22131 2018-03-28  Carl Love  <cel@us.ibm.com>
22133         Revert
22134         2017-09-27  Carl Love  <cel@us.ibm.com>
22136         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22137         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22138         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22139         fctiw instruction.
22141 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22143         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22144         instead of __vector bool.
22145         (_mm_max_pu8): Likewise.
22146         (_mm_min_pi16): Likewise.
22148 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22150         PR target/84912
22151         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22152         (DIVWEUO): Likewise.
22153         (DIVDEO): Likewise.
22154         (DIVDEUO): Likewise.
22155         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22156         DIVWEUO and DIVDEUO.
22157         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22158         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22159         (div_extend): Likewise.
22160         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22161         builtin function.
22162         (__builtin_divweuo): Likewise.
22163         (__builtin_divdeo): Likewise.
22164         (__builtin_divdeuo): Likewise.
22166 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22168         PR target/85095
22169         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22170         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22172         PR tree-optimization/82004
22173         * gimple-match-head.c (optimize_pow_to_exp): New function.
22174         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22175         Don't fold to exp if optimize_pow_to_exp is false.
22177 2018-03-28  Martin Liska  <mliska@suse.cz>
22179         PR other/84819
22180         * calls.c (initialize_argument_information): Fix trailing space.
22181         * common.opt: Fix typo and provide better explanation for
22182         -fsanitize-coverage option.
22183         * config/i386/i386.opt: Fix typo.
22185 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22186             Martin Liska  <mliska@suse.cz>
22188         PR sanitizer/85081
22189         * gimplify.c (asan_poison_variable): Don't do the check for
22190         gimplify_omp_ctxp here.
22191         (gimplify_decl_expr): Do it here.
22192         (gimplify_target_expr): Likewise.
22194 2018-03-28  Martin Liska  <mliska@suse.cz>
22196         PR target/84988
22197         * config/i386/i386.c (ix86_function_arg_advance): Do not call
22198         chkp_type_bounds_count if MPX is not enabled.
22200 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
22202         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22204 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
22206         PR target/84914
22207         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22208         function to create the function decl for complex long double
22209         multiply and divide for -mabi=ieeelongdouble.
22210         (init_float128_ieee): Call it.
22212 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22214         PR target/85044
22215         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22216         -fcf-protection=branch -mibt.
22217         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22219 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22221         PR target/81863
22222         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22223         arm_word_relocations.
22225 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22227         PR target/85056
22228         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22229         extern array declarations.
22231 2018-03-27  Richard Biener  <rguenther@suse.de>
22233         PR middle-end/84067
22234         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22235         explicit single_use checks.
22237 2018-03-27  Richard Biener  <rguenther@suse.de>
22239         PR tree-optimization/85082
22240         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22241         Valueize the VUSE.
22243 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22245         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22246         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22247         Turn on fasynchronous-unwind-tables and funwind-tables.
22249 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22251         PR target/85073
22252         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22253         (*bmi_blsr_<mode>_ccz): Ditto.
22255 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22257         PR tree-optimization/85063
22258         * omp-general.c (offloading_function_p): New function.  Factor out
22259         of ...
22260         * omp-offload.c (pass_omp_target_link::gate): ... here.
22261         * omp-general.h (offloading_function_p): Declare.
22262         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22263         with attribute omp declare target for offloading functions.
22265 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
22267         PR tree-optimization/84005
22268         * tree-data-ref.h (get_base_for_alignment): Declare.
22269         * tree-data-ref.c (get_base_for_alignment_1): New function.
22270         (get_base_for_alignment): Likewise.
22271         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22272         get_base_for_alignment to find a suitable base object, instead
22273         of always using drb->base_address.
22275 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22277         PR inline-asm/85022
22278         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22279         known size by default.
22281 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
22283         PR inline-asm/85030
22284         * lra-constraints.c (process_alt_operands): Don't match BLKmode
22285         and non BLKmode operands.
22287 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22289         PR target/85026
22290         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22291         Clean up attributes.
22293 2018-03-23  Richard Biener  <rguenther@suse.de>
22295         PR debug/85020
22296         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22297         we are going to emit early debug for LTO.
22299 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22301         PR inline-asm/85034
22302         * function.c (match_asm_constraints_1): Don't optimize if input
22303         doesn't satisfy general_operand predicate for output's mode.
22305         PR inline-asm/85022
22306         * alias.c (write_dependence_p): Don't require for x_canonicalized
22307         non-VOIDmode if x has VOIDmode.
22309         PR sanitizer/85029
22310         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22311         just don't try to optimize it rather than assert it never happens.
22313 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22315         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22316         macro expansions for definition of ST_INTERNAL_<mode> and
22317         LD_INTERNAL_<mode> builtins.
22318         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22319         Remove prototype.
22320         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22321         function.
22322         (altivec_expand_st_builtin): Likewise.
22323         (altivec_expand_builtin): Remove calls to deleted functions.
22324         (rs6000_address_for_altivec): Delete this function.
22325         * config/rs6000/vector.md: Remove expands for
22326         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22328 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
22330         PR target/84826
22331         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22332         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22333         re-computing once computed.
22334         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22335         (arm_init_machine_status): Initialize
22336         machine->static_chain_stack_bytes.
22338 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22340         PR target/84760
22341         * doc/extend.texi: Add four new prototypes for vec_ld.
22342         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22343         definitions for more logical presentation.
22344         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22345         entries for V1TI variants of __builtin_altivec_ld builtin.
22346         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22347         handling of V1TI variant of LVX icode pattern.
22348         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22349         (rs6000_gimple_fold_builtin): Likewise.
22350         (altivec_init_builtins): Add code to define
22351         __builtin_altivec_lvx_v1ti function.
22353 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22355         PR inline-asm/84941
22356         * function.c (match_asm_constraints_1): Don't do the optimization
22357         if input isn't a REG, SUBREG, MEM or constant.
22359 2018-03-22  Tom de Vries  <tom@codesourcery.com>
22361         PR tree-optimization/84956
22362         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22363         bb_has_abnormal_pred.
22365 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22367         PR sanitizer/85018
22368         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22369         DECL_INITIAL (decl) to decl at the end.
22370         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22371         adjust the comment.
22373 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22375         * doc/extend.texi (__builtin_tgmath): Document when complex
22376         integer types are treated as _Complex _Float64.
22378 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22380         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22382 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22384         PR tree-optimization/84960
22385         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22386         if it is ENTRY block, move them into single succ of ENTRY in that case.
22388 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
22390         PR tree-optimization/84811
22391         * poly-int.h (poly_span_traits): Remove the T3 parameter and
22392         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22393         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22394         (known_subrange_p): Update accordingly.  Cast each value involved
22395         in the size comparison, rather than casting the result of the
22396         subtraction.
22398 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22400         PR tree-optimization/84982
22401         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22402         by flipping the least significant bit rather than all bits from
22403         bitpos to bitpos + bitsize - 1.
22405 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22407         * doc/extend.texi (Deprecated Features): Remove mention of
22408         long-deleted deprecations.
22410 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22412         PR jit/84288
22413         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22414         * configure: Regenerate.
22416 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22418         PR tree-optimization/83126
22419         * tree-parloops.c (num_phis): New function.
22420         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22422 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22424         * doc/extend.texi (Deprecated Features): Update deprecated flags,
22425         mention anon-struct/union members and trailing attributes.
22427 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
22429         PR tree-optimization/84969
22430         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22431         builtin memset partitions if they set different rhs values.
22433 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22435         PR rtl-optimization/84989
22436         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22437         VEC_DUPLICATE with scalar result mode.
22439 2018-03-21  Martin Liska  <mliska@suse.cz>
22441         PR ipa/84963
22442         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22443         not intended return statement.
22445 2018-03-21  Martin Liska  <mliska@suse.cz>
22447         PR target/84988
22448         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22449         (chkp_find_bound_slots_1): Limit number of iterations.
22451 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
22453         PR target/84838
22454         * Minor grammar fixes for x86 options.
22456 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22458         PR debug/84875
22459         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22460         holding REG_CFA_RESTORE notes, instead turn them into a USE.
22462 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
22464         PR target/83789
22465         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22466         (altivec_lvx_<mode>_1op): Likewise.
22467         (altivec_stvx_<mode>_2op): Likewise.
22468         (altivec_stvx_<mode>_1op): Likewise.
22469         (altivec_lvx_<VM2:mode>): New define_expand.
22470         (altivec_stvx_<VM2:mode>): Likewise.
22471         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22472         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22473         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22474         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22475         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22476         (rs6000_gen_lvx): Likewise.
22477         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22478         (altivec_expand_stv_builtin): Likewise.
22479         (altivec_expand_builtin): Likewise.
22480         * config/rs6000/vector.md: Likewise.
22482 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22484         PR target/82518
22485         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22486         BYTES_BIG_ENDIAN.
22488 2018-03-20  Richard Biener  <rguenther@suse.de>
22490         PR target/84986
22491         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22492         sign-conversions as zero, fall back to standard scalar_stmt
22493         cost for the rest.
22495 2018-03-20  Martin Liska  <mliska@suse.cz>
22497         PR ipa/84825
22498         * predict.c (rebuild_frequencies): Handle case when we have
22499         PROFILE_ABSENT, but flag_guess_branch_prob is false.
22501 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22503         PR target/84990
22504         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22505         flag_section_anchors.
22506         * varasm.c (use_blocks_for_decl_p): Remove hack for
22507         dw2_force_const_mem.
22509         PR target/84845
22510         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22511         to ...
22512         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
22513         be created, use lowpart_subreg of operands[0] rather than operands[0]
22514         itself.
22515         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22516         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22517         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22518         and n constraint instead of aarch64_shift_imm_di and Usd.
22519         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22520         (*aarch64_<optab>_reg_minus<mode>3): ... this.
22522 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
22524         PR target/82989
22525         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
22526         to favor GPR over NEON registers.
22527         (<shift>di3_neon): Likewise.
22529 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22531         PR target/84952
22532         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
22533         (nvptx_process_pars): Emit bar.sync asap and alap.
22535 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22537         PR target/84954
22538         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
22539         seen_label if seen_label is already set.
22541 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22543         PR target/84945
22544         * config/i386/i386.c (fold_builtin_cpu): For features above 31
22545         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
22546         Use 1U instead of 1.  Formatting fixes.
22548         PR c/84953
22549         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
22550         instead of TREE_TYPE (s1) for the return value.
22552 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
22554         PR tree-optimization/84946
22555         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
22556         bitsize + bitsize in poly_uint64 rather than poly_int64.
22558         PR sanitizer/78651
22559         * dwarf2asm.c: Include fold-const.c.
22560         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
22561         of decl rather than decl itself.
22563         PR rtl-optimization/84643
22564         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
22566 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
22568         PR sanitizer/78651
22569         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
22570         calling assemble_variable.
22572 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
22574         PR target/81647
22575         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
22576         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
22578 2018-03-19  Jim Wilson  <jimw@sifive.com>
22580         PR bootstrap/84856
22581         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
22582         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
22583         (riscv_first_stack_step): Likewise.
22584         (riscv_option_override): Use STACK_BOUNDARY instead of
22585         MIN_STACK_BOUNDARY.
22586         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
22587         MIN_STACK_BOUNDARY.
22588         (BIGGEST_ALIGNMENT): Set to 128.
22589         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
22590         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22591         STACK_BOUNDARY.
22593 2018-03-19  Richard Biener  <rguenther@suse.de>
22595         PR tree-optimization/84933
22596         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
22597         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
22599 2018-03-19  Richard Biener  <rguenther@suse.de>
22601         PR tree-optimization/84859
22602         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
22603         (cond_if_else_store_replacement): Perform sinking operation on
22604         single-store BBs regardless of MAX_STORES_TO_SINK setting.
22605         Generalize what a BB with a single eligible store is.
22607 2018-03-19  Richard Biener  <rguenther@suse.de>
22609         PR tree-optimization/84929
22610         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
22611         chrec_is_positive against non-chrec arg.
22613 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
22615         PR target/84711
22616         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
22618 2018-03-18  Martin Liska  <mliska@suse.cz>
22620         PR rtl-optimization/84635
22621         * regrename.c (build_def_use): Use matches_mode only when
22622         matches >= 0.
22624 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
22626         PR tree-optimization/84913
22627         * tree-vect-loop.c (vectorizable_reduction): Don't try to
22628         vectorize chains of COND_EXPRs.
22630 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22632         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
22634 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22636         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
22638 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22640         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
22642 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22643             Kito Cheng  <kito.cheng@gmail.com>
22645         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
22646         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
22647         (nds32_adjust_reg_alloc_order): New function.
22648         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
22650 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
22652         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
22653         nds32_print_operand, nds32_print_operand_address): Use
22654         HOST_WIDE_INT_PRINT_DEC instead.
22656 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22658         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
22660 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
22662         PR target/84902
22663         * config/i386/i386.c (initial_ix86_tune_features,
22664         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
22665         unsigned long long.
22666         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
22667         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
22668         rather than 1u << ix86_tune.  Formatting fix.
22669         (ix86_option_override_internal): Change ix86_arch_mask from
22670         unsigned int to unsigned HOST_WIDE_INT, initialize to
22671         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
22672         (ix86_function_specific_restore): Likewise.
22674 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22676         PR target/84899
22677         * postreload.c (reload_combine_recognize_pattern): Perform
22678         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
22679         truncate_int_for_mode the result for the destination's mode.
22681         PR c/84909
22682         * hsa-gen.c (mem_type_for_type): Fix comment typo.
22683         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
22684         Likewise.
22685         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22686         Likewise.
22688 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
22690         PR target/84876
22691         * lra-assigns.c (lra_split_hard_reg_for): Don't use
22692         regno_allocno_class_array and sorted_pseudos.
22693         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
22694         insns where regno is used.
22696 2018-03-16  Martin Liska  <mliska@suse.cz>
22698         PR ipa/84833
22699         * multiple_target.c (create_dispatcher_calls): Redirect
22700         reference in the symbol table.
22702 2018-03-16  Martin Liska  <mliska@suse.cz>
22704         PR ipa/84722
22705         * multiple_target.c (create_dispatcher_calls): Redirect also
22706         an alias.
22708 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22710         PR c++/79937
22711         PR c++/82410
22712         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
22713         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
22714         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
22716 2018-03-16  Julia Koval  <julia.koval@intel.com>
22718         * doc/invoke.texi (Skylake Server): Add CLWB.
22719         Cannonlake): Remove CLWB.
22721 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22723         PR tree-optimization/84841
22724         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
22725         1 << 3.
22726         (FLOAT_ONE_CONST_TYPE): Define.
22727         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
22728         (sort_by_operand_rank): Put entries with higher constant_type last
22729         rather than first to match comments.
22731 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
22733         * config/nios2/nios2.md (movsi_internal): Fix thinko in
22734         split predicate.
22736 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22738         PR c++/79085
22739         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
22740         check and use address of target always.
22742 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
22744         PR target/84574
22745         * config/i386/i386.c (indirect_thunk_needed): Update comments.
22746         (indirect_thunk_bnd_needed): Likewise.
22747         (indirect_thunks_used): Likewise.
22748         (indirect_thunks_bnd_used): Likewise.
22749         (indirect_return_needed): New.
22750         (indirect_return_bnd_needed): Likewise.
22751         (output_indirect_thunk_function): Add a bool argument for
22752         function return.
22753         (output_indirect_thunk_function): Don't generate alias for
22754         function return thunk.
22755         (ix86_code_end): Call output_indirect_thunk_function to generate
22756         function return thunks.
22757         (ix86_output_function_return): Set indirect_return_bnd_needed
22758         and indirect_return_needed instead of indirect_thunk_bnd_needed
22759         and indirect_thunk_needed.
22761 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
22763         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
22764         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
22765         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
22767 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
22768             Paul Hua <paul.hua.gm@gmail.com>
22770         PR c/84852
22771         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
22773 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
22775         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
22776         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
22777         resp. SFmode cases.
22779 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
22781         PR target/84711
22782         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
22783         instead of GET_MODE_SIZE when comparing Units.
22785 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
22787         PR target/68256
22788         * varasm.c (hash_section): Return an unchangeble hash value
22789         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
22790         Return !aarch64_can_use_per_function_literal_pools_p ().
22792 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22794         PR target/84860
22795         * optabs.c (emit_conditional_move): Pass address of cmode's copy
22796         rather than address of cmode as last argument to prepare_cmp_insn.
22798 2018-03-15  Julia Koval  <julia.koval@intel.com>
22800         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
22801         F_AVX512VNNI, F_AVX512BITALG): New.
22803 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
22805         PR target/83451
22806         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
22807         insn for floating-point loads and stores.
22809 2018-03-14  Carl Love  <cel@us.ibm.com>
22811         * config/rs6000/rs6000-c.c: Add macro definitions for
22812         ALTIVEC_BUILTIN_VEC_PERMXOR.
22813         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
22814         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
22815         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
22816         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
22817         UNSPEC_VPERMXOR.
22818         * config/doc/extend.texi: Add prototypes for vec_permxor.
22820 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
22822         PR c/84852
22823         * diagnostic-show-locus.c (class layout_point): Convert m_line
22824         from int to linenum_type.
22825         (line_span::comparator): Use linenum "compare" function when
22826         comparing line numbers.
22827         (test_line_span): New function.
22828         (layout_range::contains_point): Convert param "row" from int to
22829         linenum_type.
22830         (layout_range::intersects_line_p): Likewise.
22831         (layout::will_show_line_p): Likewise.
22832         (layout::print_source_line): Likewise.
22833         (layout::should_print_annotation_line_p): Likewise.
22834         (layout::print_annotation_line): Likewise.
22835         (layout::print_leading_fixits): Likewise.
22836         (layout::annotation_line_showed_range_p): Likewise.
22837         (struct line_corrections): Likewise for field m_row.
22838         (line_corrections::line_corrections): Likewise for param "row".
22839         (layout::print_trailing_fixits): Likewise.
22840         (layout::get_state_at_point): Likewise.
22841         (layout::get_x_bound_for_row): Likewise.
22842         (layout::print_line): Likewise.
22843         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
22844         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
22845         * input.c (selftest::test_linenum_comparisons): New function.
22846         (selftest::input_c_tests): Call it.
22847         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
22848         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
22849         * selftest.h (ASSERT_GT): New macro.
22850         (ASSERT_GT_AT): New macro.
22851         (ASSERT_LT): New macro.
22852         (ASSERT_LT_AT): New macro.
22854 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
22856         PR rtl-optimization/84780
22857         * combine.c (distribute_links): Don't make a link based on pc_rtx.
22859 2018-03-14  Martin Liska  <mliska@suse.cz>
22861         * tree.c (record_node_allocation_statistics): Use
22862         get_stats_node_kind.
22863         (get_stats_node_kind): New function extracted from
22864         record_node_allocation_statistics.
22865         (free_node): Use get_stats_node_kind.
22867 2018-03-14  Richard Biener  <rguenther@suse.de>
22869         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
22870         that the value-set of ANTIC_IN doesn't grow.
22872         Revert
22873         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22874         member.
22875         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22876         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22878 2018-03-14  Julia Koval  <julia.koval@intel.com>
22880         * config.gcc (icelake-client, icelake-server): New.
22881         (icelake): Remove.
22882         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
22883         (initial_ix86_arch_features): Ditto.
22884         (PTA_SKYLAKE): Add SGX.
22885         (PTA_ICELAKE): Remove.
22886         (PTA_ICELAKE_CLIENT): New.
22887         (PTA_ICELAKE_SERVER): New.
22888         (ix86_option_override_internal): Split up icelake on icelake client and
22889         icelake server.
22890         (get_builtin_code_for_version): Ditto.
22891         (fold_builtin_cpu): Ditto.
22892         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
22893         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
22894         * config/i386/i386.h (processor_type): Ditto.
22895         * doc/invoke.texi: Ditto.
22897 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
22899         PR sanitizer/83392
22900         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
22901         INTEGER_CST offset, add it together with bitpos / 8 and
22902         sign extend based on POINTER_SIZE.
22904         PR target/84844
22905         Revert
22906         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
22908         PR target/78090
22909         * config/i386/constraints.md (Yc): New register constraint.
22910         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
22911         Use Yc constraint for alternative 2 of operand 0.  Remove
22912         preferred_for_speed attribute.
22914 2018-03-14  Richard Biener  <rguenther@suse.de>
22916         PR tree-optimization/84830
22917         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
22918         with the old one to avoid oscillations.
22920 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
22922         PR target/83712
22923         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
22924         pseudos.
22925         (assign_by_spills): Return a flag of reload assignment failure.
22926         Do not process the reload assignment failures.  Do not spill other
22927         reload pseudos if they has the same reg class.  Update n if
22928         necessary.
22929         (lra_assign): Add a return arg.  Set up from the result of
22930         assign_by_spills call.
22931         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22932         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22933         usage_insns if it is not NULL.
22934         (spill_hard_reg_in_range): New function.
22935         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22936         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22937         function prototypes.
22938         (lra_assign): Change prototype.
22939         * lra.c (lra): Add code to deal with fails by splitting hard reg
22940         live ranges.
22942 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
22944         * config/riscv/riscv.opt (mrelax): New option.
22945         * config/riscv/riscv.c (riscv_file_start): Emit ".option
22946         "norelax" when riscv_mrelax is disabled.
22947         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
22949 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22951         PR target/84743
22952         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
22953         reassociation for int modes.
22955 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22957         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
22958         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
22959         for big-endian.
22960         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
22961         * config/aarch64/aarch64-sve.md
22962         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
22963         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
22964         (*extend<mode><Vwide>2): Rename to...
22965         (aarch64_sve_extend<mode><Vwide>2): ...this.
22966         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
22967         renaming the old pattern to...
22968         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
22969         unsigned packs.
22970         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
22971         define_expand, renaming the old pattern to...
22972         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
22973         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
22974         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
22975         account when deciding which SVE instruction the optab should use.
22976         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
22978 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22980         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
22981         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
22982         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
22983         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
22984         (tlsdesc_small_<mode>): Turn a define_expand and use
22985         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
22986         (tlsdesc_small_advsimd_<mode>): ...this.
22987         (tlsdesc_small_sve_<mode>): New pattern.
22989 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22991         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
22992         (UNSPEC_UMUL_HIGHPART): New constants.
22993         (MUL_HIGHPART): New int iteraor.
22994         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
22995         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
22996         define_expand.
22997         (*<su>mul<mode>3_highpart): New define_insn.
22999 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23001         PR lto/84805
23002         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23003         incomplete types.
23005 2018-03-13  Martin Liska  <mliska@suse.cz>
23007         PR ipa/84658.
23008         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23009         vector.
23010         (sem_item_optimizer::~sem_item_optimizer): Release it.
23011         (sem_item_optimizer::merge_classes): Register variable aliases.
23012         (sem_item_optimizer::fixup_pt_set): New function.
23013         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23014         * ipa-icf.h: Declare new variables and functions.
23016 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23018         PR middle-end/84834
23019         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23020         integer_pow2p@2 and test integer_pow2p in condition.
23021         (A < 0 ? C : 0): Similarly for @1.
23023         PR middle-end/84831
23024         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23025         characters starting at p contain '\0' character, don't look beyond
23026         that.
23028         PR target/84827
23029         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23030         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23032         PR target/84828
23033         * reg-stack.c (change_stack): Change update_end var from int to
23034         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23035         also call set_block_for_insn on the newly added insns and rescan.
23037         PR target/84786
23038         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23039         on the last operand.
23041         PR c++/84704
23042         * tree.c (stabilize_reference_1): Return save_expr (e) for
23043         STATEMENT_LIST even if it doesn't have side-effects.
23045 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23047         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23049 2018-03-12  Renlin Li  <renlin.li@arm.com>
23051         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23052         aarch64_output_scalar_simd_mov_immediate.
23054 2018-03-12  Martin Sebor  <msebor@redhat.com>
23056         PR tree-optimization/83456
23057         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23058         for perfectly overlapping calls to memcpy.
23059         (gimple_fold_builtin_memory_chk): Same.
23060         (gimple_fold_builtin_strcpy): Handle no-warning.
23061         (gimple_fold_builtin_stxcpy_chk): Same.
23062         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23064 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23066         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23067         parameter.  Use it for SFmode.
23068         (rs6000_function_arg_advance_1): Adjust.
23069         (rs6000_function_arg): Adjust.
23070         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23072 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23074         PR rtl-optimization/84169
23075         PR rtl-optimization/84780
23076         * combine.c (can_combine_p): Check for a 2-insn combination whether
23077         the destination register is used between the two insns, too.
23079 2018-03-12  Richard Biener  <rguenther@suse.de>
23081         PR tree-optimization/84803
23082         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23083         for refs DR analysis didn't process.
23085 2018-03-12  Richard Biener  <rguenther@suse.de>
23087         PR tree-optimization/84777
23088         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23089         force-vectorize loops ignore whether we are optimizing for size.
23091 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23093         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23094         (TARGET_MD_ASM_ADJUST): Define.
23096 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23097             Kito Cheng  <kito.cheng@gmail.com>
23098             Chung-Ju Wu  <jasonwucj@gmail.com>
23100         * config/nds32/nds32.c (nds32_compute_stack_frame,
23101         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23102         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23103         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23104         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23105         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23106         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23107         * config/nds32/nds32.md (prologue, epilogue): Use macro
23108         NDS32_V3PUSH_AVAILABLE_P to do checking.
23110 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23112         PR debug/58150
23113         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23114         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23115         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23116         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23117         addition of most attributes on !orig_type_die or the attribute not
23118         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23120 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23121             Chung-Ju Wu  <jasonwucj@gmail.com>
23123         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23124         __NDS32_VH__ macro.
23125         * config/nds32/nds32.opt (mvh): New option.
23127 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23128             Chung-Ju Wu  <jasonwucj@gmail.com>
23130         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23131         function.
23132         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23133         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23134         definition.
23136 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23137             Chung-Ju Wu  <jasonwucj@gmail.com>
23139         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23140         function.
23141         * config/nds32/nds32-multiple.md (strlensi): New pattern.
23142         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23144 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
23145             Kito Cheng  <kito.cheng@gmail.com>
23146             Chung-Ju Wu  <jasonwucj@gmail.com>
23148         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23149         UNSPEC_FFMISM and UNSPEC_FLMISM.
23150         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23151         for ffb, ffmism and flmism.
23152         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23153         (unspec_ffmism): Ditto.
23154         (unspec_flmism): Ditto.
23155         (nds32_expand_builtin_impl): Check if string extension is available.
23156         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23157         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23159 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
23161         Reverting patch:
23162         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23164         PR target/83712
23165         * lra-assigns.c (assign_by_spills): Return a flag of reload
23166         assignment failure.  Do not process the reload assignment
23167         failures.  Do not spill other reload pseudos if they has the same
23168         reg class.
23169         (lra_assign): Add a return arg.  Set up from the result of
23170         assign_by_spills call.
23171         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23172         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23173         usage_insns if it is not NULL.
23174         (spill_hard_reg_in_range): New function.
23175         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23176         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23177         function prototypes.
23178         (lra_assign): Change prototype.
23179         * lra.c (lra): Add code to deal with fails by splitting hard reg
23180         live ranges.
23182 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
23184         PR target/84807
23185         * config/i386/i386.opt: Replace Enforcment with Enforcement.
23187 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
23189         PR debug/84620
23190         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23191         (dw_val_node): Add val_symbolic_view.
23192         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23193         (symview_upper_bound): New.
23194         (new_line_info_table): Initialize symviews_since_reset.
23195         (dwarf2out_source_line): Count symviews_since_reset and set
23196         symview_upper_bound.
23197         (dw_val_equal_p): Handle symview.
23198         (add_AT_symview): New.
23199         (print_dw_val): Handle symview.
23200         (attr_checksum, attr_checksum_ordered): Likewise.
23201         (same_dw_val_p, size_of_die): Likewise.
23202         (value_format, output_die): Likewise.
23203         (add_high_low_attributes): Use add_AT_symview for entry_view.
23204         (dwarf2out_finish): Reset symview_upper_bound, clear
23205         zero_view_p.
23207 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
23209         PR target/83969
23210         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23211         Add strict argument and use it.
23212         (rs6000_split_multireg_move): Update for new strict argument.
23213         (mem_operand_gpr): Disallow all non-offsettable addresses.
23214         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23216 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23218         PR target/84772
23219         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23220         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23221         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23223         PR c++/84767
23224         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23225         decl, use remap_type if we want to use the type.
23227 2018-03-09  Martin Sebor  <msebor@redhat.com>
23229         PR tree-optimization/84526
23230         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23231         Remove dead code.
23232         (builtin_access::generic_overlap): Be prepared to handle non-array
23233         base objects.
23235 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23237         PR rtl-optimization/84682
23238         * lra-constraints.c (process_address_1): Check is_address flag
23239         for address constraints.
23240         (process_alt_operands): Likewise.
23241         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23242         preprocess_constraints.
23243         * recog.h (preprocess_constraints): Add oploc parameter.
23244         Adjust callers.
23245         * recog.c (preprocess_constraints): Test address_operand for
23246         CT_ADDRESS constraints.
23248 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23250         PR target/83712
23251         * lra-assigns.c (assign_by_spills): Return a flag of reload
23252         assignment failure.  Do not process the reload assignment
23253         failures.  Do not spill other reload pseudos if they has the same
23254         reg class.
23255         (lra_assign): Add a return arg.  Set up from the result of
23256         assign_by_spills call.
23257         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23258         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23259         usage_insns if it is not NULL.
23260         (spill_hard_reg_in_range): New function.
23261         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23262         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23263         function prototypes.
23264         (lra_assign): Change prototype.
23265         * lra.c (lra): Add code to deal with fails by splitting hard reg
23266         live ranges.
23268 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23270         PR target/83193
23271         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23272         Accept complain bool parameter.  Only emit errors if it is true.
23273         (arm_parse_cpu_option_name): Likewise.
23274         (arm_target_thumb_only): Adjust callers of the above.
23275         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23276         prototype to take a default true bool parameter.
23277         (arm_parse_arch_option_name): Likewise.
23279 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
23280             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
23282         PR jit/64089
23283         PR jit/84288
23284         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23285         * configure: Regenerate.
23286         * configure.ac ("linker --version-script option"): New.
23287         ("linker soname option"): New.
23289 2018-03-09  Richard Biener  <rguenther@suse.de>
23291         PR tree-optimization/84775
23292         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23293         immediate uses of predicate stmts and mark them modified.
23295         Revert
23296         PR tree-optimization/84178
23297         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23298         to caller.
23299         (version_loop_for_if_conversion): Delay update_ssa call.
23300         (tree_if_conversion): Delay update_ssa until after predicate
23301         insertion.
23303 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
23305         PR target/84763
23306         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23307         when the function accesses prior frames.
23309 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23311         PR debug/84456
23312         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23313         gen_llsym, otherwise call maybe_gen_llsym.
23315         PR inline-asm/84742
23316         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23317         has ',' character inside of it.
23319 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23321         PR target/84748
23322         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23323         as clobbering CC_REGNUM.
23325 2018-03-08  Richard Biener  <rguenther@suse.de>
23327         PR middle-end/84552
23328         * tree-scalar-evolution.c: Include tree-into-ssa.h.
23329         (follow_copies_to_constant): Do not follow SSA names registered
23330         for update.
23332 2018-03-08  Richard Biener  <rguenther@suse.de>
23334         PR tree-optimization/84178
23335         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23336         to caller.
23337         (version_loop_for_if_conversion): Delay update_ssa call.
23338         (tree_if_conversion): Delay update_ssa until after predicate
23339         insertion.
23341 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
23343         PR tree-optimization/84178
23344         * tree-if-conv.c (release_bb_predicate): Remove the
23345         the assertion that the stmts have NULL use_ops.
23346         Discard the statements, asserting that they haven't
23347         yet been added to a BB.
23349 2018-03-08  Richard Biener  <rguenther@suse.de>
23351         PR tree-optimization/84746
23352         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23353         (phi_translate): Pass in destination ANTIC_OUT set.
23354         (phi_translate_1): Likewise.  For a simplified result lookup
23355         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23356         (phi_translate_set): Adjust.
23357         (do_pre_regular_insertion): Likewise.
23358         (do_pre_partial_partial_insertion): Likewise.
23360 2018-03-08  Martin Liska  <mliska@suse.cz>
23362         PR gcov-profile/84735
23363         * doc/gcov.texi: Document usage of profile files.
23364         * gcov-io.h: Document changes in the format.
23366 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23368         PR debug/84404
23369         PR debug/84408
23370         * dwarf2out.c (struct dw_line_info_table): Update comments for
23371         view == -1.
23372         (FORCE_RESET_NEXT_VIEW): New.
23373         (FORCE_RESETTING_VIEW_P): New.
23374         (RESETTING_VIEW_P): Check for -1 too.
23375         (ZERO_VIEW_P): Likewise.
23376         (new_line_info_table): Force-reset next view.
23377         (dwarf2out_begin_function): Likewise.
23378         (dwarf2out_source_line): Simplify zero_view_p initialization.
23379         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23380         view directly.  Omit view when omitting .loc at line 0.
23382 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23384         PR tree-optimization/84740
23385         * tree-switch-conversion.c (process_switch): Call build_constructors
23386         only if info.phi_count is non-zero.
23388         PR tree-optimization/84739
23389         * tree-tailcall.c (find_tail_calls): Check call arguments against
23390         DECL_ARGUMENTS (current_function_decl) rather than
23391         DECL_ARGUMENTS (func) when checking for tail recursion.
23393 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23395         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23396         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23397         Volker Reichelt's entry and add entries for people that perform
23398         GCC fuzzy testing and report numerous bugs.
23400 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
23402         PR target/82411
23403         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23404         readonly data in sdata, if that is disabled.
23405         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23406         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23407         -mreadonly-in-sdata option.
23409 2018-03-07  Martin Sebor  <msebor@redhat.com>
23411         PR tree-optimization/84468
23412         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23413         basic block when looking for nul assignment.
23415 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
23417         PR target/84277
23418         * except.h (output_function_exception_table): Adjust prototype.
23419         * except.c (output_function_exception_table): Remove FNNAME parameter
23420         and add SECTION parameter.  Ouput one part of the table at a time.
23421         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23422         the first part of the exception table and emit unwind directives.
23423         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23424         (i386_pe_seh_cold_init): Likewise.
23425         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23426         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23427         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23428         (ix86_output_call_insn): Emit a nop in one more case for SEH.
23429         * config/i386/winnt.c: Include except.h.
23430         (struct seh_frame_state): Add reg_offset, after_prologue and
23431         in_cold_section fields.
23432         (i386_pe_seh_end_prologue): Set seh->after_prologue.
23433         (i386_pe_seh_cold_init): New function.
23434         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23435         to seh->in_cold_section.
23436         (seh_emit_push): Record the offset of the push.
23437         (seh_emit_save): Record the offet of the save.
23438         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23439         Test seh->after_prologue to disregard the epilogue.
23440         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23441         (i386_pe_end_cold_function): New function.
23443 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23445         PR fortran/84565
23446         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23447         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23449         PR c++/84704
23450         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23451         on tmp_var.
23452         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23453         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23455         PR middle-end/84723
23456         * multiple_target.c: Include tree-inline.h and intl.h.
23457         (expand_target_clones): Diagnose and fail if node->definition and
23458         !tree_versionable_function_p (node->decl).
23460 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
23462         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23463         sprint_ul.
23464         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23465         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23466         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23468 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23470         PR target/84710
23471         * combine.c (try_combine): Use reg_or_subregno instead of handling
23472         just paradoxical SUBREGs and REGs.
23474 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
23476         * config/arc/arc.c (arc_finalize_pic): Remove function.
23477         (arc_must_save_register): We use single base PIC register, remove
23478         checks to save/restore the PIC register.
23479         (arc_expand_prologue): Likewise.
23480         * config/arc/arc-protos.h (arc_set_default_type_attributes):
23481         Remove.
23482         (arc_verify_short): Likewise.
23483         (arc_attr_type): Likewise.
23484         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23485         (walk_stores): Likewise.
23486         (arc_address_cost): Make it static.
23487         (arc_verify_short): Likewise.
23488         (branch_dest): Likewise.
23489         (arc_attr_type): Likewise.
23490         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23491         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23492         (arc_final_prescan_insn): Remove inserting the nops due to
23493         hardware hazards.  It is done in reorg step.
23494         (insn_length_variant_t): Remove.
23495         (insn_length_parameters_t): Likewise.
23496         (arc_insn_length_parameters): Likewise.
23497         (arc_get_insn_variants): Likewise.
23498         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23500 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23502         PR inline-asm/84683
23503         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23504         assertion failure.
23506         PR tree-optimization/84687
23507         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23508         on new_node->decl.
23509         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23511 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23513         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23514         Rename to ppc_speculation_barrier.
23515         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23516         __builtin_ppc_speculation_barrier.
23518 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23520         PR target/84700
23521         * combine.c (combine_simplify_rtx): Don't try to simplify if
23522         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23523         are equal to x.
23525 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
23527         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
23528         to 32 bytes when compiling for POWER9.
23530 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23532         PR target/84564
23533         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
23534         regparm >= 3 with no arg reg available also for calls with
23535         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
23537         PR target/84524
23538         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
23539         orig,vex.
23540         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
23542 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
23544         PR target/84264
23545         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
23547 2018-03-05  Richard Biener  <rguenther@suse.de>
23549         PR tree-optimization/84486
23550         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
23551         When inserting a __builtin_assume_aligned call set the LHS
23552         SSA name alignment info accordingly.
23554 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
23556         PR tree-optimization/84114
23557         * config/aarch64/aarch64.c (aarch64_reassociation_width)
23558         Avoid reassociation of FLOAT_MODE addition.
23560 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
23562         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
23563         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
23564         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
23565         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
23566         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
23567         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
23568         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
23569         and -mwbnoinvd.
23570         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
23571         __builtin_ia32_wbinvd): New builtins.
23572         (SPECIAL_ARGS2): New.
23573         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
23574         (SPECIAL_ARGS2): New.
23575         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
23576         (ix86_valid_target_attribute_inner_p): Ditto.
23577         (ix86_init_mmx_sse_builtins): Add special_args2.
23578         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
23579         TARGET_WBNOINVD_P): New.
23580         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
23581         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
23582         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
23583         * config/i386/immintrin.h (_wbinvd): New intrinsic.
23584         * config/i386/pconfigintrin.h: New file.
23585         * config/i386/wbnoinvdintrin.h: Ditto.
23586         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
23587         wbnoinvdintrin.h.
23588         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
23590 2018-03-05  Richard Biener  <rguenther@suse.de>
23592         PR tree-optimization/84670
23593         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23594         member.
23595         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23596         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23597         (compute_antic_aux): Only assert the number of values in ANTIC_IN
23598         doesn't grow if all successors (recursively) were visited at least
23599         once.
23601 2018-03-05  Richard Biener  <rguenther@suse.de>
23603         PR tree-optimization/84650
23604         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
23605         if executed in the loop pipeline.
23607 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
23609         * doc/configfiles.texi (Configuration Files): Move info about
23610         conditionalizing $target-protos.h to...
23611         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
23612         differs from $target-protos.h.
23614 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
23615             Chung-Ju Wu  <jasonwucj@gmail.com>
23617         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
23618         * config/nds32/nds32-multiple.md (setmemsi): Define.
23619         * config/nds32/nds32-memory-manipulation.c
23620         (nds32_gen_dup_4_byte_to_word_value): New.
23621         (emit_setmem_word_loop): New.
23622         (emit_setmem_byte_loop): New.
23623         (nds32_expand_setmem_loop): New.
23624         (nds32_expand_setmem_loop_v3m): New.
23625         (nds32_expand_setmem_unroll): New.
23626         (nds32_expand_setmem): New.
23628 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23629             Chung-Ju Wu  <jasonwucj@gmail.com>
23631         * config/nds32/nds32-memory-manipulation.c
23632         (nds32_emit_load_store): New.
23633         (nds32_emit_post_inc_load_store): New.
23634         (nds32_emit_mem_move): New.
23635         (nds32_emit_mem_move_block): New.
23636         (nds32_expand_movmemsi_loop_unknown_size): New.
23637         (nds32_expand_movmemsi_loop_known_size): New.
23638         (nds32_expand_movmemsi_loop): New.
23639         (nds32_expand_movmemsi_unroll): New.
23640         (nds32_expand_movmemqi): Rename ...
23641         (nds32_expand_movmemsi): ... to this.
23642         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
23643         (movmemsi): ... to this.
23644         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
23645         (nds32_expand_movmemsi): ... to this.
23647 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23648             Monk Chiang  <sh.chiang04@gmail.com>
23649             Chung-Ju Wu  <jasonwucj@gmail.com>
23651         * config/nds32/nds32-protos.h
23652         (nds32_expand_load_multiple): New arguments.
23653         (nds32_expand_store_multiple): Ditto.
23654         (nds32_valid_multiple_load_store): Rename ...
23655         (nds32_valid_multiple_load_store_p): ... to this.
23656         * config/nds32/nds32-memory-manipulation.c
23657         (nds32_expand_load_multiple): Refine implementation.
23658         (nds32_expand_store_multiple): Ditto.
23659         * config/nds32/nds32-multiple.md
23660         (load_multiple): Update nds32_expand_load_multiple interface.
23661         (store_multiple): Update nds32_expand_store_multiple interface.
23662         * config/nds32/nds32-predicates.c
23663         (nds32_valid_multiple_load_store): Rename ...
23664         (nds32_valid_multiple_load_store_p): ... to this and refine
23665         implementation.
23666         * config/nds32/predicates.md
23667         (nds32_load_multiple_and_update_address_operation): New predicate.
23668         (nds32_store_multiple_and_update_address_operation): New predicate.
23670 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23671             Chung-Ju Wu  <jasonwucj@gmail.com>
23673         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
23674         (combo): New attribute.
23675         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
23677 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
23679         * config/nds32/nds32.opt: Change -mcmodel= default value.
23681 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
23682             Monk Chiang  <sh.chiang04@gmail.com>
23683             Chung-Ju Wu  <jasonwucj@gmail.com>
23685         * config/nds32/constants.md (unspec_element): New enum.
23686         * config/nds32/constraints.md (Umw): New constraint.
23687         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
23688         * config/nds32/nds32-intrinsic.md: Likewise.
23689         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
23690         (nds32_valid_smw_lwm_base_p): New.
23691         (nds32_output_smw_single_word): New.
23692         (nds32_output_lmw_single_word): New.
23693         (nds32_expand_unaligned_load): New.
23694         (nds32_expand_unaligned_store): New.
23695         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
23696         (nds32_output_smw_single_word): Declare.
23697         (nds32_output_lmw_single_word): Declare.
23698         (nds32_expand_unaligned_load): Declare.
23699         (nds32_expand_unaligned_store): Declare.
23700         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
23701         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
23702         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
23703         NDS32_BUILTIN_UASTORE_DW.
23704         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
23705         predicate.
23707 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
23708             Kito Cheng  <kito.cheng@gmail.com>
23709             Chung-Ju Wu  <jasonwucj@gmail.com>
23711         * config/nds32/nds32-intrinsic.c
23712         (nds32_expand_builtin_null_ftype_reg): Delete.
23713         (nds32_expand_builtin_reg_ftype_imm): Ditto.
23714         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
23715         (nds32_read_argument): New.
23716         (nds32_legitimize_target): Ditto.
23717         (nds32_legitimize_argument): Ditto.
23718         (nds32_check_constant_argument): Ditto.
23719         (nds32_expand_unop_builtin): Ditto.
23720         (nds32_expand_unopimm_builtin): Ditto.
23721         (nds32_expand_binop_builtin): Ditto.
23722         (nds32_builtin_decl_impl): Ditto.
23723         (builtin_description): Ditto.
23724         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
23725         (nds32_init_builtins_impl): Ditto.
23726         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
23727         (nds32_builtin_decl): New.
23728         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
23729         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
23731 2018-03-02  Jeff Law  <law@redhat.com>
23733         * reorg.c (stop_search_p): Handle DEBUG_INSN.
23734         (redundant_insn, fill_simple_delay_slots): Likewise.
23735         (fill_slots_from_thread): Likewise.
23736         * resource.c (mark_referenced_resources): Likewise.
23737         (mark_set_resources, find_dead_or_set_registers): Likewise.
23739 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23741         * substring-locations.h (format_warning_va): Formatting fix for
23742         ATTRIBUTE_GCC_DIAG.
23743         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
23744         argument.
23745         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
23746         * substring-locations.c: Include intl.h.
23747         (format_warning_va): Turned into small wrapper around
23748         format_warning_n_va, renamed to ...
23749         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
23750         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
23751         use ngettext.
23752         (format_warning_at_substring_n): New function.
23753         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
23754         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
23755         format_warning_at_substring with just a shorter name instead of
23756         const function pointer.
23757         (fmtwarn_n): New function.
23758         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
23759         appropriate, get rid of all the fmtstr temporaries, move conditionals
23760         with G_() wrapped string literals directly into fmtwarn arguments,
23761         cast dir.len to (int), formatting fixes.
23763 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
23765         * doc/invoke.texi: Remove "Cilk Plus" references.
23767 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23768             Richard Biener  <rguenther@suse.de>
23770         PR ipa/84628
23771         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
23772         for error or warning attributes if CALL_FROM_THUNK_P is set.
23773         Formatting fixes.
23775 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23777         PR target/56540
23778         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
23779         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
23781         PR target/56540
23782         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
23783         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
23785         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
23786         instead of -1U in last predictors element's probability member.
23788 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
23790         PR ipa/83983
23791         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23792         arguments if they are comparable.
23794 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23796         PR tree-optimization/84634
23797         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
23798         masks and masked_loop_p with a single loop_masks, making sure it's
23799         null for bb vectorization.
23801 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23803         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
23804         (vect_analyze_data_ref_access): Use loop->safe_len rather than
23805         loop->force_vectorize to check whether there is no alias.
23807 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23809         PR target/84614
23810         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
23811         prototypes.
23812         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
23813         comments.
23814         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
23815         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
23816         instead of a loop around prev_real_insn.
23817         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
23818         prev_real_insn.
23820         PR inline-asm/84625
23821         * config/i386/i386.c (ix86_print_operand): Use conditional
23822         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
23823         zero vector.
23825 2018-03-02  Richard Biener  <rguenther@suse.de>
23827         PR tree-optimization/84427
23828         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
23829         (bitmap_set_subtract_values): Rewrite to handle multiple
23830         exprs per value.
23831         (clean): Likewise.
23832         (prune_clobbered_mems): Likewise.
23833         (phi_translate): Take edge instead of pred/phiblock.
23834         (phi_translate_1): Likewise.
23835         (phi_translate_set): Likewise.  Insert all translated
23836         exprs for a value into the set, keeping possibly multiple
23837         expressions per value.
23838         (compute_antic_aux): Adjust for phi_translate changes.
23839         When intersecting union the expressions and prune those
23840         not in the final value set, keeping possibly multiple
23841         expressions per value.  Do not use value-insertion
23842         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
23843         all expressions.  Add verification that the value-sets
23844         only shrink during iteration.
23845         (compute_partial_antic_aux): Adjust for the phi_translate changes.
23846         (do_pre_regular_insertion): Likewise.
23847         (do_pre_partial_partial_insertion): Likewise.
23849 2018-03-02  Richard Biener  <rguenther@suse.de>
23851         PR target/82005
23852         * config/darwin.c (saved_debug_info_level): New static global.
23853         (darwin_asm_lto_start): Disable debug info generation for LTO out.
23854         (darwin_asm_lto_end): Restore debug info generation settings.
23856 2018-03-01  Martin Liska  <mliska@suse.cz>
23858         PR sanitizer/82484
23859         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
23860         volatile arguments.
23862 2018-03-01  Richard Biener  <rguenther@suse.de>
23864         PR debug/84645
23865         * dwarf2out.c (gen_variable_die): Properly handle late VLA
23866         type annotation with LTO when debug was disabled at compile-time.
23868 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
23870         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
23871         XINT with INTVAL.
23872         (mips_final_postscan_insn): Likewise.
23874 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
23876         PR rtl-optimization/84528
23877         * alias.c (init_alias_target): Add commentary.
23878         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
23879         a unique base value if the frame pointer is not eliminated
23880         to the stack pointer.
23882 2018-03-01  Tom de Vries  <tom@codesourcery.com>
23884         PR rtl-optimization/83327
23885         * lra-int.h (hard_regs_spilled_into): Declare.
23886         * lra.c (hard_regs_spilled_into): Define.
23887         (init_reg_info): Init hard_regs_spilled_into.
23888         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
23889         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
23890         (process_bb_lives): Handle hard_regs_spilled_into.
23891         (lra_create_live_ranges_1): Before doing liveness propagation, clear
23892         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
23894 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
23896         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
23897         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
23898         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
23899         * config/rs6000/aix72.h: New file.
23901 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
23903         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
23904         instead of warning_at with conditional singular and plural messages
23905         where possible.
23907         PR target/52991
23908         * stor-layout.c (update_alignment_for_field): For
23909         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
23910         && !DECL_PACKED (field), do the alignment update, just use
23911         only desired_align instead of MAX (type_align, desired_align)
23912         as the alignment.
23913         (place_field): Don't do known_align < desired_align handling
23914         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
23915         is non-NULL, instead do it after rli->prev_field handling and
23916         only if not within a bitfield word.  For DECL_PACKED (field)
23917         use type_align of BITS_PER_UNIT.
23919 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
23921         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
23922         superfluous parentheses and trailing spaces.
23924 2018-02-28  Richard Biener  <rguenther@suse.de>
23926         PR tree-optimization/84584
23927         * graphite-scop-detection.c (scop_detection::add_scop): Discard
23928         SCoPs with fake exit edge.
23930 2018-02-28  Martin Liska  <mliska@suse.cz>
23932         PR testsuite/84597
23933         * timevar.c (timer::print): Fix format to properly print 100%
23934         values.
23936 2018-02-28  Richard Biener  <rguenther@suse.de>
23938         PR middle-end/84607
23939         * genmatch.c (capture_info::walk_match): Do not mark
23940         captured expressions without operands as expr_p given
23941         they act more like predicates and should be subject to
23942         "lost tail" side-effect preserving.
23944 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
23946         PR rtl-optimization/81611
23947         * auto-inc-dec.c (attempt_change): Move dead note from
23948         mem_insn if it's the next use of regno
23949         (find_address): Take address use of reg holding
23950         non-incremented value.  Add parm to limit search to the named
23951         reg only.
23952         (merge_in_block): Attempt to use a mem insn that is the next
23953         use of the original regno.
23955 2018-02-27  Martin Sebor  <msebor@redhat.com>
23957         PR c++/83871
23958         * doc/invoke.texi (-Wmissing-attributes): New option.
23959         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
23961 2018-02-27  Martin Sebor  <msebor@redhat.com>
23963         PR translation/84207
23964         * diagnostic-core.h (warning_n, error_n, inform_n): Change
23965         n argument to unsigned HOST_WIDE_INT.
23966         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
23967         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
23968         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
23969         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
23971 2018-02-27  Richard Biener  <rguenther@suse.de>
23973         PR tree-optimization/84512
23974         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
23975         Do not use the estimate returned from record_stmt_cost for
23976         the scalar iteration cost but sum properly using add_stmt_cost.
23978 2018-02-27  Richard Biener  <rguenther@suse.de>
23980         PR tree-optimization/84466
23981         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23982         Adjust last change to less strictly validate use operands.
23984 2018-02-27  Martin Liska  <mliska@suse.cz>
23986         PR gcov-profile/84548
23987         * gcov.c (process_file): Allow partial overlap and consider it
23988         also as group functions.
23989         (output_lines): Properly calculate range of lines for a group.
23991 2018-02-27  Martin Liska  <mliska@suse.cz>
23993         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
23994         'ggc' suffixes.  Change first column width.
23995         (timer::print): Fix formatting of the column.
23997 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
23999         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24000         preserve inline entry blocks for the sake of debug inline
24001         entry point markers alone.
24002         (remove_unused_locals): Suggest in comments a better place to
24003         force the preservation of inline entry blocks that are
24004         otherwise unused, but do not preserve them.
24006 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24008         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24010 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24012         PR target/84039
24013         * config/i386/constraints.md (Bs): Replace
24014         ix86_indirect_branch_register with
24015         TARGET_INDIRECT_BRANCH_REGISTER.
24016         (Bw): Likewise.
24017         * config/i386/i386.md (indirect_jump): Likewise.
24018         (tablejump): Likewise.
24019         (*sibcall_memory): Likewise.
24020         (*sibcall_value_memory): Likewise.
24021         Peepholes of indirect call and jump via memory: Likewise.
24022         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24023         (*sibcall_value_GOT_32): Likewise.
24024         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24025         (GOT_memory_operand): Likewise.
24026         (call_insn_operand): Likewise.
24027         (sibcall_insn_operand): Likewise.
24028         (GOT32_symbol_operand): Likewise.
24029         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24031 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24033         PR rtl-optimization/83496
24034         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24035         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24036         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24037         redundant insn, if any.
24038         (relax_delay_slots): Likewise.
24039         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24041 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24043         PR tree-optimization/83965
24044         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24045         that grouped statements are part of a reduction chain.  Return
24046         true if the statement is not marked as a reduction itself but
24047         is part of a group.
24048         (vect_recog_dot_prod_pattern): Don't check whether the statement
24049         is part of a group here.
24050         (vect_recog_sad_pattern): Likewise.
24051         (vect_recog_widen_sum_pattern): Likewise.
24053 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24055         PR debug/84545
24056         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24058 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24060         PR target/84530
24061         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24062         the bool argument.
24063         (ix86_output_indirect_function_return): New prototype.
24064         (ix86_split_simple_return_pop_internal): Likewise.
24065         * config/i386/i386.c (indirect_return_via_cx): New.
24066         (indirect_return_via_cx_bnd): Likewise.
24067         (indirect_thunk_name): Handle return va CX_REG.
24068         (output_indirect_thunk_function): Create alias for
24069         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24070         (ix86_output_indirect_jmp): Remove the bool argument.
24071         (ix86_output_indirect_function_return): New function.
24072         (ix86_split_simple_return_pop_internal): Likewise.
24073         * config/i386/i386.md (*indirect_jump): Don't pass false
24074         to ix86_output_indirect_jmp.
24075         (*tablejump_1): Likewise.
24076         (simple_return_pop_internal): Change it to define_insn_and_split.
24077         Call ix86_split_simple_return_pop_internal to split it for
24078         -mfunction-return=.
24079         (simple_return_indirect_internal): Call
24080         ix86_output_indirect_function_return instead of
24081         ix86_output_indirect_jmp.
24083 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24085         PR bootstrap/84405
24086         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24087         memset and value initialization afterwards.
24089 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24091         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24093 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24095         PR target/84521
24096         * common/config/aarch64/aarch64-common.c
24097         (aarch_option_optimization_table[]): Switch
24098         off fomit-frame-pointer
24100 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24101             Chung-Ju Wu  <jasonwucj@gmail.com>
24103         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24104         volatile memory.
24105         (store_multiple): Ditto.
24107 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24109         * config.gcc: Add --with-cpu support for nds32 target.
24110         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24111         * config/nds32/nds32.opt: Add -mcpu= option.
24113 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24115         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24116         isel=yes): Warn for these deprecated options.
24118 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24120         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24121         ISA_2_5_MASKS_EMBEDDED.
24123 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24125         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24126         p->max as pointers rather than using iterative_hash_expr.
24128 2018-02-23  Carl Love  <cel@us.ibm.com>
24130         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24131         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24132         BU_P8V_OVERLOAD_2.
24133         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24134         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24135         P8V_BUILTIN_VEC_VUNSIGNED2.
24137 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
24139         PR target/81572
24140         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24141         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24142         LRA_UNKNOWN_ALT.
24143         * lra-constraints.c (curr_insn_transform): Set up
24144         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
24145         LRA_UNKNOWN_ALT.
24146         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24147         * lra-eliminations.c (spill_pseudos): Ditto.
24148         (process_insn_for_elimination): Ditto.
24149         * lra-lives.c (reg_early_clobber_p): Use the new macros.
24150         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24151         LRA_NON_CLOBBERED_ALT.
24153 2018-02-22  Martin Sebor  <msebor@redhat.com>
24155         PR tree-optimization/84480
24156         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24157         to maybe_diag_stxncpy_trunc.  Call it.
24158         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24159         from gimple_fold_builtin_strcpy.  Print inlining stack.
24160         (handle_builtin_stxncpy): Print inlining stack.
24161         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24163 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
24165         PR target/84176
24166         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24167         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24168         and -fcheck-pointer-bounds are used together.
24169         (indirect_thunk_prefix): New enum.
24170         (indirect_thunk_need_prefix): New function.
24171         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
24172         "_nt" instead of "_bnd" for NOTRACK prefix.
24173         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24174         (output_indirect_thunk_function): Likewise.
24175         (): Likewise.
24176         (ix86_code_end): Update output_indirect_thunk_function calls.
24177         (ix86_output_indirect_branch_via_reg): Replace
24178         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24179         (ix86_output_indirect_branch_via_push): Likewise.
24180         (ix86_output_function_return): Likewise.
24181         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24182         incompatible with -fcf-protection=branch and
24183         -fcheck-pointer-bounds.
24185 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24187         PR target/83335
24188         * config/aarch64/aarch64.c (aarch64_print_address_internal):
24189         Change gcc_assert call to output_operand_lossage.
24191 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24193         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24195 2018-02-22  DJ Delorie  <dj@redhat.com>
24196             Sebastian Perta  <sebastian.perta@renesas.com>
24197             Oleg Endo  <olegendo@gcc.gnu.org>
24199         * config/rx/rx.c (rx_rtx_costs): New function.
24200         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24202 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24204         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24206 2018-02-22  Martin Liska  <mliska@suse.cz>
24208         PR driver/83193
24209         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24210         Add "native" as a possible value.
24212 2018-02-22  Martin Liska  <mliska@suse.cz>
24214         PR driver/83193
24215         * config/i386/i386.c (ix86_option_override_internal):
24216         Add "native" as a possible value for -march and -mtune.
24218 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24220         PR target/84502
24221         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24222         to all type variants.
24224         PR tree-optimization/84503
24225         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24226         width as info->bitpos + info->bitsize - start.
24227         (merged_store_group::merge_overlapping): Simplify width computation.
24228         (check_no_overlap): New function.
24229         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24230         start + width and last_order of the group, fail if check_no_overlap
24231         fails.
24232         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24233         to group if check_no_overlap fails.
24235 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24237         * config/rs6000/altivec.md: Delete contraint arguments to
24238         define_expand, define_split, and define_peephole2, and in
24239         define_insn_and_split if always unused.
24240         * config/rs6000/darwin.md: Ditto.
24241         * config/rs6000/dfp.md: Ditto.
24242         * config/rs6000/rs6000.md: Ditto.
24243         * config/rs6000/sync.md: Ditto.
24244         * config/rs6000/vector.md: Ditto.
24245         * config/rs6000/vsx.md: Ditto.
24247 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24249         * config/rs6000/altivec.md: Write output control strings as braced
24250         blocks instead of double-quoted strings.
24251         * config/rs6000/darwin.md: Ditto.
24252         * config/rs6000/rs6000.md: Ditto.
24253         * config/rs6000/vector.md: Ditto.
24254         * config/rs6000/vsx.md: Ditto.
24256 2018-02-21  Jason Merrill  <jason@redhat.com>
24258         PR c++/84314 - ICE with templates and fastcall attribute.
24259         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24261 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24263         * ipa-cp.c (determine_versionability): Fix comment typos.
24265 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24267         PR c/84229
24268         * ipa-cp.c (determine_versionability): Do not version functions caling
24269         va_arg_pack.
24271 2018-02-21  Martin Liska  <mliska@suse.cz>
24273         PR driver/83193
24274         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24275         Add "native" as a possible value.
24276         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
24277         the macro when native cpu detection is available.
24279 2018-02-21  Martin Liska  <mliska@suse.cz>
24281         PR driver/83193
24282         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24283         Add "native" as a possible value.
24284         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24285         when native cpu detection is available.
24287 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
24288             Martin Sebor  <msebor@redhat.com>
24290         PR tree-optimization/84478
24291         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24292         false.
24293         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24294         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24295         support which is conservatively correct, for 2 only stay conservative
24296         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
24297         argument to the 2 argument get_range_strlen, adjust 6 arg
24298         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24299         false.
24300         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24301         (gimple_fold_builtin_strlen): Pass true as last argument to
24302         get_range_strlen.
24304 2018-02-20  Martin Sebor  <msebor@redhat.com>
24306         PR middle-end/84095
24307         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24308         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
24309         (builtin_memref::builtin_memref): Factor out parts into
24310         set_base_and_offset and call it.
24312 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24314         PR middle-end/84406
24315         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24316         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24317         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
24318         search at the associated MODE_INT.
24320 2018-02-20  Jeff Law  <law@redhat.com>
24322         PR middle-end/82123
24323         PR tree-optimization/81592
24324         PR middle-end/79257
24325         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24326         for range data rather than using global data.
24327         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24328         range data rather than using global data.
24329         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24330         pass it to children as needed.
24331         (struct directive::fmtresult): Similarly.
24332         (struct directive::set_width): Similarly.
24333         (struct directive::set_precision): Similarly.
24334         (format_integer, format_directive, parse_directive): Similarly.
24335         (format_none): Accept unnamed vr_values parameter.
24336         (format_percent, format_floating, format_character): Similarly.
24337         (format_string, format_plain): Similarly.
24338         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24339         the EVRP range analyzer for range data rather than using global data.
24340         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24341         gimple-ssa-evrp-analyze.h
24342         (class sprintf_dom_walker): Add after_dom_children member function.
24343         Add evrp_range_analyzer member.
24344         (sprintf_dom_walker::before_dom_children): Call into the EVRP
24345         range analyzer as needed.
24346         (sprintf_dom_walker::after_dom_children): New member function.
24347         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24348         if not optimizing.
24349         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24350         (evrp_range_analyzer::pop_to_marker): Likewise.
24352 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24354         PR tree-optimization/84419
24355         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24356         with the required type if its current type is compatible but
24357         different.
24359 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
24361         PR middle-end/82004
24362         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24363         after vectorization.
24365 2018-02-20  Martin Liska  <mliska@suse.cz>
24367         PR driver/83193
24368         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24369         possible values if we don't have a hint.
24371 2018-02-20  Martin Liska  <mliska@suse.cz>
24373         PR c/84310
24374         PR target/79747
24375         * final.c (shorten_branches): Build align_tab array with one
24376         more element.
24377         * opts.c (finish_options): Add alignment option limit check.
24378         (MAX_CODE_ALIGN): Likewise.
24379         (MAX_CODE_ALIGN_VALUE): Likewise.
24380         * doc/invoke.texi: Document maximum allowed option value for
24381         all -falign-* options.
24383 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
24385         PR target/84146
24386         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24387         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24388         * var-tracking.c (emit_note_insn_var_location): Remove all references
24389         to NOTE_INSN_CALL_ARG_LOCATION.
24390         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24391         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24392         Use copy_rtx_if_shared.
24393         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24394         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24395         (dwarf2out_var_location): Remove handling of
24396         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24397         on call_insn.
24398         * final.c (final_scan_insn): Remove all references to
24399         NOTE_INSN_CALL_ARG_LOCATION.
24400         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
24401         before dumping final insns.
24402         * except.c (emit_note_eh_region_end): Remove all references to
24403         NOTE_INSN_CALL_ARG_LOCATION.
24404         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24405         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24406         * config/arc/arc.c (hwloop_optimize): Likewise.
24407         * config/arm/arm.c (create_fix_barrier): Likewise.
24408         * config/s390/s390.c (s390_chunkify_start): Likewise.
24409         * config/sh/sh.c (find_barrier): Likewise.
24410         * config/i386/i386.c (rest_of_insert_endbranch,
24411         ix86_seh_fixup_eh_fallthru): Likewise.
24412         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24413         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24414         * config/frv/frv.c (frv_function_prologue): Likewise.
24415         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
24416         reg note.
24417         (note_outside_basic_block_p): Remove all references to
24418         NOTE_INSN_CALL_ARG_LOCATION.
24419         * gengtype.c (adjust_field_rtx_def): Likewise.
24420         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24421         Likewise.
24422         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24423         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24425         PR c++/84444
24426         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24427         is ADDR_EXPR.
24429         PR tree-optimization/84452
24430         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24431         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24432         is NULL.
24434 2018-02-19  Martin Liska  <mliska@suse.cz>
24436         PR sanitizer/82183
24437         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24439 2018-02-19  Martin Liska  <mliska@suse.cz>
24440             Richard Sandiford  <richard.sandiford@linaro.org>
24442         PR tree-optimization/82491
24443         * gimple-fold.c (get_base_constructor): Make earlier bail out
24444         to prevent ubsan.
24446 2018-02-19  Carl Love  <cel@us.ibm.com>
24448         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24449         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24450         BU_P8V_OVERLOAD_1.
24451         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24452         P8V_BUILTIN_VEC_NEG.
24454 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
24456         * config/rl78/rl78.md (movdf): New define expand.
24458 2018-02-19  Martin Liska  <mliska@suse.cz>
24460         PR other/80589
24461         * doc/invoke.texi: Fix typo.
24462         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24464 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
24466         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24467         handle rs6000_single_float and rs6000_double_float specially for
24468         e500 family CPUs.
24470 2018-02-16  Jeff Law  <law@redhat.com>
24472         * config/rx/rx.c (add_pop_cfi_notes): New function.;
24473         (pop_regs): Use it.
24475 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24477         PR ipa/84425
24478         * ipa-inline.c (inline_small_functions): Fix a typo.
24480 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24482         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24484 2018-02-16  Carl Love  <cel@us.ibm.com>
24486         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24487         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24488         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24489         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24490         expansion to P8V_BUILTIN_VEC_FLOAT2.
24492 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
24494         PR rtl-optimization/70023
24495         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24496         src_regno into account.
24498 2018-02-16  Carl Love  <cel@us.ibm.com>
24500         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24501         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24502         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24503         * config/rs6000/rs6000.c: Remove case statements for
24504         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24505         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24506         and P9V_BUILTIN_VEC_VINSERT4B.
24507         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24508         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24509         * config/rs6000/vsx.md:
24510         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24511         vec_insert4b.
24513 2018-02-16  Carl Love  <cel@us.ibm.com>
24515         * config/rs6000/altivec.h: Add builtin names vec_extract4b
24516         vec_insert4b.
24517         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24518         definitions.
24519         * config/rs6000/rs6000-c.c: Add the definitions for
24520         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24521         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24522         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24523         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
24524         definition for insert4b and define insn *insert3b_internal.
24525         * doc/extend.texi: Add documentation for vec_extract4b.
24527 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24529         * doc/extend.texi (Backwards Compatibility): Mention friend
24530         injection.  Note for-scope is deprecated.
24531         * doc/invoke.texi (-ffriend-injection): Deprecate.
24533 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
24535         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
24536         that moved to I2, also allow destinations that are a paradoxical
24537         subreg (instead of a normal reg).
24539 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
24541         PR target/83831
24542         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
24543         to QImode.
24545 2018-02-16  Richard Biener  <rguenther@suse.de>
24547         PR tree-optimization/84037
24548         PR tree-optimization/84016
24549         PR target/82862
24550         * config/i386/i386.c (ix86_builtin_vectorization_cost):
24551         Adjust vec_construct for the fact we need additional higher latency
24552         128bit inserts for AVX256 and AVX512 vector builds.
24553         (ix86_add_stmt_cost): Scale vector construction cost for
24554         elementwise loads.
24556 2018-02-16  Richard Biener  <rguenther@suse.de>
24558         PR tree-optimization/84417
24559         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
24560         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
24561         (non_rewritable_lvalue_p): Likewise, use poly-ints.
24563 2018-02-16  Martin Liska  <mliska@suse.cz>
24565         PR sanitizer/84307
24566         * internal-fn.def (ASAN_CHECK): Set proper flags.
24567         (ASAN_MARK): Likewise.
24569 2018-02-16  Julia Koval  <julia.koval@intel.com>
24571         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
24572         from PTA_CANNONLAKE.
24574 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24576         PR target/84272
24577         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
24578         Use ++iter rather than iter++ for std::list iterators.
24579         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
24580         defer deleting them until all nodes in the forest are processed.  Do
24581         free even leaf nodes.  Change to_process into auto_vec.
24583         PR bootstrap/84405
24584         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
24585         * vec.h (vec_default_construct): Use memset instead of placement new
24586         if BROKEN_VALUE_INITIALIZATION is defined.
24587         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
24588         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
24589         is defined.
24591         PR rtl-optimization/83723
24592         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
24593         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
24594         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
24595         recursive calls.
24596         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
24597         callers.
24598         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
24600 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
24602         PR rtl-optimization/81443
24603         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
24604         from inner REGs to paradoxical SUBREGs.
24606 2018-02-16  Richard Biener  <rguenther@suse.de>
24608         PR tree-optimization/84399
24609         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24610         For operands we can analyze at their definition make sure we can
24611         analyze them at each use as well.
24613 2018-02-16  Richard Biener  <rguenther@suse.de>
24615         PR tree-optimization/84190
24616         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
24617         volatile accesses if the decl isn't volatile.
24619 2018-02-15  Jason Merrill  <jason@redhat.com>
24621         PR c++/84314 - ICE with templates and fastcall attribute.
24622         * attribs.c (build_type_attribute_qual_variant): Don't clobber
24623         TYPE_CANONICAL on an existing type.
24625 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
24627         PR tree-optimization/84383
24628         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
24629         dstoff nor call operand_equal_p if dstbase is NULL.
24631         PR tree-optimization/84334
24632         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
24633         also a CONSTANT_CLASS_P, punt.
24635 2018-02-14  Jim Wilson  <jimw@sifive.com>
24637         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
24638         first SMALL_OPERAND check.  New local min_second_step.  Move assert
24639         to where locals are set.  Add TARGET_RVC support.
24640         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
24642 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
24644         * doc/invoke.texi: Correct -Wformat-overflow code sample.
24646 2018-02-14  Martin Sebor  <msebor@redhat.com>
24648         PR tree-optimization/83698
24649         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
24650         arrays constrain the offset range to their bounds.
24651         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
24652         (builtin_access::overlap): Avoid setting the size of overlap if it's
24653         already been set.
24654         (maybe_diag_overlap): Also consider arrays when deciding what values
24655         of offsets to include in diagnostics.
24657 2018-02-14  Martin Sebor  <msebor@redhat.com>
24659         PR c/84108
24660         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
24661         that correspond to the kind of a declaration.
24663 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
24665         PR target/83984
24666         * config/pa/pa.md: Load address of PIC label using the linkage table
24667         if the label is nonlocal.
24669 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24671         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
24672         warning message if user requests -maltivec=be.
24673         * doc/invoke.texi: Document deprecation of -maltivec=be.
24675 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
24677         PR target/84220
24678         * config/rs6000/rs6000-c.c: Update definitions for
24679         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
24680         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
24682 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24684         PR target/84239
24685         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
24686         add _get_ssp intrinsics. Remove argument from
24687         __builtin_ia32_rdssp[d|q].
24688         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
24689         * config/i386/i386-builtin.def: Remove argument from
24690         __builtin_ia32_rdssp[d|q].
24691         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
24692         ix86_expand_special_args_builtin for _rdssp[d|q].
24693         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
24694         Clear register before usage.
24695         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
24696         Add documentation for new _get_ssp and _inc_ssp intrinsics.
24698 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
24700         PR tree-optimization/84357
24701         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
24702         operand 1 of an ARRAY_REF too.
24704 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
24706         PR target/83831
24707         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
24708         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24709         declarations.
24710         (set_of_reg): New struct.
24711         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
24712         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
24713         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24714         functions.
24715         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
24716         Split into bitclr, bitset, bitinvert patterns if appropriate.
24717         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
24718         use rx_fuse_in_memory_bitop.
24719         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
24720         to named insn, correct maximum insn length.
24722 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
24724         PR target/79242
24725         * machmode.def: Define a complex mode for PARTIAL_INT.
24726         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
24727         MODE_PARTIAL_INT.
24728         * doc/rtl.texi: Document CSPImode.
24729         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
24730         handling.
24731         (msp430_hard_regno_nregs_with_padding): Likewise.
24733 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
24735         PR target/84279
24736         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
24738 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
24740         PR rtl-optimization/84169
24741         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
24742         we generated a parallel as new i3 and we split that to new i2 and i3
24743         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
24744         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
24745         those to i2, not i1.  Partially rewrite this scan code.
24747 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24749         PR c/82210
24750         * stor-layout.c (place_field): For variable length fields, adjust
24751         offset_align afterwards not just based on the field's alignment,
24752         but also on the size.
24754         PR middle-end/84309
24755         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
24756         of exps and logs in the use_exp2 case.
24758 2018-02-13  Jeff Law  <law@redhat.com>
24760         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
24761         entry for "vector".
24763         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
24764         ARGS as unused.
24766 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
24768         PR debug/84342
24769         PR debug/84319
24770         * common.opt (gas-loc-support, gas-locview-support): New.
24771         (ginline-points, ginternal-reset-location-views): New.
24772         * doc/invoke.texi: Document them.  Use @itemx where intended.
24773         (gvariable-location-views): Adjust.
24774         * target.def (reset_location_view): New.
24775         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
24776         (TARGET_RESET_LOCATION_VIEW): New.
24777         * doc/tm.texi: Rebuilt.
24778         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
24779         (dwarf2out_default_as_locview_support): New.
24780         (output_asm_line_debug_info): Use option variables.
24781         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
24782         (output_loc_list): Likewise.
24783         (add_high_low_attributes): Check option variables.
24784         Don't output entry view attribute in strict mode.
24785         (gen_inlined_subroutine_die): Check option variables.
24786         (dwarf2out_inline_entry): Likewise.
24787         (init_sections_and_labels): Likewise.
24788         (dwarf2out_early_finish): Likewise.
24789         (maybe_reset_location_view): New, from...
24790         (dwarf2out_var_location): ... here.  Call it.
24791         * debug.h (dwarf2out_default_as_loc_support): Declare.
24792         (dwarf2out_default_as_locview_support): Declare.
24793         * hooks.c (hook_int_rtx_insn_0): New.
24794         * hooks.h (hook_int_rtx_insn_0): Declare.
24795         * toplev.c (process_options): Take -gas-loc-support and
24796         -gas-locview-support from dwarf2out.  Enable
24797         -gvariable-location-views by default only with locview
24798         assembler support.  Enable -ginternal-reset-location-views by
24799         default only if the target defines the corresponding hook.
24800         Enable -ginline-points by default if location views are
24801         enabled; force it disabled if statement frontiers are
24802         disabled.
24803         * tree-inline.c (expand_call_inline): Check option variables.
24804         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
24806 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
24808         PR tree-optimization/84321
24809         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
24810         handling.  Also check whether the anti-range contains any values
24811         that satisfy the mask; switch to a VR_RANGE if not.
24813 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
24815         PR sanitizer/84340
24816         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
24818 2018-02-13  Martin Jambor  <mjambor@suse.cz>
24820         PR c++/83990
24821         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
24822         of call statements, also set location of a load to a temporary.
24824 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
24826         * config/rl78/rl78.c (add_vector_labels): New function.
24827         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
24828         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
24829         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
24830         which checks that no arguments are passed.
24831         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
24832         * doc/extend.texi: Documentation for the new attribute.
24834 2018-02-13  Andreas Schwab  <schwab@suse.de>
24836         * config/riscv/linux.h (CPP_SPEC): Define.
24838 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24840         PR target/84335
24841         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
24842         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
24843         OPTION_MASK_ISA_AES as first argument to def_builtin_const
24844         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
24845         instead of OPTION_MASK_ISA_PCLMUL as first argument to
24846         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
24847         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
24848         temporarily for AES and PCLMUL builtins.
24850         PR tree-optimization/84339
24851         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
24852         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
24853         Formatting fixes.
24855         PR middle-end/84309
24856         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
24857         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
24858         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
24859         inline function.
24860         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
24861         inline function.
24862         * omp-simd-clone.h: New file.
24863         * omp-simd-clone.c: Include omp-simd-clone.h.
24864         (expand_simd_clones): No longer static.
24865         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
24866         cgraph.h and omp-simd-clone.h.
24867         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
24868         (vect_recog_widen_shift_pattern): Formatting fix.
24869         (vect_pattern_recog_1): Don't check optab for calls.
24871         PR target/84336
24872         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
24873         operands[2] into a REG before using gen_lowpart on it.
24875 2018-02-12  Jeff Law  <law@redhat.com>
24877         PR target/83760
24878         * config/sh/sh.c (find_barrier): Consider a sibling call
24879         a barrier as well.
24881         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
24882         successfully back substituting a reg.
24884 2018-02-12  Richard Biener  <rguenther@suse.de>
24886         PR tree-optimization/84037
24887         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
24888         parameter, move visited init to caller.
24889         (vect_slp_analyze_operations): Separate cost from validity
24890         check, initialize visited once for all instances.
24891         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
24892         for all instances.
24893         * tree-vect-stmts.c (vect_model_simple_cost): Make early
24894         out an assert.
24895         (vect_model_promotion_demotion_cost): Likewise.
24896         (vectorizable_bswap): Guard cost modeling with !slp_node
24897         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
24898         SLP stmts.
24899         (vectorizable_call): Likewise.
24900         (vectorizable_conversion): Likewise.
24901         (vectorizable_assignment): Likewise.
24902         (vectorizable_shift): Likewise.
24903         (vectorizable_operation): Likewise.
24904         (vectorizable_store): Likewise.
24905         (vectorizable_load): Likewise.
24906         (vectorizable_condition): Likewise.
24907         (vectorizable_comparison): Likewise.
24909 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
24911         PR sanitizer/84307
24912         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
24913         (ASAN_MARK): Fix fnspec to account for return value, change pointer
24914         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
24916 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24918         PR middle-end/83665
24919         * params.def (inline-min-speedup): Increase from 8 to 15.
24920         (max-inline-insns-auto): Decrease from 40 to 30.
24921         * ipa-split.c (consider_split): Add some buffer for function to
24922         be considered inlining candidate.
24923         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
24924         default values.
24926 2018-02-12  Richard Biener  <rguenther@suse.de>
24928         PR tree-optimization/84037
24929         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
24930         matched stmts if we cannot swap the non-matched ones.
24932 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24934         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
24935         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
24936         _mm_maskz_scalef_round_ss): New intrinsics.
24937         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
24938         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
24939         __builtin_ia32_scalefss_round): Remove.
24940         (__builtin_ia32_scalefsd_mask_round,
24941         __builtin_ia32_scalefss_mask_round): New intrinsics.
24942         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
24943         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
24944         ((match_operand:VF_128 2 "<round_nimm_predicate>"
24945         "<round_constraint>")): Changed to ...
24946         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
24947         "<round_scalar_constraint>")): ... this.
24948         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
24949         %0, %1, %2<round_op3>}"): Changed to ...
24950         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
24951         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
24952         %2<round_scalar_mask_op3>}"): ... this.
24953         * config/i386/subst.md (round_scalar_nimm_predicate): New.
24955 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24957         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
24958         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
24959         (_mm_maskz_sqrt_round_ss): New intrinsics.
24960         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
24961         (__builtin_ia32_sqrtsd_mask_round)
24962         (__builtin_ia32_sqrtss_mask_round): New builtins.
24963         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
24964         (__builtin_ia32_sqrtss_round): Remove.
24965         (__builtin_ia32_sqrtsd_mask_round)
24966         (__builtin_ia32_sqrtss_mask_round): New builtins.
24967         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
24968         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
24969         ((match_operand:VF_128 1 "vector_operand"
24970         "xBm,<round_constraint>")): Changed to ...
24971         ((match_operand:VF_128 1 "vector_operand"
24972         "xBm,<round_scalar_constraint>")): ... this.
24973         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
24974         %0, %2, %<iptr>1<round_op3>}): Changed to ...
24975         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
24976         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
24977         %<iptr>1<round_scalar_mask_op3>}): ... this.
24978         ((set_attr "prefix" "<round_prefix>")): Changed to ...
24979         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
24981 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
24983         PR target/84266
24984         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
24985         Cast vec_cmpeq result to correct type.
24986         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
24987         Cast vec_cmpgt result to correct type.
24989 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
24991         * final.c (final_scan_insn_1): Renamed from...
24992         (final_scan_insn): ... this.  New wrapper, to recover
24993         seen from the outermost call in recursive ones.
24994         * config/sparc/sparc.c (output_return): Drop seen from call.
24995         (output_sibcall): Likewise.
24996         * config/visium/visium.c (output_branch): Likewise.
24998 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25000         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25001         function label.
25003 2018-02-10  Alan Modra  <amodra@gmail.com>
25005         PR target/84300
25006         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25007         Specify LR as an input.
25009 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25011         PR sanitizer/83987
25012         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25013         remove_member_access_dummy_vars): New functions.
25014         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25015         lower_omp_1, execute_lower_omp): Use them.
25017         PR rtl-optimization/84308
25018         * shrink-wrap.c (spread_components): Release todo vector.
25020 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25022         PR rtl-optimization/57193
25023         * ira-color.c (struct allocno_color_data): Add member
25024         conflict_allocno_hard_prefs.
25025         (update_conflict_allocno_hard_prefs): New.
25026         (bucket_allocno_compare_func): Add a preference based on
25027         conflict_allocno_hard_prefs.
25028         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25029         (color_allocnos): Remove a dead code.  Initiate
25030         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25032 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25034         PR target/84226
25035         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25036         constraint from =wa to wa.  Avoid a subreg on the output operand,
25037         instead use a pseudo and subreg it in a move.
25038         (p9_xxbrd_<mode>): Changed to ...
25039         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25040         (p9_xxbrd_v2df): New expander.
25041         (p9_xxbrw_<mode>): Changed to ...
25042         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25043         (p9_xxbrw_v4sf): New expander.
25045 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25047         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25049 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25051         PR target/83926
25052         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25053         multiply in 32-bit mode.
25054         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25055         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25056         mode.
25058 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25060         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25061         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25062         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25063         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25065 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25067         PR lto/84213
25068         * dwarf2out.c (is_trivial_indirect_ref): New function.
25069         (dwarf2out_late_global_decl): Do not generate a location
25070         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25071         and that are not defined in the current unit.
25073 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25075         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25076         instead of a libcall for UNORDERED.
25078 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25080         PR target/82641
25081         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25082         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25084 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25086         PR target/PR84295
25087         * config/s390/s390.c (s390_set_current_function): Invoke
25088         s390_indirect_branch_settings also if fndecl didn't change.
25090 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25092         * config/rs6000/rs6000.md (blockage): Set length to zero.
25094 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25096         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25098 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25100         PR sanitizer/84285
25101         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25102         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25103         -static-lib*san.
25105         PR debug/84252
25106         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25107         PARALLEL incoming that failed vt_get_decl_and_offset check.
25109         PR middle-end/84237
25110         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25111         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25112         TREE_READONLY bit.
25113         (get_variable_section): For decls in named .bss* sections pass true as
25114         second argument to bss_initializer_p.
25116 2018-02-09  Marek Polacek  <polacek@redhat.com>
25117             Jakub Jelinek  <jakub@redhat.com>
25119         PR c++/83659
25120         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25121         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25122         Sync some changes from cxx_fold_indirect_ref.
25124 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25126         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25127         markers.
25128         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25129         (BLOCK_INLINE_ENTRY_LABEL): New.
25130         (dwarf2out_var_location): Disregard inline entry markers.
25131         (inline_entry_data): New struct.
25132         (inline_entry_data_hasher): New hashtable type.
25133         (inline_entry_data_hasher::hash): New.
25134         (inline_entry_data_hasher::equal): New.
25135         (inline_entry_data_table): New variable.
25136         (add_high_low_attributes): Add DW_AT_entry_pc and
25137         DW_AT_GNU_entry_view attributes if a pending entry is found
25138         in inline_entry_data_table.  Add old entry_pc attribute only
25139         if debug nonbinding markers are disabled.
25140         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25141         markers are enabled.
25142         (block_within_block_p, dwarf2out_inline_entry): New.
25143         (dwarf2out_finish): Check that no entries remained in
25144         inline_entry_data_table.
25145         * final.c (reemit_insn_block_notes): Handle inline entry notes.
25146         (final_scan_insn, notice_source_line): Likewise.
25147         (rest_of_clean_state): Skip inline entry markers.
25148         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25149         markers.
25150         * gimple.c (gimple_build_debug_inline_entry): New.
25151         * gimple.h (enum gimple_debug_subcode): Add
25152         GIMPLE_DEBUG_INLINE_ENTRY.
25153         (gimple_build_debug_inline_entry): Declare.
25154         (gimple_debug_inline_entry_p): New.
25155         (gimple_debug_nonbind_marker_p): Adjust.
25156         * insn-notes.def (INLINE_ENTRY): New.
25157         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25158         inline entry marker notes.
25159         (print_insn): Likewise.
25160         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25161         (INSN_DEBUG_MARKER_KIND): Likewise.
25162         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25163         * tree-inline.c (expand_call_inline): Build and insert
25164         debug_inline_entry stmt.
25165         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25166         inline entry blocks early, if nonbind markers are enabled.
25167         (dump_scope_block): Dump fragment info.
25168         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25169         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25170         (gimple_build_debug_inline_entry): New.
25171         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25172         Enable/disable inline entry points too.
25173         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25174         (DEBUG_INSN): Describe inline entry markers.
25176         * common.opt (gvariable-location-views): New.
25177         (gvariable-location-views=incompat5): New.
25178         * config.in: Rebuilt.
25179         * configure: Rebuilt.
25180         * configure.ac: Test assembler for view support.
25181         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25182         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25183         * dwarf2out.c (var_loc_view): New typedef.
25184         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25185         (dwarf2out_locviews_in_attribute): New.
25186         (dwarf2out_locviews_in_loclist): New.
25187         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25188         (enum dw_line_info_opcode): Add LI_adv_address.
25189         (struct dw_line_info_table): Add view.
25190         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25191         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25192         (zero_view_p): New variable.
25193         (ZERO_VIEW_P): New macro.
25194         (output_asm_line_debug_info): New.
25195         (struct var_loc_node): Add view.
25196         (add_AT_view_list, AT_loc_list): New.
25197         (add_var_loc_to_decl): Add view param.  Test it against last.
25198         (new_loc_list): Add view params.  Record them.
25199         (AT_loc_list_ptr): Handle loc and view lists.
25200         (view_list_to_loc_list_val_node): New.
25201         (print_dw_val): Handle dw_val_class_view_list.
25202         (size_of_die): Likewise.
25203         (value_format): Likewise.
25204         (loc_list_has_views): New.
25205         (gen_llsym): Set vl_symbol too.
25206         (maybe_gen_llsym, skip_loc_list_entry): New.
25207         (dwarf2out_maybe_output_loclist_view_pair): New.
25208         (output_loc_list): Output view list or entries too.
25209         (output_view_list_offset): New.
25210         (output_die): Handle dw_val_class_view_list.
25211         (output_dwarf_version): New.
25212         (output_compilation_unit_header): Use it.
25213         (output_skeleton_debug_sections): Likewise.
25214         (output_rnglists, output_line_info): Likewise.
25215         (output_pubnames, output_aranges): Update version comments.
25216         (output_one_line_info_table): Output view numbers in asm comments.
25217         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25218         Call maybe_gen_llsym.
25219         (loc_list_from_tree_1): Adjust.
25220         (add_AT_location_description): Create view list attribute if
25221         needed, check it's absent otherwise.
25222         (convert_cfa_to_fb_loc_list): Adjust.
25223         (maybe_emit_file): Call output_asm_line_debug_info for test.
25224         (dwarf2out_var_location): Reset views as needed.  Precompute
25225         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25226         attribute.  Set view.
25227         (new_line_info_table): Reset next view.
25228         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25229         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25230         the assembler, or select appropriate line info opcodes.
25231         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25232         (optimize_string_length): Catch it.  Adjust.
25233         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25234         dw_val_class_view_list, and remove it if no longer needed.
25235         (hash_loc_list): Hash view numbers.
25236         (loc_list_hasher::equal): Compare them.
25237         (optimize_location_lists): Check whether a view list symbol is
25238         needed, and whether the locview attribute is present, and
25239         whether they match.  Remove the locview attribute if no longer
25240         needed.
25241         (index_location_lists): Call skip_loc_list_entry for test.
25242         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25243         Use output_dwarf_version.
25244         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25245         (struct dw_val_node): Add val_view_list.
25246         * final.c (SEEN_NEXT_VIEW): New.
25247         (set_next_view_needed): New.
25248         (clear_next_view_needed): New.
25249         (maybe_output_next_view): New.
25250         (final_start_function): Rename to...
25251         (final_start_function_1): ... this.  Take pointer to FIRST,
25252         add SEEN parameter.  Emit param bindings in the initial view.
25253         (final_start_function): Reintroduce SEEN-less interface.
25254         (final): Rename to...
25255         (final_1): ... this.  Take SEEN parameter.  Output final pending
25256         next view at the end.
25257         (final): Reintroduce seen-less interface.
25258         (final_scan_insn): Output pending next view before switching
25259         sections or ending a block.  Mark the next view as needed when
25260         outputting variable locations.  Notify debug backend of section
25261         changes, and of location view changes.
25262         (rest_of_handle_final): Adjust.
25263         * toplev.c (process_options): Autodetect value for debug variable
25264         location views option.  Warn on incompat5 without -gdwarf-5.
25265         * doc/invoke.texi (gvariable-location-views): New.
25266         (gvariable-location-views=incompat5): New.
25267         (gno-variable-location-views): New.
25269 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
25271         PR tree-optimization/84136
25272         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25273         that the result of find_edge is non-NULL.
25275 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25277         PR target/83008
25278         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25279         storing integer register in SImode.  Fix cost of 256 and 512
25280         byte aligned SSE register store.
25282 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25284         * config/i386/i386.c (ix86_multiplication_cost): Fix
25285         multiplication cost for TARGET_AVX512DQ.
25287 2018-02-08  Marek Polacek  <polacek@redhat.com>
25289         PR tree-optimization/84238
25290         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25291         get_range_strlen.
25293 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25295         PR tree-optimization/84265
25296         * tree-vect-stmts.c (vectorizable_store): Don't treat
25297         VMAT_CONTIGUOUS accesses as grouped.
25298         (vectorizable_load): Likewise.
25300 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25302         PR tree-optimization/81635
25303         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25304         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25305         (test_round_for_mask): New functions.
25306         (wide_int_cc_tests): Call test_round_for_mask.
25307         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25308         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25309         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25310         range returned by get_range_info.
25312 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25314         PR ipa/81360
25315         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25316         * symtab.c: Include builtins.h
25317         (symtab_node::output_to_lto_symbol_table_p): Move here
25318         from lto-streamer-out.c:output_symbol_p.
25319         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25320         (output_symbol_p): Move all logic to symtab.c
25321         (produce_symtab): Update.
25323 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25325         * config/s390/s390-opts.h (enum indirect_branch): Define.
25326         * config/s390/s390-protos.h (s390_return_addr_from_memory)
25327         (s390_indirect_branch_via_thunk)
25328         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25329         (enum s390_indirect_branch_type): Define.
25330         * config/s390/s390.c (struct s390_frame_layout, struct
25331         machine_function): Remove.
25332         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25333         (indirect_branch_table_label_no, indirect_branch_table_name):
25334         Define variables.
25335         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25336         (enum s390_indirect_branch_option): Define.
25337         (s390_return_addr_from_memory): New function.
25338         (s390_handle_string_attribute): New function.
25339         (s390_attribute_table): Add new attribute handler.
25340         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25341         (s390_indirect_branch_via_thunk): New function.
25342         (s390_indirect_branch_via_inline_thunk): New function.
25343         (s390_function_ok_for_sibcall): When jumping via thunk disallow
25344         sibling call optimization for non z10 compiles.
25345         (s390_emit_call): Force indirect branch target to be a single
25346         register.  Add r1 clobber for non-z10 compiles.
25347         (s390_emit_epilogue): Emit return jump via return_use expander.
25348         (s390_reorg): Handle JUMP_INSNs as execute targets.
25349         (s390_option_override_internal): Perform validity checks for the
25350         new command line options.
25351         (s390_indirect_branch_attrvalue): New function.
25352         (s390_indirect_branch_settings): New function.
25353         (s390_set_current_function): Invoke s390_indirect_branch_settings.
25354         (s390_output_indirect_thunk_function):  New function.
25355         (s390_code_end): Implement target hook.
25356         (s390_case_values_threshold): Implement target hook.
25357         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25358         macros.
25359         * config/s390/s390.h (struct s390_frame_layout)
25360         (struct machine_function): Move here from s390.c.
25361         (TARGET_INDIRECT_BRANCH_NOBP_RET)
25362         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25363         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25364         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25365         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25366         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25367         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25368         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25369         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25370         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25371         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25372         (mnemonic attribute): Add values which aren't recognized
25373         automatically.
25374         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25375         pattern for branch conversion.  Fix mnemonic attribute.
25376         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25377         indirect branch via thunk if requested.
25378         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25379         ("*indirect_jump"): Disable for branch conversion using out of
25380         line thunks.
25381         ("indirect_jump_via_thunk<mode>_z10")
25382         ("indirect_jump_via_thunk<mode>")
25383         ("indirect_jump_via_inlinethunk<mode>_z10")
25384         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25385         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25386         ("casesi_jump_via_inlinethunk<mode>_z10")
25387         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25388         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25389         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25390         ("*indirect2_jump"): Disable for branch conversion.
25391         ("casesi_jump"): Turn into expander and expand patterns for branch
25392         conversion.
25393         ("return_use"): New expander.
25394         ("*return"): Emit return via thunk and rename it to ...
25395         ("*return<mode>"): ... this one.
25396         * config/s390/s390.opt: Add new options and and enum for the
25397         option values.
25399 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25401         * lra-constraints.c (match_reload): Unconditionally use
25402         gen_lowpart_SUBREG, rather than selecting between that
25403         and equivalent gen_rtx_SUBREG code.
25405 2018-02-08  Richard Biener  <rguenther@suse.de>
25407         PR tree-optimization/84233
25408         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25409         changed flag instead of boguously re-using phi_inserted.
25411 2018-02-08  Martin Jambor  <mjambor@suse.cz>
25413         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25414         static local variables.
25416 2018-02-08  Richard Biener  <rguenther@suse.de>
25418         PR tree-optimization/84278
25419         * tree-vect-stmts.c (vectorizable_store): When looking for
25420         smaller vector types to perform grouped strided loads/stores
25421         make sure the mode is supported by the target.
25422         (vectorizable_load): Likewise.
25424 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25426         * config/aarch64/aarch64.c (aarch64_components_for_bb):
25427         Increase LDP/STP opportunities by adding adjacent callee-saves.
25429 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25431         PR rtl-optimization/84068
25432         PR rtl-optimization/83459
25433         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25435 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
25437         PR tree-optimization/84224
25438         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25439         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25440         non-zero arguments.
25442 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
25444         PR target/84113
25445         * config/rs6000/altivec.md (*restore_world): Remove LR use.
25446         * config/rs6000/predicates.md (restore_world_operation): Adjust op
25447         count, remove one USE.
25449 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25451         * doc/install.texi (Configuration): Document the
25452         --with-long-double-format={ibm,ieee} PowerPC configuration
25453         options.
25455         PR target/84154
25456         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25457         Convert from define_expand to be define_insn_and_split.  Rework
25458         float/double/_Float128 conversions to QI/HI/SImode to work with
25459         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
25460         conversions to QI/HImode types did a store and then a load to
25461         truncate the value.  For conversions to VSX registers, don't split
25462         the insn, instead emit the code directly.  Use the code iterator
25463         any_fix to combine signed and unsigned conversions.
25464         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25465         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25466         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25467         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25468         (fix_<mode>di2_hw): Likewise.
25469         (fixuns_<mode>di2_hw): Likewise.
25470         (fix_<mode>si2_hw): Likewise.
25471         (fixuns_<mode>si2_hw): Likewise.
25472         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25473         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25474         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25475         fix<uns>_trunc<SFDF:mode>si2_p8.
25476         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25477         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25478         (fix<uns>_<mode>_mem): Likewise.
25479         (fctiw<u>z_<mode>_mem): Likewise.
25480         (fix<uns>_<mode>_mem): Likewise.
25481         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25482         the register allocator from doing a direct move to the GPRs to do
25483         a store, and instead use the ISA 3.0 store byte/half-word from
25484         vector register instruction.  For IEEE 128-bit floating point,
25485         also optimize stores of 32-bit ints.
25486         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25488 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
25490         * genextract.c (push_pathstr_operand): New function to support
25491         [a-zA-Z].
25492         (walk_rtx): Call push_pathstr_operand.
25493         (print_path): Support [a-zA-Z].
25495 2018-02-07  Richard Biener  <rguenther@suse.de>
25497         PR tree-optimization/84037
25498         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25499         (cse_and_gimplify_to_preheader): Declare.
25500         (vect_get_place_in_interleaving_chain): Likewise.
25501         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25502         ivexpr_map.
25503         (_loop_vec_info::~_loop_vec_info): Delete it.
25504         (cse_and_gimplify_to_preheader): New function.
25505         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25506         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25507         (vectorizable_load): Likewise.  For grouped stores always base
25508         the IV on the first element.
25509         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25510         condition before gimplifying.
25512 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25514         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25515         *DIV_EXPR and *MOD_EXPR.
25517 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
25519         PR target/84248
25520         * config/i386/i386.c (ix86_option_override_internal): Mask out
25521         the CF_SET bit when checking -fcf-protection.
25523 2018-02-07  Tom de Vries  <tom@codesourcery.com>
25525         PR libgomp/84217
25526         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
25527         enough.
25529 2018-02-07  Richard Biener  <rguenther@suse.de>
25531         PR tree-optimization/84204
25532         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
25533         this place.
25535         PR tree-optimization/84205
25536         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
25537         special-case isl_ast_op_zdiv_r.
25539         PR tree-optimization/84223
25540         * graphite-scop-detection.c (gather_bbs::before_dom_children):
25541         Only add conditions from within the region.
25542         (gather_bbs::after_dom_children): Adjust.
25544 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
25546         PR target/84209
25547         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
25548         * config/avr/avr.md: Only post-reload split REG-REG moves if
25549         either register is GENERAL_REG_P.
25551 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25553         PR tree-optimization/84235
25554         * tree-ssa-scopedtables.c
25555         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
25556         if the subtraction is performed in floating point type where NaNs are
25557         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
25558         build 1.  Formatting fix.
25560 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
25562         PR target/84146
25563         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
25564         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
25565         and skip it regardless of bb boundaries.  Use CALL_P macro,
25566         don't test INSN_P (insn) together with CALL_P or JUMP_P check
25567         unnecessarily, formatting fix.
25569 2018-02-06  Michael Collison  <michael.collison@arm.com>
25571         * config/arm/thumb2.md:
25572         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
25573         (*thumb_mov_notscc): Ditto.
25575 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
25577         PR target/84154
25578         * config/rs6000/rs6000.md (su code attribute): Use "u" for
25579         unsigned_fix, not "s".
25581 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25583         * configure.ac (gcc_fn_eh_frame_ro): New function.
25584         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
25585         correct .eh_frame permissions.
25586         * configure: Regenerate.
25588 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
25590         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
25591         irrelevant options.
25593 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25595         * config/rs6000/rs6000.c (rs6000_option_override_internal):
25596         Display warning message for -mno-speculate-indirect-jumps.
25598 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
25600         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
25601         Undocumented.
25602         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
25604 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
25606         PR tree-optimization/84225
25607         * tree-eh.c (find_trapping_overflow): Only call
25608         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
25610 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25612         PR target/84145
25613         * config/i386/i386.c: Reimplement the check of possible options
25614         -mibt/-mshstk conbination. Change error messages.
25615         * doc/invoke.texi: Fix a typo: remove extra '='.
25617 2018-02-06  Marek Polacek  <polacek@redhat.com>
25619         PR tree-optimization/84228
25620         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
25622 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
25624         PR target/82641
25625         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
25626         emitted arch directives.
25627         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
25628         __ARM_FEATURE_COPROC before changing architectures.
25630 2018-02-06  Richard Biener  <rguenther@suse.de>
25632         * config/i386/i386.c (print_reg): Fix typo.
25633         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
25635 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
25637         * configure: Regenerate.
25639 2018-02-05  Martin Sebor  <msebor@redhat.com>
25641         PR tree-optimization/83369
25642         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
25643         inlining context.
25645 2018-02-05  Martin Liska  <mliska@suse.cz>
25647         * doc/invoke.texi: Cherry-pick upstream r323995.
25649 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
25651         * ira.c (ira_init_register_move_cost): Adjust comment.
25653 2018-02-05  Martin Liska  <mliska@suse.cz>
25655         PR gcov-profile/84137
25656         * doc/gcov.texi: Fix typo in documentation.
25658 2018-02-05  Martin Liska  <mliska@suse.cz>
25660         PR gcov-profile/83879
25661         * doc/gcov.texi: Document necessity of --dynamic-list-data when
25662         using dlopen functionality.
25664 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
25666         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
25667         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
25668         _mm_maskz_range_ss, _mm_mask_range_round_ss,
25669         _mm_maskz_range_round_ss): New intrinsics.
25670         (__builtin_ia32_rangesd128_round)
25671         (__builtin_ia32_rangess128_round): Remove.
25672         (__builtin_ia32_rangesd128_mask_round,
25673         __builtin_ia32_rangess128_mask_round): New builtins.
25674         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
25675         __builtin_ia32_rangess128_round): Remove.
25676         (__builtin_ia32_rangesd128_mask_round,
25677         __builtin_ia32_rangess128_mask_round): New builtins.
25678         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
25679         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
25680         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
25681         "<round_saeonly_constraint>")): Changed to ...
25682         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
25683         "<round_saeonly_scalar_constraint>")): ... this.
25684         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25685         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
25686         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
25687         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
25688         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
25690 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
25692         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
25693         options.
25694         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
25695         Remove all values except native, 8540 and 8548.
25697 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
25699         * config/i386/i386.c (ix86_output_function_return): Pass
25700         INVALID_REGNUM, instead of -1, as invalid register number to
25701         indirect_thunk_name and output_indirect_thunk.
25703 2018-02-02  Julia Koval  <julia.koval@intel.com>
25705         * config.gcc: Add -march=icelake.
25706         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
25707         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
25708         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
25709         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
25710         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
25711         (processor_target_table): Add icelake.
25712         (ix86_option_override_internal): Handle new PTAs.
25713         (get_builtin_code_for_version): Handle icelake.
25714         (M_INTEL_COREI7_ICELAKE): New.
25715         (fold_builtin_cpu): Handle icelake.
25716         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
25717         * doc/invoke.texi: Add -march=icelake.
25719 2018-02-02  Julia Koval  <julia.koval@intel.com>
25721         * config/i386/i386.c (ix86_option_override_internal): Change flags type
25722         to wide_int_bitmask.
25723         * wide-int-bitmask.h: New.
25725 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25727         PR target/84066
25728         * config/i386/i386.md: Replace Pmode with word_mode in
25729         builtin_setjmp_setup and builtin_longjmp to support x32.
25731 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
25733         PR target/56010
25734         PR target/83743
25735         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
25736         #include "opts.h".
25737         (rs6000_supported_cpu_names): New static variable.
25738         (linux_cpu_translation_table): Likewise.
25739         (elf_platform) <cpu>: Define new static variable and use it.
25740         Translate kernel AT_PLATFORM name to canonical name if needed.
25741         Error if platform name is unknown.
25743 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
25745         PR target/84089
25746         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
25748 2018-02-01  Jeff Law  <law@redhat.com>
25750         PR target/84128
25751         * config/i386/i386.c (release_scratch_register_on_entry): Add new
25752         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
25753         the scratch if RELEASE_VIA_POP is false.
25754         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
25755         If we have to save a temporary register, decrement SIZE appropriately.
25756         Pass new arguments to release_scratch_register_on_entry.
25757         (ix86_adjust_stack_and_probe): Likewise.
25758         (ix86_emit_probe_stack_range): Pass new arguments to
25759         release_scratch_register_on_entry.
25761 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
25763         PR rtl-optimization/84157
25764         * combine.c (change_zero_ext): Use REG_P predicate in
25765         front of HARD_REGISTER_P predicate.
25767 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
25769         * config/avr/avr.c (avr_option_override): Move disabling of
25770         -fdelete-null-pointer-checks to...
25771         * common/config/avr/avr-common.c (avr_option_optimization_table):
25772         ...here.
25774 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25776         PR tree-optimization/81635
25777         * tree-data-ref.c (split_constant_offset_1): For types that
25778         wrap on overflow, try to use range info to prove that wrapping
25779         cannot occur.
25781 2018-02-01  Renlin Li  <renlin.li@arm.com>
25783         PR target/83370
25784         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
25785         TAILCALL_ADDR_REGS.
25786         (aarch64_register_move_cost): Likewise.
25787         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
25788         TAILCALL_ADDR_REGS.
25789         (REG_CLASS_NAMES): Likewise.
25790         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
25791         TAILCALL_ADDR_REGS. Remove IP registers.
25792         * config/aarch64/aarch64.md (Ucs): Update register constraint.
25794 2018-02-01  Richard Biener  <rguenther@suse.de>
25796         * domwalk.h (dom_walker::dom_walker): Add additional constructor
25797         for specifying RPO order and allow NULL for that.
25798         * domwalk.c (dom_walker::dom_walker): Likewise.
25799         (dom_walker::walk): Handle NULL RPO order.
25800         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
25801         in RPO order.
25802         (rewrite_update_dom_walker): Likewise.
25803         (mark_def_dom_walker): Likewise.
25805 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25807         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
25808         (aarch64_maybe_expand_sve_subreg_move): Declare.
25809         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
25810         * config/aarch64/predicates.md (aarch64_any_register_operand): New
25811         predicate.
25812         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
25813         that are semantically a reverse operation.
25814         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
25815         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
25816         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
25817         functions.
25818         (aarch64_can_change_mode_class): For big-endian, forbid changes
25819         between two SVE modes if they have different element sizes.
25821 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25823         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
25824         the TImode handling for big-endian targets.
25826 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25828         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
25829         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
25830         not just bytes.
25831         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
25832         Remove BSWAP handing for big-endian targets and use the form of
25833         LD1RQ appropariate for the mode.
25835 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25837         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
25838         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
25839         duplicated element.
25841 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25843         PR tearget/83845
25844         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
25845         check for operands that need to go through aarch64_sve_reload_be.
25847 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
25849         PR tree-optimization/81661
25850         PR tree-optimization/84117
25851         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
25852         * tree-eh.c: Include gimplify.h.
25853         (find_trapping_overflow, replace_trapping_overflow,
25854         rewrite_to_non_trapping_overflow): New functions.
25855         * tree-vect-loop.c: Include tree-eh.h.
25856         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
25857         * tree-data-ref.c: Include tree-eh.h.
25858         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
25860 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
25862         PR rtl-optimization/84123
25863         * combine.c (change_zero_ext): Check if hard register satisfies
25864         can_change_dest_mode before calling gen_lowpart_SUBREG.
25866 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
25868         PR target/82444
25869         * ira.c (ira_init_register_move_cost): Remove assert.
25871 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25873         PR rtl-optimization/84071
25874         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
25875         * doc/tm.texi: Regenerate.
25877 2018-01-31  Richard Biener  <rguenther@suse.de>
25879         PR tree-optimization/84132
25880         * tree-data-ref.c (analyze_miv_subscript): Properly
25881         check whether evolution_function_is_affine_multivariate_p
25882         before calling gcd_of_steps_may_divide_p.
25884 2018-01-31  Julia Koval  <julia.koval@intel.com>
25886         PR target/83618
25887         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
25888         * config/i386/i386.md (rdpid_rex64) New.
25889         (rdpid): Make 32bit only.
25891 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
25893         PR lto/84105
25894         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
25895         an IDENTIFIER_NODE for FUNCTION_TYPE's.
25897 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25899         Revert
25900         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25902         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25904 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25906         PR rtl-optimization/84071
25907         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
25908         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
25910 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25912         * config/arc/arc.c (arc_handle_aux_attribute): New function.
25913         (arc_attribute_table): Add 'aux' attribute.
25914         (arc_in_small_data_p): Consider aux like variables.
25915         (arc_is_aux_reg_p): New function.
25916         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
25917         (arc_get_aux_arg): New function.
25918         (prepare_move_operands): Handle aux-register access.
25919         (arc_handle_aux_attribute): New function.
25920         * doc/extend.texi (ARC Variable attributes): Add subsection.
25922 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25924         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
25925         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
25926         (arc_attribute_table): Add 'uncached' attribute.
25927         (arc_print_operand): Print '.di' flag for uncached memory
25928         accesses.
25929         (arc_in_small_data_p): Do not consider for small data the uncached
25930         types.
25931         (arc_is_uncached_mem_p): New function.
25932         * config/arc/predicates.md (compact_store_memory_operand): Check
25933         for uncached memory accesses.
25934         (nonvol_nonimm_operand): Likewise.
25935         * doc/extend.texi (ARC Type Attribute): New subsection.
25937 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
25939         PR c/84100
25940         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
25941         falign-loops=): Add Optimization flag.
25943 2018-01-30  Jeff Law  <law@redhat.com>
25945         PR target/84064
25946         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
25947         INT_REGISTERS_SAVED.  Check it prior to calling
25948         get_scratch_register_on_entry.
25949         (ix86_adjust_stack_and_probe): Similarly.
25950         (ix86_emit_probe_stack_range): Similarly.
25951         (ix86_expand_prologue): Corresponding changes.
25953 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25955         PR target/40411
25956         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
25957         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
25959 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
25961         PR target/84112
25962         * lra-constraints.c (curr_insn_transform): Process AND in the
25963         address.
25965 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
25967         PR rtl-optimization/83986
25968         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
25969         dependence against last_pending_memory_flush in addition to
25970         pending_jump_insns.
25972 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
25974         PR tree-optimization/81611
25975         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
25976         copies.
25978 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25980         PR target/83758
25981         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
25982         a reg rtx.
25984 2018-01-30  Richard Biener  <rguenther@suse.de>
25985             Jakub Jelinek  <jakub@redhat.com>
25987         PR tree-optimization/84111
25988         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
25989         inner loops added during recursion, as they don't have up-to-date
25990         SSA form.
25992 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
25994         PR ipa/81360
25995         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
25996         (can_inline_edge_by_limits_p): ... here.
25997         (can_early_inline_edge_p, check_callers,
25998         update_caller_keys, update_callee_keys, recursive_inlining,
25999         add_new_edges_to_heap, speculation_useful_p,
26000         inline_small_functions,
26001         inline_small_functions, flatten_function,
26002         inline_to_all_callers_1): Update.
26004 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26006         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26007         zeros correctly.
26009 2018-01-30  Richard Biener  <rguenther@suse.de>
26011         PR tree-optimization/83008
26012         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26013         invariant and constant vector uses in stmts when they need
26014         more than one stmt.
26016 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26018         PR bootstrap/84017
26019         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26020         * configure: Regenerate.
26022 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26024         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26025         pattern.
26026         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26027         Use gen_rtx_REG rather than gen_lowpart.
26029 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26031         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26032         rather than 0 when creating partial subregs.
26034 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26036         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26037         of usage.
26039 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26041         PR target/81550
26042         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26043         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26044         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26045         flags.  This restores the settings used before the 2017-07-24.
26046         Turning off pre increment/decrement/modify allows IVOPTS to
26047         optimize DF/SF loops where the index is an int.
26049 2018-01-29  Richard Biener  <rguenther@suse.de>
26050             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26052         PR bootstrap/80867
26053         * tree-vect-stmts.c (vectorizable_call): Don't call
26054         targetm.vectorize_builtin_md_vectorized_function if callee is
26055         NULL.
26057 2018-01-22  Carl Love  <cel@us.ibm.com>
26059         * doc/extend.tex: Fix typo in second arg in
26060         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26062 2018-01-29  Richard Biener  <rguenther@suse.de>
26064         PR tree-optimization/84086
26065         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26066         (flush_ssaname_freelist): When SSA names were released reset
26067         the SCEV hash table.
26069 2018-01-29  Richard Biener  <rguenther@suse.de>
26071         PR tree-optimization/84057
26072         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26073         removed paths when removing edges.
26075 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26077         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26078         -mfunction-return=@var{choice}.
26080 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26082         PR diagnostic/84034
26083         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26084         Handle CR like TAB.
26085         (layout::print_source_line): Likewise.
26086         (test_get_line_width_without_trailing_whitespace): Add test cases.
26088 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26090         PR middle-end/84040
26091         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26092         debug insns.
26094 2018-01-26  Jim Wilson  <jimw@sifive.com>
26096         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26098         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26099         specified.
26101 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26103         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26104         and CMP + SUB-immediate -> SUBS.
26106 2018-01-26  Martin Sebor  <msebor@redhat.com>
26108         PR tree-optimization/83896
26109         * tree-ssa-strlen.c (get_string_len): Rename...
26110         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26111         Avoid assuming length is constant.
26112         (handle_char_store): Use HOST_WIDE_INT for string length.
26114 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26116         PR target/81763
26117         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26118         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26120 2018-01-26  Richard Biener  <rguenther@suse.de>
26122         PR rtl-optimization/84003
26123         * dse.c (record_store): Only record redundant stores when
26124         the earlier store aliases at least all accesses the later one does.
26126 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26128         PR rtl-optimization/83985
26129         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26130         REG_CFA_RESTORE insns.
26131         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26132         REG_CFA_RESTORE insns here.
26134         PR c/83989
26135         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26136         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26138 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26140         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26141         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26142         (arc_init): Likewise.
26143         (arc_override_options): Likewise.
26144         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26145         value.
26146         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26147         support.
26148         * config/arc/arc.h (TARGET_DBNZ): Define.
26149         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26150         properly set the tune attribute.
26151         (dbnz): Use TARGET_DBNZ guard.
26152         * config/arc/arc.opt (mtune): Add core3 option.
26154 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26156         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26157         recognize new pic like addresses.
26158         (arc_delegitimize_address): Clean up.
26160 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26162         * config/arc/arc-arches.def: Option mrf16 valid for all
26163         architectures.
26164         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26165         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26166         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26167         * config/arc/arc-tables.opt: Regenerate.
26168         * config/arc/arc.c (arc_conditional_register_usage): Handle
26169         reduced register file case.
26170         (arc_file_start): Set must have build attributes.
26171         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26172         mrf16 option value.
26173         * config/arc/arc.opt (mrf16): Add new option.
26174         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26175         * config/arc/genmultilib.awk: Handle new mrf16 option.
26176         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26177         * config/arc/t-multilib: Regenerate.
26178         * doc/invoke.texi (ARC Options): Document mrf16 option.
26180 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26182         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26183         * config/arc/arc.c (arc_handle_secure_attribute): New function.
26184         (arc_attribute_table): Add 'secure_call' attribute.
26185         (arc_print_operand): Print secure call operand.
26186         (arc_function_ok_for_sibcall): Don't optimize tail calls when
26187         secure.
26188         (arc_is_secure_call_p): New function.  * config/arc/arc.md
26189         (call_i): Add support for sjli instruction.
26190         (call_value_i): Likewise.
26191         * config/arc/constraints.md (Csc): New constraint.
26193 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26194             John Eric Martin  <John.Martin@emmicro-us.com>
26196         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26197         * config/arc/arc.c (_arc_jli_section): New struct.
26198         (arc_jli_section): New type.
26199         (rc_jli_sections): New static variable.
26200         (arc_handle_jli_attribute): New function.
26201         (arc_attribute_table): Add jli_always and jli_fixed attribute.
26202         (arc_file_end): New function.
26203         (TARGET_ASM_FILE_END): Define.
26204         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26205         (arc_add_jli_section): New function.
26206         (jli_call_scan): Likewise.
26207         (arc_reorg): Call jli_call_scan.
26208         (arc_output_addsi): Remove 'S' from printing asm operand.
26209         (arc_is_jli_call_p): New function.
26210         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26211         operand.
26212         (movhi_insn): Likewise.
26213         (movsi_insn): Likewise.
26214         (movsi_set_cc_insn): Likewise.
26215         (loadqi_update): Likewise.
26216         (load_zeroextendqisi_update): Likewise.
26217         (load_signextendqisi_update): Likewise.
26218         (loadhi_update): Likewise.
26219         (load_zeroextendhisi_update): Likewise.
26220         (load_signextendhisi_update): Likewise.
26221         (loadsi_update): Likewise.
26222         (loadsf_update): Likewise.
26223         (movsicc_insn): Likewise.
26224         (bset_insn): Likewise.
26225         (bxor_insn): Likewise.
26226         (bclr_insn): Likewise.
26227         (bmsk_insn): Likewise.
26228         (bicsi3_insn): Likewise.
26229         (cmpsi_cc_c_insn): Likewise.
26230         (movsi_ne): Likewise.
26231         (movsi_cond_exec): Likewise.
26232         (clrsbsi2): Likewise.
26233         (norm_f): Likewise.
26234         (normw): Likewise.
26235         (swap): Likewise.
26236         (divaw): Likewise.
26237         (flag): Likewise.
26238         (sr): Likewise.
26239         (kflag): Likewise.
26240         (ffs): Likewise.
26241         (ffs_f): Likewise.
26242         (fls): Likewise.
26243         (call_i): Remove 'S' asm letter, add jli instruction.
26244         (call_value_i): Likewise.
26245         * config/arc/arc.op (mjli-always): New option.
26246         * config/arc/constraints.md (Cji): New constraint.
26247         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26248         operand.
26249         (subsf3_fpx): Likewise.
26250         (mulsf3_fpx): Likewise.
26251         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26252         asm operand.
26253         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26254         function attrbutes.
26255         * doc/invoke.texi (ARC): Document mjli-always option.
26257 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26259         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26260         avoid addition with 0 and use incw and decw where possible.
26262 2018-01-26  Richard Biener  <rguenther@suse.de>
26264         PR tree-optimization/81082
26265         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26266         association if it requires casting to unsigned.
26267         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26268         from fold_plusminus_mult_expr to catch important cases late when
26269         range info is available.
26271 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26273         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26274         * configure.ac (hidden_linkonce): New test.
26275         * configure: Regenerate.
26276         * config.in: Regenerate.
26278 2018-01-26  Julia Koval  <julia.koval@intel.com>
26280         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26281         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26282         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26283         _mm_mask_bitshuffle_epi64_mask): Fix type.
26284         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26285         USI_FTYPE_V4DI_V4DI_USI): Remove.
26286         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26287         __builtin_ia32_vpshufbitqmb256_mask,
26288         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26289         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26290         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26292 2018-01-26  Alan Modra  <amodra@gmail.com>
26294         PR target/84033
26295         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26296         UNSPEC_VBPERMQ.  Sort other unspecs.
26298 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
26300         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26302 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
26304         PR middle-end/83055
26305         * predict.c (drop_profile): Do not push/pop cfun; update also
26306         node->count.
26307         (handle_missing_profiles): Fix logic looking for zero profiles.
26309 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
26311         PR middle-end/83977
26312         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26313         on functions with #pragma omp declare simd or functions with simd
26314         attribute.
26315         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26316         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26317         Remove trailing \n from warning_at calls.
26319 2018-01-25  Tom de Vries  <tom@codesourcery.com>
26321         PR target/84028
26322         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26323         for neutered workers.
26325 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
26327         PR target/68467
26328         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26329         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26331 2018-01-24  Jeff Law  <law@redhat.com>
26333         PR target/83994
26334         * i386.c (get_probe_interval): Move to earlier point.
26335         (ix86_compute_frame_layout): If -fstack-clash-protection and
26336         the frame is larger than the probe interval, then use pushes
26337         to save registers rather than reg->mem moves.
26338         (ix86_expand_prologue): Remove conditional for int_registers_saved
26339         assertion.
26341 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
26343         PR target/84014
26344         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26345         min/max for never referenced object.
26347 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
26349         PR middle-end/83977
26350         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26351         here.
26352         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26353         attributes from DECL_ATTRIBUTES (decl) without affecting
26354         DECL_ATTRIBUTES (current_function_decl).
26355         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26356         functions with non-NULL DECL_ABSTRACT_ORIGIN.
26358 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
26360         PR tree-optimization/83979
26361         * fold-const.c (fold_comparison): Use constant_boolean_node
26362         instead of boolean_{true,false}_node.
26364 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
26366         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26367         with zero counts.
26369 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26371         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26372         Simplify the clause that sets the length attribute.
26373         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26374         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26375         clause that sets the length attribute.
26376         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26378 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26380         PR target/83589
26381         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26382         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26383         Add strict parameter.
26384         (prevent_branch_around_nothing): Insert dummy insn between branch to
26385         label and label with no ptx insn inbetween.
26386         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26388 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26390         PR target/81352
26391         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26392         for neutered threads in warp.
26393         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26395 2018-01-24  Richard Biener  <rguenther@suse.de>
26397         PR tree-optimization/83176
26398         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26399         operands.
26401 2018-01-24  Richard Biener  <rguenther@suse.de>
26403         PR tree-optimization/82819
26404         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26405         code generating pluses that are no-ops in the target precision.
26407 2018-01-24  Richard Biener  <rguenther@suse.de>
26409         PR middle-end/84000
26410         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26412 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26414         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26415         to merge probabilities.
26416         * predict.c (probably_never_executed): Also mark as cold functions
26417         with global 0 profile and guessed local profile.
26418         * profile-count.c (profile_probability::combine_with_count): New
26419         member function.
26420         * profile-count.h (profile_probability::operator*,
26421         profile_probability::operator*=, profile_probability::operator/,
26422         profile_probability::operator/=): Reduce precision to adjusted
26423         and set value to guessed on contradictory divisions.
26424         (profile_probability::combine_with_freq): Remove.
26425         (profile_probability::combine_wiht_count): Declare.
26426         (profile_count::force_nonzero):: Set to adjusted.
26427         (profile_count::probability_in):: Set quality to adjusted.
26428         * tree-ssa-tail-merge.c (replace_block_by): Use
26429         combine_with_count.
26431 2018-01-23  Andrew Waterman  <andrew@sifive.com>
26432             Jim Wilson  <jimw@sifive.com>
26434         * config/riscv/riscv.c (riscv_stack_boundary): New.
26435         (riscv_option_override): Set riscv_stack_boundary.  Handle
26436         riscv_preferred_stack_boundary_arg.
26437         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26438         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26439         (STACK_BOUNDARY): Set to riscv_stack_boundary.
26440         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26441         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26442         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26444 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
26446         PR target/83905
26447         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26448         of struct ix86_frame.
26449         (ix86_expand_epilogue): Likewise.  Add a local variable for
26450         the reg_save_offset field in struct ix86_frame.
26452 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
26454         PR tree-optimization/82604
26455         * tree-loop-distribution.c (enum partition_kind): New enum item
26456         PKIND_PARTIAL_MEMSET.
26457         (partition_builtin_p): Support above new enum item.
26458         (generate_code_for_partition): Ditto.
26459         (compute_access_range): Differentiate cases that equality can be
26460         proven at all loops, the innermost loops or no loops.
26461         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26462         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
26463         (finalize_partitions, distribute_loop): Don't fuse partition of
26464         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26465         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26466         parloop is enabled.
26468 2018-01-23  Martin Liska  <mliska@suse.cz>
26470         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26471         order to ignore the predictor.
26472         (PRED_POLYMORPHIC_CALL): Likewise.
26473         (PRED_RECURSIVE_CALL): Likewise.
26475 2018-01-23  Martin Liska  <mliska@suse.cz>
26477         * tree-profile.c (tree_profiling): Print function header to
26478         aware reader which function we are working on.
26479         * value-prof.c (gimple_find_values_to_profile): Do not print
26480         not interesting value histograms.
26482 2018-01-23  Martin Liska  <mliska@suse.cz>
26484         * profile-count.h (enum profile_quality): Add
26485         profile_uninitialized as the first value. Do not number values
26486         as they are zero based.
26487         (profile_count::verify): Update sanity check.
26488         (profile_probability::verify): Likewise.
26490 2018-01-23  Nathan Sidwell  <nathan@acm.org>
26492         * doc/invoke.texi (ffor-scope): Deprecate.
26494 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26496         PR tree-optimization/83510
26497         * domwalk.c (set_all_edges_as_executable): New function.
26498         (dom_walker::dom_walker): Convert bool param
26499         "skip_unreachable_blocks" to enum reachability.  Move setup of
26500         edge flags to set_all_edges_as_executable and only do it when
26501         reachability is REACHABLE_BLOCKS.
26502         * domwalk.h (enum dom_walker::reachability): New enum.
26503         (dom_walker::dom_walker): Convert bool param
26504         "skip_unreachable_blocks" to enum reachability.
26505         (set_all_edges_as_executable): New decl.
26506         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
26507         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26508         "reachability".
26509         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26510         but converting true to REACHABLE_BLOCKS.
26511         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26512         * tree-vrp.c
26513         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26514         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26515         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26516         REACHABLE_BLOCKS.
26517         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26518         if check_all_array_refs will be called.
26520 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26522         * tree.c (selftest::test_location_wrappers): Add more test
26523         coverage.
26525 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26527         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
26528         (selftest::test_bit_in_range): Likewise.
26530 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26532         PR testsuite/83888
26533         * doc/sourcebuild.texi (vect_float): Say that the selector
26534         only describes the situation when -funsafe-math-optimizations is on.
26535         (vect_float_strict): Document.
26537 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26539         PR tree-optimization/83965
26540         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
26541         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
26542         instead of checking only for a reduction.
26543         (vect_recog_widen_sum_pattern): Likewise.
26545 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26547         * predict.c (probably_never_executed): Only use precise profile info.
26548         (compute_function_frequency): Skip after inlining hack since we now
26549         have quality checking.
26551 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26553         * profile-count.h (profile_probability::very_unlikely,
26554         profile_probability::unlikely, profile_probability::even): Set
26555         precision to guessed.
26557 2018-01-23  Richard Biener  <rguenther@suse.de>
26559         PR tree-optimization/83963
26560         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
26561         Properly terminate dominator walk when crossing the exit edge not
26562         when visiting its source block.
26564 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
26566         PR c++/83918
26567         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
26568         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
26570 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
26572         PR tree-optimization/83957
26573         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
26574         semicolon after for body surrounded by braces.
26576         PR tree-optimization/83081
26577         * profile-count.h (profile_probability::split): New method.
26578         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
26579         Use profile_probability::split.
26580         (do_compare_rtx_and_jump): Fix adjustment of probabilities
26581         when splitting a single conditional jump into 2.
26583 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
26585         PR tree-optimization/69452
26586         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
26587         decl.
26589 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26591         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
26592         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
26593         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
26595 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26597         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
26598         declaration.
26599         * config/rl78/rl78.md (movdi): New define_expand.
26600         * config/rl78/rl78.c (rl78_split_movdi): New function.
26602 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
26604         PR target/83862
26605         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
26606         no longer used.
26607         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
26608         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
26609         128-bit to produce an UNSPEC move to get the double word with the
26610         signbit and then a shift directly to do signbit.
26611         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
26612         implementation with a new version that just does either a direct
26613         move or a regular move.  Move memory interface to separate insns.
26614         Move insns so they are next to the expander.
26615         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
26616         with signbit move.  Split big and little endian case.
26617         (signbit<mode>2_dm_mem_le): Likewise.
26618         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
26619         (signbit<mode>2_dm2): Likewise.
26621 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26623         * config/rl78/rl78.md (anddi3): New define_expand.
26625 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26627         * config/rl78/rl78.md (umindi3): New define_expand.
26629 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26631         * config/rl78/rl78.md (smindi3): New define_expand.
26633 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26635         * config/rl78/rl78.md (smaxdi3): New define_expand.
26637 2018-01-22  Carl Love  <cel@us.ibm.com>
26639         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
26640         LVX_V1TI): Add macro expansion.
26641         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
26642         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
26643         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
26644         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
26645         Change check to determine if the instruction is a byte reversing
26646         entry.  Fix typo in comment.
26647         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
26648         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
26649         Add def_builtin calls for new builtins.
26650         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
26651         Add define_insn expansion.
26653 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26655         * config/rl78/rl78.md (umaxdi3): New define_expand.
26657 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26659         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
26660         for non-QImode registers.
26662 2018-01-22  Richard Biener  <rguenther@suse.de>
26664         PR tree-optimization/83963
26665         * graphite-scop-detection.c (scop_detection::get_sese): Delay
26666         including the loop exit block.
26667         (scop_detection::merge_sese): Likewise.
26668         (scop_detection::add_scop): Do it here instead.
26670 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26672         * doc/sourcebuild.texi (arm_softfloat): Document.
26674 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
26676         PR gcc/77734
26677         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
26678         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
26679         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
26681 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26682             David Edelsohn  <dje.gcc@gmail.com>
26684         PR target/83946
26685         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26686         Change "crset eq" to "crset 2".
26687         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26688         (*call_indirect_aix<mode>_nospec): Likewise.
26689         (*call_value_indirect_aix<mode>_nospec): Likewise.
26690         (*call_indirect_elfv2<mode>_nospec): Likewise.
26691         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
26692         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
26693         change assembly output from . to $.
26694         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26695         (indirect_jump<mode>_nospec): Change assembly output from . to $.
26696         (*tablejump<mode>_internal1_nospec): Likewise.
26698 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
26700         PR target/80870
26701         * config/sh/sh_optimize_sett_clrt.cc:
26702         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
26704 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
26706         PR tree-optimization/83940
26707         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
26708         offset_dt to vect_constant_def rather than vect_unknown_def_type.
26709         (vect_check_load_store_mask): Add a mask_dt_out parameter and
26710         use it to pass back the definition type.
26711         (vect_check_store_rhs): Likewise rhs_dt_out.
26712         (vect_build_gather_load_calls): Add a mask_dt argument and use
26713         it instead of a call to vect_is_simple_use.
26714         (vectorizable_store): Update calls to vect_check_load_store_mask
26715         and vect_check_store_rhs.  Use the dt returned by the latter instead
26716         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
26717         instead of calls to vect_is_simple_use.  Pass the scalar rather
26718         than the vector operand to vect_is_simple_use when handling
26719         second and subsequent copies of an rhs value.
26720         (vectorizable_load): Update calls to vect_check_load_store_mask
26721         and vect_build_gather_load_calls.  Use the cached mask_dt and
26722         gs_info.offset_dt instead of calls to vect_is_simple_use.
26724 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
26726         PR middle-end/83945
26727         * tree-emutls.c: Include gimplify.h.
26728         (lower_emutls_2): New function.
26729         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
26730         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
26731         it before further processing.
26733         PR target/83930
26734         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
26735         UINTVAL (trueop1) instead of INTVAL (op1).
26737 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
26739         PR debug/81570
26740         PR debug/83728
26741         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
26742         INCOMING_FRAME_SP_OFFSET if not defined.
26743         (scan_trace): Add ENTRY argument.  If true and
26744         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
26745         emit a note to adjust the CFA offset.
26746         (create_cfi_notes): Adjust scan_trace callers.
26747         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
26748         INCOMING_FRAME_SP_OFFSET in the CIE.
26749         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
26750         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
26751         Likewise.
26752         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
26753         * doc/tm.texi: Regenerated.
26755 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26757         PR rtl-optimization/83147
26758         * lra-constraints.c (remove_inheritance_pseudos): Use
26759         lra_substitute_pseudo_within_insn.
26761 2018-01-19  Tom de Vries  <tom@codesourcery.com>
26762             Cesar Philippidis  <cesar@codesourcery.com>
26764         PR target/83920
26765         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
26767 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
26769         PR target/83790
26770         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
26771         spaces for function labels.
26773 2018-01-19  Martin Liska  <mliska@suse.cz>
26775         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
26776         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
26777         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
26778         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
26779         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
26780         (PRED_CONST_RETURN): Change from 69 to 65.
26781         (PRED_NULL_RETURN): Change from 91 to 71.
26782         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
26783         (PRED_LOOP_GUARD): Change from 66 to 73.
26785 2018-01-19  Martin Liska  <mliska@suse.cz>
26787         * predict.c (predict_insn_def): Add new assert.
26788         (struct branch_predictor): Change type to signed integer.
26789         (test_prediction_value_range): Amend test to cover
26790         PROB_UNINITIALIZED.
26791         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
26792         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
26793         (PRED_LOOP_ITERATIONS_MAX): Likewise.
26794         (PRED_LOOP_IV_COMPARE): Likewise.
26795         * predict.h (PROB_UNINITIALIZED): Define new constant.
26797 2018-01-19  Martin Liska  <mliska@suse.cz>
26799         * predict.c (dump_prediction): Add new format for
26800         analyze_brprob.py script which is enabled with -details
26801         suboption.
26802         * profile-count.h (precise_p): New function.
26804 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26806         PR tree-optimization/83922
26807         * tree-vect-loop.c (vect_verify_full_masking): Return false if
26808         there are no statements that need masking.
26809         (vect_active_double_reduction_p): New function.
26810         (vect_analyze_loop_operations): Use it when handling phis that
26811         are not in the loop header.
26813 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26815         PR tree-optimization/83914
26816         * tree-vect-loop.c (vectorizable_induction): Don't convert
26817         init_expr or apply the peeling adjustment for inductions
26818         that are nested within the vectorized loop.
26820 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26822         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
26823         instead of NEG.
26825 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26827         PR sanitizer/81715
26828         PR testsuite/83882
26829         * function.h (gimplify_parameters): Add gimple_seq * argument.
26830         * function.c: Include gimple.h and options.h.
26831         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
26832         for the added local temporaries if needed.
26833         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
26834         if there are any parameter cleanups, wrap whole body into a
26835         try/finally with the cleanups.
26837 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
26839         PR target/82964
26840         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
26841         Use GET_MODE_CLASS for scalar floating point.
26843 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26845         PR ipa/82256
26846         patch by PaX Team
26847         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
26848         Fix call of call_cgraph_insertion_hooks.
26850 2018-01-18  Martin Sebor  <msebor@redhat.com>
26852         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
26854 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26856         PR ipa/83619
26857         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
26858         frequencies.
26860 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
26862         PR other/70268
26863         * common.opt: (-ffile-prefix-map): New option.
26864         * opts.c (common_handle_option): Defer it.
26865         * opts-global.c (handle_common_deferred_options): Handle it.
26866         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
26867         * file-prefix-map.h: New file.
26868         (remap_debug_filename, add_debug_prefix_map): ...here.
26869         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
26870         * final.c (debug_prefix_map, add_debug_prefix_map
26871         remap_debug_filename): Move to...
26872         * file-prefix-map.c: New file.
26873         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
26874         generalize, get rid of alloca(), use strrchr() instead of strchr().
26875         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
26876         Implement in terms of add_prefix_map().
26877         (remap_macro_filename, remap_debug_filename): Implement in term of
26878         remap_filename().
26879         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
26880         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
26881         * dbxout.c: Include file-prefix-map.h.
26882         * varasm.c: Likewise.
26883         * vmsdbgout.c: Likewise.
26884         * xcoffout.c: Likewise.
26885         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
26886         * doc/cppopts.texi (-fmacro-prefix-map): Document.
26887         * doc/invoke.texi (-ffile-prefix-map): Document.
26888         (-fdebug-prefix-map): Update description.
26890 2018-01-18  Martin Liska  <mliska@suse.cz>
26892         * config/i386/i386.c (indirect_thunk_name): Document that also
26893         lfence is emitted.
26894         (output_indirect_thunk): Document why both instructions
26895         (pause and lfence) are generated.
26897 2018-01-18  Richard Biener  <rguenther@suse.de>
26899         PR tree-optimization/83887
26900         * graphite-scop-detection.c
26901         (scop_detection::get_nearest_dom_with_single_entry): Remove.
26902         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
26903         (scop_detection::merge_sese): Re-implement with a flood-fill
26904         algorithm that properly finds a SESE region if it exists.
26906 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26908         PR c/61240
26909         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
26910         pointer_diff optimizations use view_convert instead of convert.
26912 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26914         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26915         Generate different code for -mno-speculate-indirect-jumps.
26916         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26917         (*call_indirect_aix<mode>): Disable for
26918         -mno-speculate-indirect-jumps.
26919         (*call_indirect_aix<mode>_nospec): New define_insn.
26920         (*call_value_indirect_aix<mode>): Disable for
26921         -mno-speculate-indirect-jumps.
26922         (*call_value_indirect_aix<mode>_nospec): New define_insn.
26923         (*sibcall_nonlocal_sysv<mode>): Generate different code for
26924         -mno-speculate-indirect-jumps.
26925         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26927 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
26929         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
26930         long double type, set the flags for noting the default long double
26931         type, even if we don't pass or return a long double type.
26933 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
26935         PR ipa/83051
26936         * ipa-inline.c (flatten_function): Do not overwrite final inlining
26937         failure.
26939 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
26941         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
26942         support for merge[hl].
26943         (fold_mergehl_helper): New helper function.
26944         (tree-vector-builder.h): New #include for tree_vector_builder usage.
26945         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
26946         (altivec_vmrglw_direct): Add xxmrglw insn.
26948 2018-01-17  Andrew Waterman  <andrew@sifive.com>
26950         * config/riscv/riscv.c (riscv_conditional_register_usage): If
26951         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
26953 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
26955         PR lto/83121
26956         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
26957         call the lto_location_cache before reading the
26958         DECL_SOURCE_LOCATION of the types.
26960 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
26961             Richard Sandiford  <richard.sandiford@linaro.org>
26963         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
26964         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
26965         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
26966         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
26967         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
26968         Add declaration.
26969         * config/aarch64/constraints.md (aarch64_movti_operand):
26970         Limit immediates.
26971         * config/aarch64/predicates.md (Uti): Add new constraint.
26973 2018-01-17  Carl Love  <cel@us.ibm.com>
26975         * config/rs6000/vsx.md (define_expand xl_len_r,
26976         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
26977         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
26978         lxvll.
26979         (define_expand, define_insn): Move the shift left from  the
26980         define_insn to the define_expand for lxvl and stxvl instructions.
26981         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
26982         and XL_LEN_R definitions to PURE.
26984 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
26986         * config/i386/i386.c (indirect_thunk_name): Declare regno
26987         as unsigned int.  Compare regno with INVALID_REGNUM.
26988         (output_indirect_thunk): Ditto.
26989         (output_indirect_thunk_function): Ditto.
26990         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
26991         in the call to output_indirect_thunk_function.
26993 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
26995         PR middle-end/83884
26996         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
26997         rather than the size of inner_type to determine the stack slot size
26998         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27000 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27002         PR target/83546
27003         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27004         to PTA_SILVERMONT.
27006 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27008         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27009         endian Linux systems to optionally enable multilibs for selecting
27010         the long double type if the user configured an explicit type.
27011         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27012         have no long double multilibs if not defined.
27013         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27014         warn if the user used -mabi={ieee,ibm}longdouble and we built
27015         multilibs for long double.
27016         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27017         appropriate multilib option.
27018         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27019         multilib options.
27020         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27021         for building long double multilibs.
27022         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27024 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27026         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27027         copies.
27029         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27030         64 bits.
27031         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27032         128 bits.
27034         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27035         variables.
27037         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27038         return value.
27040 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27042         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27043         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27045 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27047         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27048         different rtl trees depending on TARGET_64BIT.
27049         (rs6000_gen_lvx): Likewise.
27051 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27053         * config/visium/visium.md (nop): Tweak comment.
27054         (hazard_nop): Likewise.
27056 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27058         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27059         -mspeculate-indirect-jumps.
27060         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27061         for -mno-speculate-indirect-jumps.
27062         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27063         (*call_value_indirect_elfv2<mode>): Disable for
27064         -mno-speculate-indirect-jumps.
27065         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27066         (indirect_jump): Emit different RTL for
27067         -mno-speculate-indirect-jumps.
27068         (*indirect_jump<mode>): Disable for
27069         -mno-speculate-indirect-jumps.
27070         (*indirect_jump<mode>_nospec): New define_insn.
27071         (tablejump): Emit different RTL for
27072         -mno-speculate-indirect-jumps.
27073         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27074         (tablejumpsi_nospec): New define_expand.
27075         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27076         (tablejumpdi_nospec): New define_expand.
27077         (*tablejump<mode>_internal1): Disable for
27078         -mno-speculate-indirect-jumps.
27079         (*tablejump<mode>_internal1_nospec): New define_insn.
27080         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27081         option.
27083 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27085         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27086         callers updated.
27088 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27089             Richard Biener  <rguenth@suse.de>
27091         PR libgomp/83590
27092         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27093         return early, inline manually is_gimple_sizepos.  Make sure if we
27094         call gimplify_expr we don't end up with a gimple constant.
27095         * tree.c (variably_modified_type_p): Don't return true for
27096         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27097         * gimplify.h (is_gimple_sizepos): Remove.
27099 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27101         PR tree-optimization/83857
27102         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27103         vectorizable_live_operation for pure SLP statements.
27104         (vectorizable_live_operation): Handle PHIs.
27106 2018-01-16  Richard Biener  <rguenther@suse.de>
27108         PR tree-optimization/83867
27109         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27110         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27112 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27114         PR c/83844
27115         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27116         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27117         If off is not INTEGER_CST, issue a may not be aligned warning
27118         rather than isn't aligned.  Use isn%'t rather than isn't.
27119         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27120         into MULT_EXPR.
27121         <case MULT_EXPR>: Improve the case when bottom and one of the
27122         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27123         operand, in that case check if the other operand is multiple of
27124         bottom divided by the INTEGER_CST operand.
27126 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27128         PR target/83858
27129         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27130         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27131         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27132         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27133         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27134         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27135         (pa_function_arg_size): New function.
27137 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27139         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27140         in a separate statement.
27142 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27144         PR tree-optimization/83847
27145         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27146         group gathers and scatters.
27148 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27150         PR rtl-optimization/86620
27151         * params.def (max-sched-ready-insns): Bump minimum value to 1.
27153         PR rtl-optimization/83213
27154         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27155         to last if both are JUMP_INSNs.
27157         PR tree-optimization/83843
27158         * gimple-ssa-store-merging.c
27159         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27160         store_immediate_info for bswap/nop orig_stores.
27162 2018-01-15  Andrew Waterman  <andrew@sifive.com>
27164         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27165         !TARGET_MUL.
27166         <UDIV>: Increase cost if !TARGET_DIV.
27168 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
27170         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27171         (define_attr "cr_logical_3op"): New.
27172         (cceq_ior_compare): Adjust.
27173         (cceq_ior_compare_complement): Adjust.
27174         (*cceq_rev_compare): Adjust.
27175         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27176         (is_cracked_insn): Adjust.
27177         (insn_must_be_first_in_group): Adjust.
27178         * config/rs6000/40x.md: Adjust.
27179         * config/rs6000/440.md: Adjust.
27180         * config/rs6000/476.md: Adjust.
27181         * config/rs6000/601.md: Adjust.
27182         * config/rs6000/603.md: Adjust.
27183         * config/rs6000/6xx.md: Adjust.
27184         * config/rs6000/7450.md: Adjust.
27185         * config/rs6000/7xx.md: Adjust.
27186         * config/rs6000/8540.md: Adjust.
27187         * config/rs6000/cell.md: Adjust.
27188         * config/rs6000/e300c2c3.md: Adjust.
27189         * config/rs6000/e500mc.md: Adjust.
27190         * config/rs6000/e500mc64.md: Adjust.
27191         * config/rs6000/e5500.md: Adjust.
27192         * config/rs6000/e6500.md: Adjust.
27193         * config/rs6000/mpc.md: Adjust.
27194         * config/rs6000/power4.md: Adjust.
27195         * config/rs6000/power5.md: Adjust.
27196         * config/rs6000/power6.md: Adjust.
27197         * config/rs6000/power7.md: Adjust.
27198         * config/rs6000/power8.md: Adjust.
27199         * config/rs6000/power9.md: Adjust.
27200         * config/rs6000/rs64.md: Adjust.
27201         * config/rs6000/titan.md: Adjust.
27203 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27205         * config/i386/predicates.md (indirect_branch_operand): Rewrite
27206         ix86_indirect_branch_register logic.
27208 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27210         * config/i386/constraints.md (Bs): Update
27211         ix86_indirect_branch_register check.  Don't check
27212         ix86_indirect_branch_register with GOT_memory_operand.
27213         (Bw): Likewise.
27214         * config/i386/predicates.md (GOT_memory_operand): Don't check
27215         ix86_indirect_branch_register here.
27216         (GOT32_symbol_operand): Likewise.
27218 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27220         * config/i386/predicates.md (constant_call_address_operand):
27221         Rewrite ix86_indirect_branch_register logic.
27222         (sibcall_insn_operand): Likewise.
27224 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27226         * config/i386/constraints.md (Bs): Replace
27227         ix86_indirect_branch_thunk_register with
27228         ix86_indirect_branch_register.
27229         (Bw): Likewise.
27230         * config/i386/i386.md (indirect_jump): Likewise.
27231         (tablejump): Likewise.
27232         (*sibcall_memory): Likewise.
27233         (*sibcall_value_memory): Likewise.
27234         Peepholes of indirect call and jump via memory: Likewise.
27235         * config/i386/i386.opt: Likewise.
27236         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27237         (GOT_memory_operand): Likewise.
27238         (call_insn_operand): Likewise.
27239         (sibcall_insn_operand): Likewise.
27240         (GOT32_symbol_operand): Likewise.
27242 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27244         PR middle-end/83837
27245         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27246         type rather than type addr's type points to.
27247         (expand_omp_atomic_mutex): Likewise.
27248         (expand_omp_atomic): Likewise.
27250 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27252         PR target/83839
27253         * config/i386/i386.c (output_indirect_thunk_function): Use
27254         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27255         for  __x86_return_thunk.
27257 2018-01-15  Richard Biener  <rguenther@suse.de>
27259         PR middle-end/83850
27260         * expmed.c (extract_bit_field_1): Fix typo.
27262 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27264         PR target/83687
27265         * config/arm/iterators.md (VF): New mode iterator.
27266         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27267         Remove integer-related logic from pattern.
27268         (neon_vabd<mode>_3): Likewise.
27270 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27272         PR middle-end/82694
27273         * common.opt (fstrict-overflow): No longer an alias.
27274         (fwrapv-pointer): New option.
27275         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27276         also for pointer types based on flag_wrapv_pointer.
27277         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27278         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27279         opts->x_flag_wrapv got set.
27280         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27281         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27282         POINTER_TYPE_OVERFLOW_UNDEFINED.
27283         * match.pd: Likewise in address comparison pattern.
27284         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27286 2018-01-15  Richard Biener  <rguenther@suse.de>
27288         PR lto/83804
27289         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27290         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
27291         Reset type names to their identifier if their TYPE_DECL doesn't
27292         have linkage (and thus is used for ODR and devirt).
27293         (save_debug_info_for_decl): Remove.
27294         (save_debug_info_for_type): Likewise.
27295         (add_tree_to_fld_list): Adjust.
27296         * tree-pretty-print.c (dump_generic_node): Make dumping of
27297         type names more robust.
27299 2018-01-15  Richard Biener  <rguenther@suse.de>
27301         * BASE-VER: Bump to 8.0.1.
27303 2018-01-14  Martin Sebor  <msebor@redhat.com>
27305         PR other/83508
27306         * builtins.c (check_access): Avoid warning when the no-warning bit
27307         is set.
27309 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
27311         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27312         * ira-color (allocno_hard_regs_compare): Likewise.
27314 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
27316         PR target/83013
27317         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27318         Use .pushsection/.popsection.
27320 2018-01-14  Martin Sebor  <msebor@redhat.com>
27322         PR c++/81327
27323         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27325 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
27327         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27328         entry from extra_headers.
27329         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27330         extra_headers, make the list bitwise identical to the i?86-*-* one.
27332 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27334         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27335         -mcmodel=large with -mindirect-branch=thunk,
27336         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27337         -mfunction-return=thunk-extern.
27338         * doc/invoke.texi: Document -mcmodel=large is incompatible with
27339         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27340         -mfunction-return=thunk and -mfunction-return=thunk-extern.
27342 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27344         * config/i386/i386.c (print_reg): Print the name of the full
27345         integer register without '%'.
27346         (ix86_print_operand): Handle 'V'.
27347         * doc/extend.texi: Document 'V' modifier.
27349 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27351         * config/i386/constraints.md (Bs): Disallow memory operand for
27352         -mindirect-branch-register.
27353         (Bw): Likewise.
27354         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27355         (GOT_memory_operand): Likewise.
27356         (call_insn_operand): Likewise.
27357         (sibcall_insn_operand): Likewise.
27358         (GOT32_symbol_operand): Likewise.
27359         * config/i386/i386.md (indirect_jump): Call convert_memory_address
27360         for -mindirect-branch-register.
27361         (tablejump): Likewise.
27362         (*sibcall_memory): Likewise.
27363         (*sibcall_value_memory): Likewise.
27364         Disallow peepholes of indirect call and jump via memory for
27365         -mindirect-branch-register.
27366         (*call_pop): Replace m with Bw.
27367         (*call_value_pop): Likewise.
27368         (*sibcall_pop_memory): Replace m with Bs.
27369         * config/i386/i386.opt (mindirect-branch-register): New option.
27370         * doc/invoke.texi: Document -mindirect-branch-register option.
27372 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27374         * config/i386/i386-protos.h (ix86_output_function_return): New.
27375         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27376         set function_return_type.
27377         (indirect_thunk_name): Add ret_p to indicate thunk for function
27378         return.
27379         (output_indirect_thunk_function): Pass false to
27380         indirect_thunk_name.
27381         (ix86_output_indirect_branch_via_reg): Likewise.
27382         (ix86_output_indirect_branch_via_push): Likewise.
27383         (output_indirect_thunk_function): Create alias for function
27384         return thunk if regno < 0.
27385         (ix86_output_function_return): New function.
27386         (ix86_handle_fndecl_attribute): Handle function_return.
27387         (ix86_attribute_table): Add function_return.
27388         * config/i386/i386.h (machine_function): Add
27389         function_return_type.
27390         * config/i386/i386.md (simple_return_internal): Use
27391         ix86_output_function_return.
27392         (simple_return_internal_long): Likewise.
27393         * config/i386/i386.opt (mfunction-return=): New option.
27394         (indirect_branch): Mention -mfunction-return=.
27395         * doc/extend.texi: Document function_return function attribute.
27396         * doc/invoke.texi: Document -mfunction-return= option.
27398 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27400         * config/i386/i386-opts.h (indirect_branch): New.
27401         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27402         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27403         with local indirect jump when converting indirect call and jump.
27404         (ix86_set_indirect_branch_type): New.
27405         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27406         (indirectlabelno): New.
27407         (indirect_thunk_needed): Likewise.
27408         (indirect_thunk_bnd_needed): Likewise.
27409         (indirect_thunks_used): Likewise.
27410         (indirect_thunks_bnd_used): Likewise.
27411         (INDIRECT_LABEL): Likewise.
27412         (indirect_thunk_name): Likewise.
27413         (output_indirect_thunk): Likewise.
27414         (output_indirect_thunk_function): Likewise.
27415         (ix86_output_indirect_branch_via_reg): Likewise.
27416         (ix86_output_indirect_branch_via_push): Likewise.
27417         (ix86_output_indirect_branch): Likewise.
27418         (ix86_output_indirect_jmp): Likewise.
27419         (ix86_code_end): Call output_indirect_thunk_function if needed.
27420         (ix86_output_call_insn): Call ix86_output_indirect_branch if
27421         needed.
27422         (ix86_handle_fndecl_attribute): Handle indirect_branch.
27423         (ix86_attribute_table): Add indirect_branch.
27424         * config/i386/i386.h (machine_function): Add indirect_branch_type
27425         and has_local_indirect_jump.
27426         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27427         to true.
27428         (tablejump): Likewise.
27429         (*indirect_jump): Use ix86_output_indirect_jmp.
27430         (*tablejump_1): Likewise.
27431         (simple_return_indirect_internal): Likewise.
27432         * config/i386/i386.opt (mindirect-branch=): New option.
27433         (indirect_branch): New.
27434         (keep): Likewise.
27435         (thunk): Likewise.
27436         (thunk-inline): Likewise.
27437         (thunk-extern): Likewise.
27438         * doc/extend.texi: Document indirect_branch function attribute.
27439         * doc/invoke.texi: Document -mindirect-branch= option.
27441 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
27443         PR ipa/83051
27444         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27446 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
27448         * ipa-inline.c (want_inline_small_function_p): Return false if
27449         inlining has already failed with CIF_FINAL_ERROR.
27450         (update_caller_keys): Call want_inline_small_function_p before
27451         can_inline_edge_p.
27452         (update_callee_keys): Likewise.
27454 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27456         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27457         New function.
27458         (rs6000_quadword_masked_address_p): Likewise.
27459         (quad_aligned_load_p): Likewise.
27460         (quad_aligned_store_p): Likewise.
27461         (const_load_sequence_p): Add comment to describe the outer-most loop.
27462         (mimic_memory_attributes_and_flags): New function.
27463         (rs6000_gen_stvx): Likewise.
27464         (replace_swapped_aligned_store): Likewise.
27465         (rs6000_gen_lvx): Likewise.
27466         (replace_swapped_aligned_load): Likewise.
27467         (replace_swapped_load_constant): Capitalize argument name in
27468         comment describing this function.
27469         (rs6000_analyze_swaps): Add a third pass to search for vector loads
27470         and stores that access quad-word aligned addresses and replace
27471         with stvx or lvx instructions when appropriate.
27472         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27473         New function prototype.
27474         (rs6000_quadword_masked_address_p): Likewise.
27475         (rs6000_gen_lvx): Likewise.
27476         (rs6000_gen_stvx): Likewise.
27477         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27478         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27479         when memory address is aligned.
27480         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27481         this split to select lvx instruction when memory address is aligned.
27482         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27483         instruction when memory address is aligned.
27484         (*vsx_le_perm_load_v16qi): Likewise.
27485         (four unnamed splitters): Modify to select the stvx instruction
27486         when memory is aligned.
27488 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
27490         * predict.c (determine_unlikely_bbs): Handle correctly BBs
27491         which appears in the queue multiple times.
27493 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27494             Alan Hayward  <alan.hayward@arm.com>
27495             David Sherwood  <david.sherwood@arm.com>
27497         * tree-vectorizer.h (vec_lower_bound): New structure.
27498         (_loop_vec_info): Add check_nonzero and lower_bounds.
27499         (LOOP_VINFO_CHECK_NONZERO): New macro.
27500         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27501         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27502         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27503         fields.  Make seg_len the distance travelled, not including the
27504         access size.
27505         (dr_direction_indicator): Declare.
27506         (dr_zero_step_indicator): Likewise.
27507         (dr_known_forward_stride_p): Likewise.
27508         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27509         tree-ssanames.h.
27510         (runtime_alias_check_p): Allow runtime alias checks with
27511         variable strides.
27512         (operator ==): Compare access_size and align.
27513         (prune_runtime_alias_test_list): Rework for new distinction between
27514         the access_size and seg_len.
27515         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
27516         segment lengths.
27517         (get_segment_min_max): New function.
27518         (create_intersect_range_checks): Use it.
27519         (dr_step_indicator): New function.
27520         (dr_direction_indicator): Likewise.
27521         (dr_zero_step_indicator): Likewise.
27522         (dr_known_forward_stride_p): Likewise.
27523         * tree-loop-distribution.c (data_ref_segment_size): Return
27524         DR_STEP * (niters - 1).
27525         (compute_alias_check_pairs): Update call to the dr_with_seg_len
27526         constructor.
27527         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
27528         (vect_preserves_scalar_order_p): New function, split out from...
27529         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
27530         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
27531         (vect_vfa_access_size): New function.
27532         (vect_vfa_align): Likewise.
27533         (vect_compile_time_alias): Take access_size_a and access_b arguments.
27534         (dump_lower_bound): New function.
27535         (vect_check_lower_bound): Likewise.
27536         (vect_small_gap_p): Likewise.
27537         (vectorizable_with_step_bound_p): Likewise.
27538         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
27539         depencies if the vectorization factor is 1.  Convert the checks
27540         for nonzero steps into checks on the bounds of DR_STEP.  Try using
27541         a bunds check for variable steps if the minimum required step is
27542         relatively small. Update calls to the dr_with_seg_len
27543         constructor and to vect_compile_time_alias.
27544         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
27545         function.
27546         (vect_loop_versioning): Call it.
27547         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
27548         when retrying.
27549         (vect_estimate_min_profitable_iters): Account for any bounds checks.
27551 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27552             Alan Hayward  <alan.hayward@arm.com>
27553             David Sherwood  <david.sherwood@arm.com>
27555         * doc/sourcebuild.texi (vect_scatter_store): Document.
27556         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
27557         optabs.
27558         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
27559         Document.
27560         * genopinit.c (main): Add supports_vec_scatter_store and
27561         supports_vec_scatter_store_cached to target_optabs.
27562         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
27563         IFN_MASK_SCATTER_STORE.
27564         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
27565         functions.
27566         * internal-fn.h (internal_store_fn_p): Declare.
27567         (internal_fn_stored_value_index): Likewise.
27568         * internal-fn.c (scatter_store_direct): New macro.
27569         (expand_scatter_store_optab_fn): New function.
27570         (direct_scatter_store_optab_supported_p): New macro.
27571         (internal_store_fn_p): New function.
27572         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
27573         IFN_MASK_SCATTER_STORE.
27574         (internal_fn_mask_index): Likewise.
27575         (internal_fn_stored_value_index): New function.
27576         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
27577         for scatter stores.
27578         * optabs-query.h (supports_vec_scatter_store_p): Declare.
27579         * optabs-query.c (supports_vec_scatter_store_p): New function.
27580         * tree-vectorizer.h (vect_get_store_rhs): Declare.
27581         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
27582         true for scatter stores.
27583         (vect_gather_scatter_fn_p): Handle scatter stores too.
27584         (vect_check_gather_scatter): Consider using scatter stores if
27585         supports_vec_scatter_store_p.
27586         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
27587         scatter stores too.
27588         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27589         internal_fn_stored_value_index.
27590         (check_load_store_masking): Handle scatter stores too.
27591         (vect_get_store_rhs): Make public.
27592         (vectorizable_call): Use internal_store_fn_p.
27593         (vectorizable_store): Handle scatter store internal functions.
27594         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
27595         when deciding whether the end of the group has been reached.
27596         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
27597         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
27598         (mask_scatter_store<mode>): New insns.
27600 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27601             Alan Hayward  <alan.hayward@arm.com>
27602             David Sherwood  <david.sherwood@arm.com>
27604         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
27605         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
27606         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
27607         function.
27608         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
27609         Use vect_truncate_gather_scatter_offset if we can't treat the
27610         operation as a normal gather load or scatter store.
27611         (get_group_load_store_type): Take the gather_scatter_info
27612         as argument.  Try using a gather load or scatter store for
27613         single-element groups.
27614         (get_load_store_type): Update calls to get_group_load_store_type
27615         and vect_use_strided_gather_scatters_p.
27617 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27618             Alan Hayward  <alan.hayward@arm.com>
27619             David Sherwood  <david.sherwood@arm.com>
27621         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
27622         optional tree argument.
27623         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
27624         null target hooks.
27625         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
27626         but continue to use the current value as a fallback.
27627         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
27628         to compare the updates.
27629         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
27630         (get_load_store_type): Use it when handling a strided access.
27631         (vect_get_strided_load_store_ops): New function.
27632         (vect_get_data_ptr_increment): Likewise.
27633         (vectorizable_load): Handle strided gather loads.  Always pass
27634         a step to vect_create_data_ref_ptr and bump_vector_ptr.
27636 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27637             Alan Hayward  <alan.hayward@arm.com>
27638             David Sherwood  <david.sherwood@arm.com>
27640         * doc/md.texi (gather_load@var{m}): Document.
27641         (mask_gather_load@var{m}): Likewise.
27642         * genopinit.c (main): Add supports_vec_gather_load and
27643         supports_vec_gather_load_cached to target_optabs.
27644         * optabs-tree.c (init_tree_optimization_optabs): Use
27645         ggc_cleared_alloc to allocate target_optabs.
27646         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
27647         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
27648         functions.
27649         * internal-fn.h (internal_load_fn_p): Declare.
27650         (internal_gather_scatter_fn_p): Likewise.
27651         (internal_fn_mask_index): Likewise.
27652         (internal_gather_scatter_fn_supported_p): Likewise.
27653         * internal-fn.c (gather_load_direct): New macro.
27654         (expand_gather_load_optab_fn): New function.
27655         (direct_gather_load_optab_supported_p): New macro.
27656         (direct_internal_fn_optab): New function.
27657         (internal_load_fn_p): Likewise.
27658         (internal_gather_scatter_fn_p): Likewise.
27659         (internal_fn_mask_index): Likewise.
27660         (internal_gather_scatter_fn_supported_p): Likewise.
27661         * optabs-query.c (supports_at_least_one_mode_p): New function.
27662         (supports_vec_gather_load_p): Likewise.
27663         * optabs-query.h (supports_vec_gather_load_p): Declare.
27664         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
27665         and memory_type field.
27666         (NUM_PATTERNS): Bump to 15.
27667         * tree-vect-data-refs.c: Include internal-fn.h.
27668         (vect_gather_scatter_fn_p): New function.
27669         (vect_describe_gather_scatter_call): Likewise.
27670         (vect_check_gather_scatter): Try using internal functions for
27671         gather loads.  Recognize existing calls to a gather load function.
27672         (vect_analyze_data_refs): Consider using gather loads if
27673         supports_vec_gather_load_p.
27674         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
27675         (vect_get_gather_scatter_offset_type): Likewise.
27676         (vect_convert_mask_for_vectype): Likewise.
27677         (vect_add_conversion_to_patterm): Likewise.
27678         (vect_try_gather_scatter_pattern): Likewise.
27679         (vect_recog_gather_scatter_pattern): New pattern recognizer.
27680         (vect_vect_recog_func_ptrs): Add it.
27681         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27682         internal_fn_mask_index and internal_gather_scatter_fn_p.
27683         (check_load_store_masking): Take the gather_scatter_info as an
27684         argument and handle gather loads.
27685         (vect_get_gather_scatter_ops): New function.
27686         (vectorizable_call): Check internal_load_fn_p.
27687         (vectorizable_load): Likewise.  Handle gather load internal
27688         functions.
27689         (vectorizable_store): Update call to check_load_store_masking.
27690         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
27691         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
27692         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
27693         (aarch64_gather_scale_operand_d): New predicates.
27694         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
27695         (mask_gather_load<mode>): New insns.
27697 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27698             Alan Hayward  <alan.hayward@arm.com>
27699             David Sherwood  <david.sherwood@arm.com>
27701         * optabs.def (fold_left_plus_optab): New optab.
27702         * doc/md.texi (fold_left_plus_@var{m}): Document.
27703         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
27704         * internal-fn.c (fold_left_direct): Define.
27705         (expand_fold_left_optab_fn): Likewise.
27706         (direct_fold_left_optab_supported_p): Likewise.
27707         * fold-const-call.c (fold_const_fold_left): New function.
27708         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
27709         * tree-parloops.c (valid_reduction_p): New function.
27710         (gather_scalar_reductions): Use it.
27711         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
27712         (vect_finish_replace_stmt): Declare.
27713         * tree-vect-loop.c (fold_left_reduction_fn): New function.
27714         (needs_fold_left_reduction_p): New function, split out from...
27715         (vect_is_simple_reduction): ...here.  Accept reductions that
27716         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
27717         (vect_force_simple_reduction): Also store the reduction type in
27718         the assignment's STMT_VINFO_REDUC_TYPE.
27719         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
27720         (merge_with_identity): New function.
27721         (vect_expand_fold_left): Likewise.
27722         (vectorize_fold_left_reduction): Likewise.
27723         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
27724         scalar phi in place for it.  Check for target support and reject
27725         cases that would reassociate the operation.  Defer the transform
27726         phase to vectorize_fold_left_reduction.
27727         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
27728         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
27729         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
27731 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27733         * tree-if-conv.c (predicate_mem_writes): Remove redundant
27734         call to ifc_temp_var.
27736 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27737             Alan Hayward  <alan.hayward@arm.com>
27738             David Sherwood  <david.sherwood@arm.com>
27740         * target.def (legitimize_address_displacement): Take the original
27741         offset as a poly_int.
27742         * targhooks.h (default_legitimize_address_displacement): Update
27743         accordingly.
27744         * targhooks.c (default_legitimize_address_displacement): Likewise.
27745         * doc/tm.texi: Regenerate.
27746         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
27747         as an argument, moving assert of ad->disp == ad->disp_term to...
27748         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
27749         Try calling targetm.legitimize_address_displacement before expanding
27750         the address rather than afterwards, and adjust for the new interface.
27751         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
27752         Match the new hook interface.  Handle SVE addresses.
27753         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
27754         new hook interface.
27756 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27758         * Makefile.in (OBJS): Add early-remat.o.
27759         * target.def (select_early_remat_modes): New hook.
27760         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
27761         * doc/tm.texi: Regenerate.
27762         * targhooks.h (default_select_early_remat_modes): Declare.
27763         * targhooks.c (default_select_early_remat_modes): New function.
27764         * timevar.def (TV_EARLY_REMAT): New timevar.
27765         * passes.def (pass_early_remat): New pass.
27766         * tree-pass.h (make_pass_early_remat): Declare.
27767         * early-remat.c: New file.
27768         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
27769         function.
27770         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
27772 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27773             Alan Hayward  <alan.hayward@arm.com>
27774             David Sherwood  <david.sherwood@arm.com>
27776         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
27777         vfm1 with a bound_epilog parameter.
27778         (vect_do_peeling): Update calls accordingly, and move the prologue
27779         call earlier in the function.  Treat the base bound_epilog as 0 for
27780         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
27781         this base when peeling for gaps.
27782         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
27783         with fully-masked loops.
27784         (vect_estimate_min_profitable_iters): Handle the single peeled
27785         iteration in that case.
27787 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27788             Alan Hayward  <alan.hayward@arm.com>
27789             David Sherwood  <david.sherwood@arm.com>
27791         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
27792         single-element interleaving even if the size is not a power of 2.
27793         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
27794         accesses for single-element interleaving if the group size is
27795         not a power of 2.
27797 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27798             Alan Hayward  <alan.hayward@arm.com>
27799             David Sherwood  <david.sherwood@arm.com>
27801         * doc/md.texi (fold_extract_last_@var{m}): Document.
27802         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
27803         * optabs.def (fold_extract_last_optab): New optab.
27804         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
27805         * internal-fn.c (fold_extract_direct): New macro.
27806         (expand_fold_extract_optab_fn): Likewise.
27807         (direct_fold_extract_optab_supported_p): Likewise.
27808         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
27809         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27810         EXTRACT_LAST_REDUCTION.
27811         (get_initial_def_for_reduction): Do not create an initial vector
27812         for EXTRACT_LAST_REDUCTION reductions.
27813         (vectorizable_reduction): Leave the scalar phi in place for
27814         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
27815         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
27816         epilogue code for EXTRACT_LAST_REDUCTION and defer the
27817         transform phase to vectorizable_condition.
27818         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
27819         split out from...
27820         (vect_finish_stmt_generation): ...here.
27821         (vect_finish_replace_stmt): New function.
27822         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
27823         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
27824         pattern.
27825         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
27827 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27828             Alan Hayward  <alan.hayward@arm.com>
27829             David Sherwood  <david.sherwood@arm.com>
27831         * doc/md.texi (extract_last_@var{m}): Document.
27832         * optabs.def (extract_last_optab): New optab.
27833         * internal-fn.def (EXTRACT_LAST): New internal function.
27834         * internal-fn.c (cond_unary_direct): New macro.
27835         (expand_cond_unary_optab_fn): Likewise.
27836         (direct_cond_unary_optab_supported_p): Likewise.
27837         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
27838         loops using EXTRACT_LAST.
27839         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
27840         (extract_last_<mode>): ...this optab.
27841         (vec_extract<mode><Vel>): Update accordingly.
27843 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27844             Alan Hayward  <alan.hayward@arm.com>
27845             David Sherwood  <david.sherwood@arm.com>
27847         * target.def (empty_mask_is_expensive): New hook.
27848         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
27849         * doc/tm.texi: Regenerate.
27850         * targhooks.h (default_empty_mask_is_expensive): Declare.
27851         * targhooks.c (default_empty_mask_is_expensive): New function.
27852         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
27853         if the target says that empty masks are expensive.
27854         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
27855         New function.
27856         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
27858 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27859             Alan Hayward  <alan.hayward@arm.com>
27860             David Sherwood  <david.sherwood@arm.com>
27862         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
27863         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
27864         (vect_use_loop_mask_for_alignment_p): New function.
27865         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
27866         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
27867         niters_skip argument.  Make sure that the first niters_skip elements
27868         of the first iteration are inactive.
27869         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
27870         Update call to vect_set_loop_masks_directly.
27871         (get_misalign_in_elems): New function, split out from...
27872         (vect_gen_prolog_loop_niters): ...here.
27873         (vect_update_init_of_dr): Take a code argument that specifies whether
27874         the adjustment should be added or subtracted.
27875         (vect_update_init_of_drs): Likewise.
27876         (vect_prepare_for_masked_peels): New function.
27877         (vect_do_peeling): Skip prologue peeling if we're using a mask
27878         instead.  Update call to vect_update_inits_of_drs.
27879         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27880         mask_skip_niters.
27881         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
27882         alignment.  Do not include the number of peeled iterations in
27883         the minimum threshold in that case.
27884         (vectorizable_induction): Adjust the start value down by
27885         LOOP_VINFO_MASK_SKIP_NITERS iterations.
27886         (vect_transform_loop): Call vect_prepare_for_masked_peels.
27887         Take the number of skipped iterations into account when calculating
27888         the loop bounds.
27889         * tree-vect-stmts.c (vect_gen_while_not): New function.
27891 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27892             Alan Hayward  <alan.hayward@arm.com>
27893             David Sherwood  <david.sherwood@arm.com>
27895         * doc/sourcebuild.texi (vect_fully_masked): Document.
27896         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
27897         default value to 0.
27898         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
27899         split out from...
27900         (vect_analyze_loop_2): ...here. Don't check the vectorization
27901         factor against the number of loop iterations if the loop is
27902         fully-masked.
27904 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27905             Alan Hayward  <alan.hayward@arm.com>
27906             David Sherwood  <david.sherwood@arm.com>
27908         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
27909         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
27910         (dump_groups): Update accordingly.
27911         (iv_use::mem_type): New member variable.
27912         (address_p): New function.
27913         (record_use): Add a mem_type argument and initialize the new
27914         mem_type field.
27915         (record_group_use): Add a mem_type argument.  Use address_p.
27916         Remove obsolete null checks of base_object.  Update call to record_use.
27917         (find_interesting_uses_op): Update call to record_group_use.
27918         (find_interesting_uses_cond): Likewise.
27919         (find_interesting_uses_address): Likewise.
27920         (get_mem_type_for_internal_fn): New function.
27921         (find_address_like_use): Likewise.
27922         (find_interesting_uses_stmt): Try find_address_like_use before
27923         calling find_interesting_uses_op.
27924         (addr_offset_valid_p): Use the iv mem_type field as the type
27925         of the addressed memory.
27926         (add_autoinc_candidates): Likewise.
27927         (get_address_cost): Likewise.
27928         (split_small_address_groups_p): Use address_p.
27929         (split_address_groups): Likewise.
27930         (add_iv_candidate_for_use): Likewise.
27931         (autoinc_possible_for_pair): Likewise.
27932         (rewrite_groups): Likewise.
27933         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
27934         (determine_group_iv_cost): Update after split of USE_ADDRESS.
27935         (get_alias_ptr_type_for_ptr_address): New function.
27936         (rewrite_use_address): Rewrite address uses in calls that were
27937         identified by find_address_like_use.
27939 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27940             Alan Hayward  <alan.hayward@arm.com>
27941             David Sherwood  <david.sherwood@arm.com>
27943         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
27944         TARGET_MEM_REFs.
27945         * gimple-expr.h (is_gimple_addressable: Likewise.
27946         * gimple-expr.c (is_gimple_address): Likewise.
27947         * internal-fn.c (expand_call_mem_ref): New function.
27948         (expand_mask_load_optab_fn): Use it.
27949         (expand_mask_store_optab_fn): Likewise.
27951 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27952             Alan Hayward  <alan.hayward@arm.com>
27953             David Sherwood  <david.sherwood@arm.com>
27955         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
27956         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
27957         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
27958         (cond_umax@var{mode}): Document.
27959         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
27960         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
27961         (cond_umin_optab, cond_umax_optab): New optabs.
27962         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
27963         (COND_IOR, COND_XOR): New internal functions.
27964         * internal-fn.h (get_conditional_internal_fn): Declare.
27965         * internal-fn.c (cond_binary_direct): New macro.
27966         (expand_cond_binary_optab_fn): Likewise.
27967         (direct_cond_binary_optab_supported_p): Likewise.
27968         (get_conditional_internal_fn): New function.
27969         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
27970         Cope with reduction statements that are vectorized as calls rather
27971         than assignments.
27972         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
27973         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
27974         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
27975         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
27976         (UNSPEC_COND_EOR): New unspecs.
27977         (optab): Add mappings for them.
27978         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
27979         (sve_int_op, sve_fp_op): New int attributes.
27981 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27982             Alan Hayward  <alan.hayward@arm.com>
27983             David Sherwood  <david.sherwood@arm.com>
27985         * optabs.def (while_ult_optab): New optab.
27986         * doc/md.texi (while_ult@var{m}@var{n}): Document.
27987         * internal-fn.def (WHILE_ULT): New internal function.
27988         * internal-fn.h (direct_internal_fn_supported_p): New override
27989         that takes two types as argument.
27990         * internal-fn.c (while_direct): New macro.
27991         (expand_while_optab_fn): New function.
27992         (convert_optab_supported_p): Likewise.
27993         (direct_while_optab_supported_p): New macro.
27994         * wide-int.h (wi::udiv_ceil): New function.
27995         * tree-vectorizer.h (rgroup_masks): New structure.
27996         (vec_loop_masks): New typedef.
27997         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
27998         and fully_masked_p.
27999         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28000         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28001         (vect_max_vf): New function.
28002         (slpeel_make_loop_iterate_ntimes): Delete.
28003         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28004         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28005         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28006         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28007         internal-fn.h, stor-layout.h and optabs-query.h.
28008         (vect_set_loop_mask): New function.
28009         (add_preheader_seq): Likewise.
28010         (add_header_seq): Likewise.
28011         (interleave_supported_p): Likewise.
28012         (vect_maybe_permute_loop_masks): Likewise.
28013         (vect_set_loop_masks_directly): Likewise.
28014         (vect_set_loop_condition_masked): Likewise.
28015         (vect_set_loop_condition_unmasked): New function, split out from
28016         slpeel_make_loop_iterate_ntimes.
28017         (slpeel_make_loop_iterate_ntimes): Rename to..
28018         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28019         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28020         (vect_do_peeling): Update call accordingly.
28021         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28022         loops.
28023         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28024         mask_compare_type, can_fully_mask_p and fully_masked_p.
28025         (release_vec_loop_masks): New function.
28026         (_loop_vec_info): Use it to free the loop masks.
28027         (can_produce_all_loop_masks_p): New function.
28028         (vect_get_max_nscalars_per_iter): Likewise.
28029         (vect_verify_full_masking): Likewise.
28030         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28031         retries, and free the mask rgroups before retrying.  Check loop-wide
28032         reasons for disallowing fully-masked loops.  Make the final decision
28033         about whether use a fully-masked loop or not.
28034         (vect_estimate_min_profitable_iters): Do not assume that peeling
28035         for the number of iterations will be needed for fully-masked loops.
28036         (vectorizable_reduction): Disable fully-masked loops.
28037         (vectorizable_live_operation): Likewise.
28038         (vect_halve_mask_nunits): New function.
28039         (vect_double_mask_nunits): Likewise.
28040         (vect_record_loop_mask): Likewise.
28041         (vect_get_loop_mask): Likewise.
28042         (vect_transform_loop): Handle the case in which the final loop
28043         iteration might handle a partial vector.  Call vect_set_loop_condition
28044         instead of slpeel_make_loop_iterate_ntimes.
28045         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28046         (check_load_store_masking): New function.
28047         (prepare_load_store_mask): Likewise.
28048         (vectorizable_store): Handle fully-masked loops.
28049         (vectorizable_load): Likewise.
28050         (supportable_widening_operation): Use vect_halve_mask_nunits for
28051         booleans.
28052         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28053         (vect_gen_while): New function.
28054         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28055         (aarch64_uqdec<mode>): New insn.
28057 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28058             Alan Hayward  <alan.hayward@arm.com>
28059             David Sherwood  <david.sherwood@arm.com>
28061         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28062         (reduc_xor_scal_optab): New optabs.
28063         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28064         (reduc_xor_scal_@var{m}): Document.
28065         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28066         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28067         internal functions.
28068         * fold-const-call.c (fold_const_call): Handle them.
28069         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28070         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28071         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28072         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28073         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28074         (UNSPEC_XORV): New unspecs.
28075         (optab): Add entries for them.
28076         (BITWISEV): New int iterator.
28077         (bit_reduc_op): New int attributes.
28079 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28080             Alan Hayward  <alan.hayward@arm.com>
28081             David Sherwood  <david.sherwood@arm.com>
28083         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28084         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28085         * optabs.def (vec_shl_insert_optab): New optab.
28086         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28087         (duplicate_and_interleave): Likewise.
28088         * tree-vect-loop.c: Include internal-fn.h.
28089         (neutral_op_for_slp_reduction): New function, split out from
28090         get_initial_defs_for_reduction.
28091         (get_initial_def_for_reduction): Handle option 2 for variable-length
28092         vectors by loading the neutral value into a vector and then shifting
28093         the initial value into element 0.
28094         (get_initial_defs_for_reduction): Replace the code argument with
28095         the neutral value calculated by neutral_op_for_slp_reduction.
28096         Use gimple_build_vector for constant-length vectors.
28097         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28098         but the first group_size elements have a neutral value.
28099         Use duplicate_and_interleave otherwise.
28100         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28101         Update call to get_initial_defs_for_reduction.  Handle SLP
28102         reductions for variable-length vectors by creating one vector
28103         result for each scalar result, with the elements associated
28104         with other scalar results stubbed out with the neutral value.
28105         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28106         Require IFN_VEC_SHL_INSERT for double reductions on
28107         variable-length vectors, or SLP reductions that have
28108         a neutral value.  Require can_duplicate_and_interleave_p
28109         support for variable-length unchained SLP reductions if there
28110         is no neutral value, such as for MIN/MAX reductions.  Also require
28111         the number of vector elements to be a multiple of the number of
28112         SLP statements when doing variable-length unchained SLP reductions.
28113         Update call to vect_create_epilog_for_reduction.
28114         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28115         and remove initial values.
28116         (duplicate_and_interleave): Make public.
28117         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28118         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28120 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28121             Alan Hayward  <alan.hayward@arm.com>
28122             David Sherwood  <david.sherwood@arm.com>
28124         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28125         (can_duplicate_and_interleave_p): New function.
28126         (vect_get_and_check_slp_defs): Take the vector of statements
28127         rather than just the current one.  Remove excess parentheses.
28128         Restriction rejectinon of vect_constant_def and vect_external_def
28129         for variable-length vectors to boolean types, or types for which
28130         can_duplicate_and_interleave_p is false.
28131         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28132         (duplicate_and_interleave): New function.
28133         (vect_get_constant_vectors): Use gimple_build_vector for
28134         constant-length vectors and suitable variable-length constant
28135         vectors.  Use duplicate_and_interleave for other variable-length
28136         vectors.  Don't defer the update when inserting new statements.
28138 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28139             Alan Hayward  <alan.hayward@arm.com>
28140             David Sherwood  <david.sherwood@arm.com>
28142         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28143         min_profitable_iters doesn't go negative.
28145 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28146             Alan Hayward  <alan.hayward@arm.com>
28147             David Sherwood  <david.sherwood@arm.com>
28149         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28150         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28151         * optabs.def (vec_mask_load_lanes_optab): New optab.
28152         (vec_mask_store_lanes_optab): Likewise.
28153         * internal-fn.def (MASK_LOAD_LANES): New internal function.
28154         (MASK_STORE_LANES): Likewise.
28155         * internal-fn.c (mask_load_lanes_direct): New macro.
28156         (mask_store_lanes_direct): Likewise.
28157         (expand_mask_load_optab_fn): Handle masked operations.
28158         (expand_mask_load_lanes_optab_fn): New macro.
28159         (expand_mask_store_optab_fn): Handle masked operations.
28160         (expand_mask_store_lanes_optab_fn): New macro.
28161         (direct_mask_load_lanes_optab_supported_p): Likewise.
28162         (direct_mask_store_lanes_optab_supported_p): Likewise.
28163         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28164         parameter.
28165         (vect_load_lanes_supported): Likewise.
28166         * tree-vect-data-refs.c (strip_conversion): New function.
28167         (can_group_stmts_p): Likewise.
28168         (vect_analyze_data_ref_accesses): Use it instead of checking
28169         for a pair of assignments.
28170         (vect_store_lanes_supported): Take a masked_p parameter.
28171         (vect_load_lanes_supported): Likewise.
28172         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28173         vect_store_lanes_supported and vect_load_lanes_supported.
28174         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28175         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28176         parameter.  Don't allow gaps for masked accesses.
28177         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
28178         and vect_load_lanes_supported.
28179         (get_load_store_type): Take a masked_p parameter and update
28180         call to get_group_load_store_type.
28181         (vectorizable_store): Update call to get_load_store_type.
28182         Handle IFN_MASK_STORE_LANES.
28183         (vectorizable_load): Update call to get_load_store_type.
28184         Handle IFN_MASK_LOAD_LANES.
28186 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28187             Alan Hayward  <alan.hayward@arm.com>
28188             David Sherwood  <david.sherwood@arm.com>
28190         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28191         modes for SVE.
28192         * config/aarch64/aarch64-protos.h
28193         (aarch64_sve_struct_memory_operand_p): Declare.
28194         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28195         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28196         (VPRED, vpred): Handle SVE structure modes.
28197         * config/aarch64/constraints.md (Utx): New constraint.
28198         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28199         (aarch64_sve_struct_nonimmediate_operand): New predicates.
28200         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28201         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28202         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28203         structure modes.  Split into pieces after RA.
28204         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28205         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28206         New patterns.
28207         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28208         SVE structure modes.
28209         (aarch64_classify_address): Likewise.
28210         (sizetochar): Move earlier in file.
28211         (aarch64_print_operand): Handle SVE register lists.
28212         (aarch64_array_mode): New function.
28213         (aarch64_sve_struct_memory_operand_p): Likewise.
28214         (TARGET_ARRAY_MODE): Redefine.
28216 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28217             Alan Hayward  <alan.hayward@arm.com>
28218             David Sherwood  <david.sherwood@arm.com>
28220         * target.def (array_mode): New target hook.
28221         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28222         * doc/tm.texi: Regenerate.
28223         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28224         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28225         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28226         targetm.array_mode.
28227         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28228         type sizes.
28230 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28231             Alan Hayward  <alan.hayward@arm.com>
28232             David Sherwood  <david.sherwood@arm.com>
28234         * fold-const.c (fold_binary_loc): Check the argument types
28235         rather than the result type when testing for a vector operation.
28237 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28239         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28240         * doc/tm.texi: Regenerate.
28242 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28243             Alan Hayward  <alan.hayward@arm.com>
28244             David Sherwood  <david.sherwood@arm.com>
28246         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28247         (sve): Document new AArch64 extension.
28248         * doc/md.texi (w): Extend the description of the AArch64
28249         constraint to include SVE vectors.
28250         (Upl, Upa): Document new AArch64 predicate constraints.
28251         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28252         enum.
28253         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28254         (msve-vector-bits=): New option.
28255         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28256         SVE when these are disabled.
28257         (sve): New extension.
28258         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28259         modes.  Adjust their number of units based on aarch64_sve_vg.
28260         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28261         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28262         aarch64_addr_query_type.
28263         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28264         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28265         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28266         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28267         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28268         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28269         (aarch64_simd_imm_zero_p): Delete.
28270         (aarch64_check_zero_based_sve_index_immediate): Declare.
28271         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28272         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28273         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28274         (aarch64_sve_float_mul_immediate_p): Likewise.
28275         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28276         rather than an rtx.
28277         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28278         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28279         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28280         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28281         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28282         (aarch64_regmode_natural_size): Likewise.
28283         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28284         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28285         left one place.
28286         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28287         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28288         for VG and the SVE predicate registers.
28289         (V_ALIASES): Add a "z"-prefixed alias.
28290         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28291         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28292         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28293         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28294         (REG_CLASS_NAMES): Add entries for them.
28295         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
28296         and the predicate registers.
28297         (aarch64_sve_vg): Declare.
28298         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28299         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28300         (REGMODE_NATURAL_SIZE): Define.
28301         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28302         SVE macros.
28303         * config/aarch64/aarch64.c: Include cfgrtl.h.
28304         (simd_immediate_info): Add a constructor for series vectors,
28305         and an associated step field.
28306         (aarch64_sve_vg): New variable.
28307         (aarch64_dbx_register_number): Handle VG and the predicate registers.
28308         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28309         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28310         (VEC_ANY_DATA, VEC_STRUCT): New constants.
28311         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28312         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28313         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28314         (aarch64_get_mask_mode): New functions.
28315         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28316         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28317         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
28318         predicate modes and predicate registers.  Explicitly restrict
28319         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
28320         to store a vector mode if it is recognized by
28321         aarch64_classify_vector_mode.
28322         (aarch64_regmode_natural_size): New function.
28323         (aarch64_hard_regno_caller_save_mode): Return the original mode
28324         for predicates.
28325         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28326         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28327         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28328         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28329         functions.
28330         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
28331         does not overlap dest if the function is frame-related.  Handle
28332         SVE constants.
28333         (aarch64_split_add_offset): New function.
28334         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28335         them aarch64_add_offset.
28336         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28337         and update call to aarch64_sub_sp.
28338         (aarch64_add_cfa_expression): New function.
28339         (aarch64_expand_prologue): Pass extra temporary registers to the
28340         functions above.  Handle the case in which we need to emit new
28341         DW_CFA_expressions for registers that were originally saved
28342         relative to the stack pointer, but now have to be expressed
28343         relative to the frame pointer.
28344         (aarch64_output_mi_thunk): Pass extra temporary registers to the
28345         functions above.
28346         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
28347         IP0 and IP1 values for SVE frames.
28348         (aarch64_expand_vec_series): New function.
28349         (aarch64_expand_sve_widened_duplicate): Likewise.
28350         (aarch64_expand_sve_const_vector): Likewise.
28351         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28352         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
28353         into the register, rather than emitting a SET directly.
28354         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28355         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28356         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28357         (offset_9bit_signed_scaled_p): New functions.
28358         (aarch64_replicate_bitmask_imm): New function.
28359         (aarch64_bitmask_imm): Use it.
28360         (aarch64_cannot_force_const_mem): Reject expressions involving
28361         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
28362         (aarch64_classify_index): Handle SVE indices, by requiring
28363         a plain register index with a scale that matches the element size.
28364         (aarch64_classify_address): Handle SVE addresses.  Assert that
28365         the mode of the address is VOIDmode or an integer mode.
28366         Update call to aarch64_classify_symbol.
28367         (aarch64_classify_symbolic_expression): Update call to
28368         aarch64_classify_symbol.
28369         (aarch64_const_vec_all_in_range_p): New function.
28370         (aarch64_print_vector_float_operand): Likewise.
28371         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28372         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28373         and the FP immediates 1.0 and 0.5.
28374         (aarch64_print_address_internal): Handle SVE addresses.
28375         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28376         (aarch64_regno_regclass): Handle predicate registers.
28377         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28378         data modes.
28379         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28380         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28381         (aarch64_convert_sve_vector_bits): New function.
28382         (aarch64_override_options): Use it to handle -msve-vector-bits=.
28383         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28384         rather than an rtx.
28385         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28386         Handle SVE vector and predicate modes.  Accept VL-based constants
28387         that need only one temporary register, and VL offsets that require
28388         no temporary registers.
28389         (aarch64_conditional_register_usage): Mark the predicate registers
28390         as fixed if SVE isn't available.
28391         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28392         Return true for SVE vector and predicate modes.
28393         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28394         rather than an unsigned int.  Handle SVE modes.
28395         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
28396         SVE modes.
28397         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28398         if SVE is enabled.
28399         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28400         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28401         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28402         (aarch64_sve_float_mul_immediate_p): New functions.
28403         (aarch64_sve_valid_immediate): New function.
28404         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28405         Explicitly reject structure modes.  Check for INDEX constants.
28406         Handle PTRUE and PFALSE constants.
28407         (aarch64_check_zero_based_sve_index_immediate): New function.
28408         (aarch64_simd_imm_zero_p): Delete.
28409         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28410         vector modes.  Accept constants in the range of CNT[BHWD].
28411         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28412         ask for an Advanced SIMD mode.
28413         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28414         (aarch64_simd_vector_alignment): Handle SVE predicates.
28415         (aarch64_vectorize_preferred_vector_alignment): New function.
28416         (aarch64_simd_vector_alignment_reachable): Use it instead of
28417         the vector size.
28418         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28419         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28420         functions.
28421         (MAX_VECT_LEN): Delete.
28422         (expand_vec_perm_d): Add a vec_flags field.
28423         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28424         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28425         (aarch64_evpc_ext): Don't apply a big-endian lane correction
28426         for SVE modes.
28427         (aarch64_evpc_rev): Rename to...
28428         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
28429         (aarch64_evpc_rev_global): New function.
28430         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28431         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28432         MAX_VECT_LEN.
28433         (aarch64_evpc_sve_tbl): New function.
28434         (aarch64_expand_vec_perm_const_1): Update after rename of
28435         aarch64_evpc_rev.  Handle SVE permutes too, trying
28436         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28437         than aarch64_evpc_tbl.
28438         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28439         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28440         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28441         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28442         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28443         (aarch64_expand_sve_vcond): New functions.
28444         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28445         of aarch64_vector_mode_p.
28446         (aarch64_dwarf_poly_indeterminate_value): New function.
28447         (aarch64_compute_pressure_classes): Likewise.
28448         (aarch64_can_change_mode_class): Likewise.
28449         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28450         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28451         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28452         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28453         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28454         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28455         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28456         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28457         constraints.
28458         (Dn, Dl, Dr): Accept const as well as const_vector.
28459         (Dz): Likewise.  Compare against CONST0_RTX.
28460         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28461         of "vector" where appropriate.
28462         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28463         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28464         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28465         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28466         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28467         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28468         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28469         (v_int_equiv): Extend to SVE modes.
28470         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28471         mode attributes.
28472         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28473         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28474         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28475         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28476         (SVE_COND_FP_CMP): New int iterators.
28477         (perm_hilo): Handle the new unpack unspecs.
28478         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28479         attributes.
28480         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28481         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28482         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28483         (aarch64_equality_operator, aarch64_constant_vector_operand)
28484         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28485         (aarch64_sve_nonimmediate_operand): Likewise.
28486         (aarch64_sve_general_operand): Likewise.
28487         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28488         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28489         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28490         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28491         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28492         (aarch64_sve_float_arith_immediate): Likewise.
28493         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28494         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28495         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28496         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28497         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28498         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28499         (aarch64_sve_float_arith_operand): Likewise.
28500         (aarch64_sve_float_arith_with_sub_operand): Likewise.
28501         (aarch64_sve_float_mul_operand): Likewise.
28502         (aarch64_sve_vec_perm_operand): Likewise.
28503         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28504         (aarch64_mov_operand): Accept const_poly_int and const_vector.
28505         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28506         as well as const_vector.
28507         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28508         in file.  Use CONST0_RTX and CONSTM1_RTX.
28509         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
28510         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28511         Use aarch64_simd_imm_zero.
28512         * config/aarch64/aarch64-sve.md: New file.
28513         * config/aarch64/aarch64.md: Include it.
28514         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28515         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28516         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28517         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28518         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28519         (sve): New attribute.
28520         (enabled): Disable instructions with the sve attribute unless
28521         TARGET_SVE.
28522         (movqi, movhi): Pass CONST_POLY_INT operaneds through
28523         aarch64_expand_mov_immediate.
28524         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28525         CNT[BHSD] immediates.
28526         (movti): Split CONST_POLY_INT moves into two halves.
28527         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
28528         Split additions that need a temporary here if the destination
28529         is the stack pointer.
28530         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
28531         (*add<mode>3_poly_1): New instruction.
28532         (set_clobber_cc): New expander.
28534 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28536         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
28537         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
28538         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
28539         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
28540         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
28541         Change innermode from fixed_mode_size to machine_mode.
28542         (simplify_subreg): Update call accordingly.  Handle a constant-sized
28543         subreg of a variable-length CONST_VECTOR.
28545 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28546             Alan Hayward  <alan.hayward@arm.com>
28547             David Sherwood  <david.sherwood@arm.com>
28549         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
28550         (add_offset_to_base): New function, split out from...
28551         (create_mem_ref): ...here.  When handling a scale other than 1,
28552         check first whether the address is valid without the offset.
28553         Add it into the base if so, leaving the index and scale as-is.
28555 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28557         PR c++/83778
28558         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
28559         fold_for_warn before checking if arg2 is INTEGER_CST.
28561 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
28563         * config/rs6000/predicates.md (load_multiple_operation): Delete.
28564         (store_multiple_operation): Delete.
28565         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
28566         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
28567         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
28568         guarded by TARGET_STRING.
28569         (rs6000_output_load_multiple): Delete.
28570         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28571         OPTION_MASK_STRING / TARGET_STRING handling.
28572         (print_operand) <'N', 'O'>: Add comment that these are unused now.
28573         (const rs6000_opt_masks) <"string">: Change mask to 0.
28574         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
28575         (MASK_STRING): Delete.
28576         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
28577         parts.  Simplify.
28578         (load_multiple): Delete.
28579         (*ldmsi8): Delete.
28580         (*ldmsi7): Delete.
28581         (*ldmsi6): Delete.
28582         (*ldmsi5): Delete.
28583         (*ldmsi4): Delete.
28584         (*ldmsi3): Delete.
28585         (store_multiple): Delete.
28586         (*stmsi8): Delete.
28587         (*stmsi7): Delete.
28588         (*stmsi6): Delete.
28589         (*stmsi5): Delete.
28590         (*stmsi4): Delete.
28591         (*stmsi3): Delete.
28592         (movmemsi_8reg): Delete.
28593         (corresponding unnamed define_insn): Delete.
28594         (movmemsi_6reg): Delete.
28595         (corresponding unnamed define_insn): Delete.
28596         (movmemsi_4reg): Delete.
28597         (corresponding unnamed define_insn): Delete.
28598         (movmemsi_2reg): Delete.
28599         (corresponding unnamed define_insn): Delete.
28600         (movmemsi_1reg): Delete.
28601         (corresponding unnamed define_insn): Delete.
28602         * config/rs6000/rs6000.opt (mno-string): New.
28603         (mstring): Replace by deprecation warning stub.
28604         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
28606 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28608         * regrename.c (regrename_do_replace): If replacing the same
28609         reg multiple times, try to reuse last created gen_raw_REG.
28611         PR debug/81155
28612         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
28613         main to workaround a bug in GDB.
28615 2018-01-12  Tom de Vries  <tom@codesourcery.com>
28617         PR target/83737
28618         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
28620 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
28622         PR rtl-optimization/80481
28623         * ira-color.c (get_cap_member): New function.
28624         (allocnos_conflict_by_live_ranges_p): Use it.
28625         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
28626         (setup_slot_coalesced_allocno_live_ranges): Ditto.
28628 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
28630         PR target/83628
28631         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
28632         (*saddl_se_1): Ditto.
28633         (*ssubsi_1): Ditto.
28634         (*ssubl_se_1): Ditto.
28636 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28638         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
28639         rather than wi::to_widest for DR_INITs.
28640         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
28641         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
28642         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
28643         INTEGER_CSTs.
28644         (vect_analyze_group_access_1): Note that here.
28646 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28648         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
28649         polynomial type sizes.
28651 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28653         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
28654         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
28655         (gimple_add_tmp_var): Likewise.
28657 2018-01-12  Martin Liska  <mliska@suse.cz>
28659         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
28660         (gimple_alloc_sizes): Likewise.
28661         (dump_gimple_statistics): Use PRIu64 in printf format.
28662         * gimple.h: Change uint64_t to int.
28664 2018-01-12  Martin Liska  <mliska@suse.cz>
28666         * tree-core.h: Use uint64_t instead of int.
28667         * tree.c (tree_node_counts): Likewise.
28668         (tree_node_sizes): Likewise.
28669         (dump_tree_statistics): Use PRIu64 in printf format.
28671 2018-01-12  Martin Liska  <mliska@suse.cz>
28673         * Makefile.in: As qsort_chk is implemented in vec.c, add
28674         vec.o to linkage of gencfn-macros.
28675         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
28676         passing the info to record_node_allocation_statistics.
28677         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
28678         and pass the info.
28679         * ggc-common.c (struct ggc_usage): Add operator== and use
28680         it in operator< and compare function.
28681         * mem-stats.h (struct mem_usage): Likewise.
28682         * vec.c (struct vec_usage): Remove operator< and compare
28683         function. Can be simply inherited.
28685 2018-01-12  Martin Jambor  <mjambor@suse.cz>
28687         PR target/81616
28688         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
28689         * tree-ssa-math-opts.c: Include domwalk.h.
28690         (convert_mult_to_fma_1): New function.
28691         (fma_transformation_info): New type.
28692         (fma_deferring_state): Likewise.
28693         (cancel_fma_deferring): New function.
28694         (result_of_phi): Likewise.
28695         (last_fma_candidate_feeds_initial_phi): Likewise.
28696         (convert_mult_to_fma): Added deferring logic, split actual
28697         transformation to convert_mult_to_fma_1.
28698         (math_opts_dom_walker): New type.
28699         (math_opts_dom_walker::after_dom_children): New method, body moved
28700         here from pass_optimize_widening_mul::execute, added deferring logic
28701         bits.
28702         (pass_optimize_widening_mul::execute): Moved most of code to
28703         math_opts_dom_walker::after_dom_children.
28704         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
28705         * config/i386/i386.c (ix86_option_override_internal): Added
28706         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
28708 2018-01-12  Richard Biener  <rguenther@suse.de>
28710         PR debug/83157
28711         * dwarf2out.c (gen_variable_die): Do not reset old_die for
28712         inline instance vars.
28714 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
28716         PR target/81819
28717         * config/rx/rx.c (rx_is_restricted_memory_address):
28718         Handle SUBREG case.
28720 2018-01-12  Richard Biener  <rguenther@suse.de>
28722         PR tree-optimization/80846
28723         * target.def (split_reduction): New target hook.
28724         * targhooks.c (default_split_reduction): New function.
28725         * targhooks.h (default_split_reduction): Declare.
28726         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
28727         target requests first reduce vectors by combining low and high
28728         parts.
28729         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
28730         (get_vectype_for_scalar_type_and_size): Export.
28731         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
28732         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
28733         * doc/tm.texi: Regenerate.
28734         * config/i386/i386.c (ix86_split_reduction): Implement
28735         TARGET_VECTORIZE_SPLIT_REDUCTION.
28737 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28739         PR target/83368
28740         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
28741         in PIC mode except for TARGET_VXWORKS_RTP.
28742         * config/sparc/sparc.c: Include cfgrtl.h.
28743         (TARGET_INIT_PIC_REG): Define.
28744         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
28745         (sparc_pic_register_p): New predicate.
28746         (sparc_legitimate_address_p): Use it.
28747         (sparc_legitimize_pic_address): Likewise.
28748         (sparc_delegitimize_address): Likewise.
28749         (sparc_mode_dependent_address_p): Likewise.
28750         (gen_load_pcrel_sym): Remove 4th parameter.
28751         (load_got_register): Adjust call to above.  Remove obsolete stuff.
28752         (sparc_expand_prologue): Do not call load_got_register here.
28753         (sparc_flat_expand_prologue): Likewise.
28754         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
28755         (sparc_use_pseudo_pic_reg): New function.
28756         (sparc_init_pic_reg): Likewise.
28757         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
28758         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
28760 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
28762         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28763         Add item for branch_cost.
28765 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28767         PR rtl-optimization/83565
28768         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
28769         not extend the result to a larger mode for rotate operations.
28770         (num_sign_bit_copies1): Likewise.
28772 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28774         PR target/40411
28775         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
28776         -symbolic.
28777         Use values-Xc.o for -pedantic.
28778         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
28780 2018-01-12  Martin Liska  <mliska@suse.cz>
28782         PR ipa/83054
28783         * ipa-devirt.c (final_warning_record::grow_type_warnings):
28784         New function.
28785         (possible_polymorphic_call_targets): Use it.
28786         (ipa_devirt): Likewise.
28788 2018-01-12  Martin Liska  <mliska@suse.cz>
28790         * profile-count.h (enum profile_quality): Use 0 as invalid
28791         enum value of profile_quality.
28793 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
28795         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
28796         -mext-string options.
28798 2018-01-12  Richard Biener  <rguenther@suse.de>
28800         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
28801         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
28802         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
28803         Likewise.
28804         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
28806 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
28808         * configure.ac (--with-long-double-format): Add support for the
28809         configuration option to change the default long double format on
28810         PowerPC systems.
28811         * config.gcc (powerpc*-linux*-*): Likewise.
28812         * configure: Regenerate.
28813         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
28814         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
28815         used without modification.
28817 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28819         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
28820         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
28821         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
28822         MISC_BUILTIN_SPEC_BARRIER.
28823         (rs6000_init_builtins): Likewise.
28824         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
28825         enum value.
28826         (speculation_barrier): New define_insn.
28827         * doc/extend.texi: Document __builtin_speculation_barrier.
28829 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28831         PR target/83203
28832         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
28833         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
28834         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
28835         iterators.
28836         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
28837         integral modes instead of "ss" and "sd".
28838         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
28839         vectors with 32-bit and 64-bit elements.
28840         (vecdupssescalarmodesuffix): New mode attribute.
28841         (vec_dup<mode>): Use it.
28843 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
28845         PR target/83330
28846         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
28847         frame if argument is passed on stack.
28849 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28851         PR target/82682
28852         * ree.c (combine_reaching_defs): Optimize also
28853         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
28854         reg2=any_extend(exp); reg1=reg2;, formatting fix.
28856 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28858         PR middle-end/83189
28859         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
28861 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28863         PR middle-end/83718
28864         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
28865         after they are computed.
28867 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
28869         PR tree-optimization/83695
28870         * gimple-loop-linterchange.cc
28871         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
28872         reset cached scev information after interchange.
28873         (pass_linterchange::execute): Remove call to scev_reset_htab.
28875 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28877         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
28878         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
28879         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
28880         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
28881         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
28882         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
28883         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
28884         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
28885         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
28886         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
28887         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
28888         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
28889         (V_lane_reg): Likewise.
28890         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
28891         New define_expand.
28892         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
28893         (vfmal_lane_low<mode>_intrinsic,
28894         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
28895         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
28896         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
28897         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
28898         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
28899         vfmsl_lane_high<mode>_intrinsic): New define_insns.
28901 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28903         * config/arm/arm-cpus.in (fp16fml): New feature.
28904         (ALL_SIMD): Add fp16fml.
28905         (armv8.2-a): Add fp16fml as an option.
28906         (armv8.3-a): Likewise.
28907         (armv8.4-a): Add fp16fml as part of fp16.
28908         * config/arm/arm.h (TARGET_FP16FML): Define.
28909         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
28910         when appropriate.
28911         * config/arm/arm-modes.def (V2HF): Define.
28912         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
28913         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
28914         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
28915         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
28916         vfmsl_low, vfmsl_high): New set of builtins.
28917         * config/arm/iterators.md (PLUSMINUS): New code iterator.
28918         (vfml_op): New code attribute.
28919         (VFMLHALVES): New int iterator.
28920         (VFML, VFMLSEL): New mode attributes.
28921         (V_reg): Define mapping for V2HF.
28922         (V_hi, V_lo): New mode attributes.
28923         (VF_constraint): Likewise.
28924         (vfml_half, vfml_half_selector): New int attributes.
28925         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
28926         define_expand.
28927         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
28928         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
28929         New define_insn.
28930         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
28931         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
28932         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
28933         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
28934         documentation.
28935         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
28936         Document new effective target and option set.
28938 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28940         * config/arm/arm-cpus.in (armv8_4): New feature.
28941         (ARMv8_4a): New fgroup.
28942         (armv8.4-a): New arch.
28943         * config/arm/arm-tables.opt: Regenerate.
28944         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
28945         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
28946         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
28947         Add matching rules for -march=armv8.4-a and extensions.
28948         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
28950 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
28952         PR target/81821
28953         * config/rx/rx.md (BW): New mode attribute.
28954         (sync_lock_test_and_setsi): Add mode suffix to insn output.
28956 2018-01-11  Richard Biener  <rguenther@suse.de>
28958         PR tree-optimization/83435
28959         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
28960         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
28961         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
28963 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28964             Alan Hayward  <alan.hayward@arm.com>
28965             David Sherwood  <david.sherwood@arm.com>
28967         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
28968         field.
28969         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
28970         (aarch64_print_address_internal): Use it to check for a zero offset.
28972 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28973             Alan Hayward  <alan.hayward@arm.com>
28974             David Sherwood  <david.sherwood@arm.com>
28976         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
28977         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
28978         Return a poly_int64 rather than a HOST_WIDE_INT.
28979         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
28980         rather than a HOST_WIDE_INT.
28981         * config/aarch64/aarch64.h (aarch64_frame): Protect with
28982         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
28983         hard_fp_offset, frame_size, initial_adjust, callee_offset and
28984         final_offset from HOST_WIDE_INT to poly_int64.
28985         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
28986         to_constant when getting the number of units in an Advanced SIMD
28987         mode.
28988         (aarch64_builtin_vectorized_function): Check for a constant number
28989         of units.
28990         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
28991         GET_MODE_SIZE.
28992         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
28993         attribute instead of GET_MODE_NUNITS.
28994         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
28995         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
28996         GET_MODE_SIZE for fixed-size registers.
28997         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
28998         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
28999         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29000         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29001         (aarch64_print_operand, aarch64_print_address_internal)
29002         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29003         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29004         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29005         Handle polynomial GET_MODE_SIZE.
29006         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29007         wider than SImode without modification.
29008         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29009         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29010         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29011         passing and returning SVE modes.
29012         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29013         rather than GEN_INT.
29014         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29015         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29016         (aarch64_allocate_and_probe_stack_space): Likewise.
29017         (aarch64_layout_frame): Cope with polynomial offsets.
29018         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29019         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29020         polynomial offsets.
29021         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29022         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29023         poly_int64 rather than a HOST_WIDE_INT.
29024         (aarch64_get_separate_components, aarch64_process_components)
29025         (aarch64_expand_prologue, aarch64_expand_epilogue)
29026         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29027         (aarch64_anchor_offset): New function, split out from...
29028         (aarch64_legitimize_address): ...here.
29029         (aarch64_builtin_vectorization_cost): Handle polynomial
29030         TYPE_VECTOR_SUBPARTS.
29031         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29032         GET_MODE_NUNITS.
29033         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29034         number of elements from the PARALLEL rather than the mode.
29035         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29036         rather than GET_MODE_BITSIZE.
29037         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29038         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29039         (aarch64_expand_vec_perm_const_1): Handle polynomial
29040         d->perm.length () and d->perm elements.
29041         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29042         Apply to_constant to d->perm elements.
29043         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29044         polynomial CONST_VECTOR_NUNITS.
29045         (aarch64_move_pointer): Take amount as a poly_int64 rather
29046         than an int.
29047         (aarch64_progress_pointer): Avoid temporary variable.
29048         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29049         the mode attribute instead of GET_MODE.
29051 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29052             Alan Hayward  <alan.hayward@arm.com>
29053             David Sherwood  <david.sherwood@arm.com>
29055         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29056         x exists before using it.
29057         (aarch64_add_constant_internal): Rename to...
29058         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29059         src and dest rtxes.  Handle the case in which they're different,
29060         including when the offset is zero.  Replace scratchreg with an rtx.
29061         Use 2 additions if there is no spare register into which we can
29062         move a 16-bit constant.
29063         (aarch64_add_constant): Delete.
29064         (aarch64_add_offset): Replace reg with separate src and dest
29065         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29066         Use aarch64_add_offset_1.
29067         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29068         an rtx rather than an int.  Take the delta as a poly_int64
29069         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29070         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29071         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29072         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29073         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29074         and aarch64_add_sp.
29075         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29076         aarch64_add_constant.
29078 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29080         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29081         Use scalar_float_mode.
29083 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29085         * config/aarch64/aarch64-simd.md
29086         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29087         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29088         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29089         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29090         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29091         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29092         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29093         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29094         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29095         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29097 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29099         PR target/83514
29100         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29101         targ_options->x_arm_arch_string is non NULL.
29103 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29105         * config/aarch64/aarch64.h
29106         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29108 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29110         PR target/82096
29111         * expmed.c (emit_store_flag_force): Swap if const op0
29112         and change VOIDmode to mode of op0.
29114 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29116         PR rtl-optimization/83761
29117         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29118         than bytes to mode_for_size.
29120 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29122         PR middle-end/83189
29123         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29124         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29125         profile.
29127 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29129         PR middle-end/83575
29130         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29131         when in layout mode.
29132         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29133         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29134         partition fixup.
29136 2018-01-10  Michael Collison  <michael.collison@arm.com>
29138         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29139         * config/aarch64/aarch64-option-extension.def: Add
29140         AARCH64_OPT_EXTENSION of 'fp16fml'.
29141         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29142         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29143         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29144         * config/aarch64/constraints.md (Ui7): New constraint.
29145         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29146         (VFMLA_SEL_W): Ditto.
29147         (f16quad): Ditto.
29148         (f16mac1): Ditto.
29149         (VFMLA16_LOW): New int iterator.
29150         (VFMLA16_HIGH): Ditto.
29151         (UNSPEC_FMLAL): New unspec.
29152         (UNSPEC_FMLSL): Ditto.
29153         (UNSPEC_FMLAL2): Ditto.
29154         (UNSPEC_FMLSL2): Ditto.
29155         (f16mac): New code attribute.
29156         * config/aarch64/aarch64-simd-builtins.def
29157         (aarch64_fmlal_lowv2sf): Ditto.
29158         (aarch64_fmlsl_lowv2sf): Ditto.
29159         (aarch64_fmlalq_lowv4sf): Ditto.
29160         (aarch64_fmlslq_lowv4sf): Ditto.
29161         (aarch64_fmlal_highv2sf): Ditto.
29162         (aarch64_fmlsl_highv2sf): Ditto.
29163         (aarch64_fmlalq_highv4sf): Ditto.
29164         (aarch64_fmlslq_highv4sf): Ditto.
29165         (aarch64_fmlal_lane_lowv2sf): Ditto.
29166         (aarch64_fmlsl_lane_lowv2sf): Ditto.
29167         (aarch64_fmlal_laneq_lowv2sf): Ditto.
29168         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29169         (aarch64_fmlalq_lane_lowv4sf): Ditto.
29170         (aarch64_fmlsl_lane_lowv4sf): Ditto.
29171         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29172         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29173         (aarch64_fmlal_lane_highv2sf): Ditto.
29174         (aarch64_fmlsl_lane_highv2sf): Ditto.
29175         (aarch64_fmlal_laneq_highv2sf): Ditto.
29176         (aarch64_fmlsl_laneq_highv2sf): Ditto.
29177         (aarch64_fmlalq_lane_highv4sf): Ditto.
29178         (aarch64_fmlsl_lane_highv4sf): Ditto.
29179         (aarch64_fmlalq_laneq_highv4sf): Ditto.
29180         (aarch64_fmlsl_laneq_highv4sf): Ditto.
29181         * config/aarch64/aarch64-simd.md:
29182         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29183         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29184         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29185         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29186         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29187         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29188         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29189         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29190         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29191         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29192         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29193         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29194         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29195         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29196         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29197         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29198         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29199         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29200         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29201         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29202         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29203         (vfmlsl_low_u32): Ditto.
29204         (vfmlalq_low_u32): Ditto.
29205         (vfmlslq_low_u32): Ditto.
29206         (vfmlal_high_u32): Ditto.
29207         (vfmlsl_high_u32): Ditto.
29208         (vfmlalq_high_u32): Ditto.
29209         (vfmlslq_high_u32): Ditto.
29210         (vfmlal_lane_low_u32): Ditto.
29211         (vfmlsl_lane_low_u32): Ditto.
29212         (vfmlal_laneq_low_u32): Ditto.
29213         (vfmlsl_laneq_low_u32): Ditto.
29214         (vfmlalq_lane_low_u32): Ditto.
29215         (vfmlslq_lane_low_u32): Ditto.
29216         (vfmlalq_laneq_low_u32): Ditto.
29217         (vfmlslq_laneq_low_u32): Ditto.
29218         (vfmlal_lane_high_u32): Ditto.
29219         (vfmlsl_lane_high_u32): Ditto.
29220         (vfmlal_laneq_high_u32): Ditto.
29221         (vfmlsl_laneq_high_u32): Ditto.
29222         (vfmlalq_lane_high_u32): Ditto.
29223         (vfmlslq_lane_high_u32): Ditto.
29224         (vfmlalq_laneq_high_u32): Ditto.
29225         (vfmlslq_laneq_high_u32): Ditto.
29226         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29227         (AARCH64_FL_FOR_ARCH8_4): New.
29228         (AARCH64_ISA_F16FML): New ISA flag.
29229         (TARGET_F16FML): New feature flag for fp16fml.
29230         (doc/invoke.texi): Document new fp16fml option.
29232 2018-01-10  Michael Collison  <michael.collison@arm.com>
29234         * config/aarch64/aarch64-builtins.c:
29235         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29236         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29237         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29238         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29239         (AARCH64_ISA_SHA3): New ISA flag.
29240         (TARGET_SHA3): New feature flag for sha3.
29241         * config/aarch64/iterators.md (sha512_op): New int attribute.
29242         (CRYPTO_SHA512): New int iterator.
29243         (UNSPEC_SHA512H): New unspec.
29244         (UNSPEC_SHA512H2): Ditto.
29245         (UNSPEC_SHA512SU0): Ditto.
29246         (UNSPEC_SHA512SU1): Ditto.
29247         * config/aarch64/aarch64-simd-builtins.def
29248         (aarch64_crypto_sha512hqv2di): New builtin.
29249         (aarch64_crypto_sha512h2qv2di): Ditto.
29250         (aarch64_crypto_sha512su0qv2di): Ditto.
29251         (aarch64_crypto_sha512su1qv2di): Ditto.
29252         (aarch64_eor3qv8hi): Ditto.
29253         (aarch64_rax1qv2di): Ditto.
29254         (aarch64_xarqv2di): Ditto.
29255         (aarch64_bcaxqv8hi): Ditto.
29256         * config/aarch64/aarch64-simd.md:
29257         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29258         (aarch64_crypto_sha512su0qv2di): Ditto.
29259         (aarch64_crypto_sha512su1qv2di): Ditto.
29260         (aarch64_eor3qv8hi): Ditto.
29261         (aarch64_rax1qv2di): Ditto.
29262         (aarch64_xarqv2di): Ditto.
29263         (aarch64_bcaxqv8hi): Ditto.
29264         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29265         (vsha512h2q_u64): Ditto.
29266         (vsha512su0q_u64): Ditto.
29267         (vsha512su1q_u64): Ditto.
29268         (veor3q_u16): Ditto.
29269         (vrax1q_u64): Ditto.
29270         (vxarq_u64): Ditto.
29271         (vbcaxq_u16): Ditto.
29272         * config/arm/types.md (crypto_sha512): New type attribute.
29273         (crypto_sha3): Ditto.
29274         (doc/invoke.texi): Document new sha3 option.
29276 2018-01-10  Michael Collison  <michael.collison@arm.com>
29278         * config/aarch64/aarch64-builtins.c:
29279         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29280         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29281         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29282         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29283         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29284         (AARCH64_ISA_SM4): New ISA flag.
29285         (TARGET_SM4): New feature flag for sm4.
29286         * config/aarch64/aarch64-simd-builtins.def
29287         (aarch64_sm3ss1qv4si): Ditto.
29288         (aarch64_sm3tt1aq4si): Ditto.
29289         (aarch64_sm3tt1bq4si): Ditto.
29290         (aarch64_sm3tt2aq4si): Ditto.
29291         (aarch64_sm3tt2bq4si): Ditto.
29292         (aarch64_sm3partw1qv4si): Ditto.
29293         (aarch64_sm3partw2qv4si): Ditto.
29294         (aarch64_sm4eqv4si): Ditto.
29295         (aarch64_sm4ekeyqv4si): Ditto.
29296         * config/aarch64/aarch64-simd.md:
29297         (aarch64_sm3ss1qv4si): Ditto.
29298         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29299         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29300         (aarch64_sm4eqv4si): Ditto.
29301         (aarch64_sm4ekeyqv4si): Ditto.
29302         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29303         (sm3part_op): Ditto.
29304         (CRYPTO_SM3TT): Ditto.
29305         (CRYPTO_SM3PART): Ditto.
29306         (UNSPEC_SM3SS1): New unspec.
29307         (UNSPEC_SM3TT1A): Ditto.
29308         (UNSPEC_SM3TT1B): Ditto.
29309         (UNSPEC_SM3TT2A): Ditto.
29310         (UNSPEC_SM3TT2B): Ditto.
29311         (UNSPEC_SM3PARTW1): Ditto.
29312         (UNSPEC_SM3PARTW2): Ditto.
29313         (UNSPEC_SM4E): Ditto.
29314         (UNSPEC_SM4EKEY): Ditto.
29315         * config/aarch64/constraints.md (Ui2): New constraint.
29316         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29317         * config/arm/types.md (crypto_sm3): New type attribute.
29318         (crypto_sm4): Ditto.
29319         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29320         (vsm3tt1aq_u32): Ditto.
29321         (vsm3tt1bq_u32): Ditto.
29322         (vsm3tt2aq_u32): Ditto.
29323         (vsm3tt2bq_u32): Ditto.
29324         (vsm3partw1q_u32): Ditto.
29325         (vsm3partw2q_u32): Ditto.
29326         (vsm4eq_u32): Ditto.
29327         (vsm4ekeyq_u32): Ditto.
29328         (doc/invoke.texi): Document new sm4 option.
29330 2018-01-10  Michael Collison  <michael.collison@arm.com>
29332         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29333         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29334         (AARCH64_FL_FOR_ARCH8_4): New.
29335         (AARCH64_FL_V8_4): New flag.
29336         (doc/invoke.texi): Document new armv8.4-a option.
29338 2018-01-10  Michael Collison  <michael.collison@arm.com>
29340         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29341         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29342         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29343         * config/aarch64/aarch64-option-extension.def: Add
29344         AARCH64_OPT_EXTENSION of 'sha2'.
29345         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29346         (crypto): Disable sha2 and aes if crypto disabled.
29347         (crypto): Enable aes and sha2 if enabled.
29348         (simd): Disable sha2 and aes if simd disabled.
29349         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29350         New flags.
29351         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29352         (TARGET_SHA2): New feature flag for sha2.
29353         (TARGET_AES): New feature flag for aes.
29354         * config/aarch64/aarch64-simd.md:
29355         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29356         conditional on TARGET_AES.
29357         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29358         (aarch64_crypto_sha1hsi): Make pattern conditional
29359         on TARGET_SHA2.
29360         (aarch64_crypto_sha1hv4si): Ditto.
29361         (aarch64_be_crypto_sha1hv4si): Ditto.
29362         (aarch64_crypto_sha1su1v4si): Ditto.
29363         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29364         (aarch64_crypto_sha1su0v4si): Ditto.
29365         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29366         (aarch64_crypto_sha256su0v4si): Ditto.
29367         (aarch64_crypto_sha256su1v4si): Ditto.
29368         (doc/invoke.texi): Document new aes and sha2 options.
29370 2018-01-10  Martin Sebor  <msebor@redhat.com>
29372         PR tree-optimization/83781
29373         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29374         as string arrays.
29376 2018-01-11  Martin Sebor  <msebor@gmail.com>
29377             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29379         PR tree-optimization/83501
29380         PR tree-optimization/81703
29382         * tree-ssa-strlen.c (get_string_cst): Rename...
29383         (get_string_len): ...to this.  Handle global constants.
29384         (handle_char_store): Adjust.
29386 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
29387             Jim Wilson  <jimw@sifive.com>
29389         * config/riscv/riscv-protos.h (riscv_output_return): New.
29390         * config/riscv/riscv.c (struct machine_function): New naked_p field.
29391         (riscv_attribute_table, riscv_output_return),
29392         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29393         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29394         (riscv_compute_frame_info): Only compute frame->mask if not a naked
29395         function.
29396         (riscv_expand_prologue): Add early return for naked function.
29397         (riscv_expand_epilogue): Likewise.
29398         (riscv_function_ok_for_sibcall): Return false for naked function.
29399         (riscv_set_current_function): New.
29400         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29401         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29402         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29403         * doc/extend.texi (RISC-V Function Attributes): New.
29405 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
29407         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29408         check for 128-bit long double before checking TCmode.
29409         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29410         128-bit long doubles before checking TFmode or TCmode.
29411         (FLOAT128_IBM_P): Likewise.
29413 2018-01-10  Martin Sebor  <msebor@redhat.com>
29415         PR tree-optimization/83671
29416         * builtins.c (c_strlen): Unconditionally return zero for the empty
29417         string.
29418         Use -Warray-bounds for warnings.
29419         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29420         for non-constant array indices with COMPONENT_REF, arrays of
29421         arrays, and pointers to arrays.
29422         (gimple_fold_builtin_strlen): Determine and set length range for
29423         non-constant character arrays.
29425 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
29427         PR middle-end/81897
29428         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29429         empty blocks.
29431 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
29433         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29435 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29437         PR target/83399
29438         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29439         VECTOR_MEM_ALTIVEC_OR_VSX_P.
29440         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29441         indexed_or_indirect_operand predicate.
29442         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29443         (*vsx_le_perm_load_v8hi): Likewise.
29444         (*vsx_le_perm_load_v16qi): Likewise.
29445         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29446         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29447         (*vsx_le_perm_store_v8hi): Likewise.
29448         (*vsx_le_perm_store_v16qi): Likewise.
29449         (eight unnamed splitters): Likewise.
29451 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29453         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29454         * config/rs6000/emmintrin.h: Likewise.
29455         * config/rs6000/mmintrin.h: Likewise.
29456         * config/rs6000/xmmintrin.h: Likewise.
29458 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
29460         PR c++/43486
29461         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29462         "public_flag".
29463         * tree.c (tree_nop_conversion): Return true for location wrapper
29464         nodes.
29465         (maybe_wrap_with_location): New function.
29466         (selftest::check_strip_nops): New function.
29467         (selftest::test_location_wrappers): New function.
29468         (selftest::tree_c_tests): Call it.
29469         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29470         (maybe_wrap_with_location): New decl.
29471         (EXPR_LOCATION_WRAPPER_P): New macro.
29472         (location_wrapper_p): New inline function.
29473         (tree_strip_any_location_wrapper): New inline function.
29475 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
29477         PR target/83735
29478         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29479         stack_realign_offset for the largest alignment of stack slot
29480         actually used.
29481         (ix86_find_max_used_stack_alignment): New function.
29482         (ix86_finalize_stack_frame_flags): Use it.  Set
29483         max_used_stack_alignment if we don't realign stack.
29484         * config/i386/i386.h (machine_function): Add
29485         max_used_stack_alignment.
29487 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
29489         * config/arm/arm.opt (-mbranch-cost): New option.
29490         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29491         account.
29493 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
29495         PR target/83629
29496         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29497         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29499 2018-01-10  Richard Biener  <rguenther@suse.de>
29501         PR debug/83765
29502         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29503         early out so it also covers the case where we have a non-NULL
29504         origin.
29506 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29508         PR tree-optimization/83753
29509         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29510         for non-strided grouped accesses if the number of elements is 1.
29512 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29514         PR target/81616
29515         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29516         * i386.h (TARGET_USE_GATHER): Define.
29517         * x86-tune.def (X86_TUNE_USE_GATHER): New.
29519 2018-01-10  Martin Liska  <mliska@suse.cz>
29521         PR bootstrap/82831
29522         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29523         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29524         partitioning.
29525         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
29526         CLEANUP_NO_PARTITIONING is not set.
29528 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29530         * doc/rtl.texi: Remove documentation of (const ...) wrappers
29531         for vectors, as a partial revert of r254296.
29532         * rtl.h (const_vec_p): Delete.
29533         (const_vec_duplicate_p): Don't test for vector CONSTs.
29534         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
29535         * expmed.c (make_tree): Likewise.
29537         Revert:
29538         * common.md (E, F): Use CONSTANT_P instead of checking for
29539         CONST_VECTOR.
29540         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
29541         checking for CONST_VECTOR.
29543 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29545         PR middle-end/83575
29546         * predict.c (force_edge_cold): Handle in more sane way edges
29547         with no prediction.
29549 2018-01-09  Carl Love  <cel@us.ibm.com>
29551         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
29552         V4SI, V4SF types.
29553         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
29554         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
29555         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
29556         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
29557         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
29558         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
29559         * config/rs6000/rs6000-protos.h: Add extern defition for
29560         rs6000_generate_float2_double_code.
29561         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
29562         function.
29563         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
29564         (float2_v2df): Add define_expand.
29566 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
29568         PR target/83628
29569         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
29570         op_mode in the force_to_mode call.
29572 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29574         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
29575         instead of checking each element individually.
29576         (aarch64_evpc_uzp): Likewise.
29577         (aarch64_evpc_zip): Likewise.
29578         (aarch64_evpc_ext): Likewise.
29579         (aarch64_evpc_rev): Likewise.
29580         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
29581         instead of checking each element individually.  Return true without
29582         generating rtl if
29583         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
29584         whether all selected elements come from the same input, instead of
29585         checking each element individually.  Remove calls to gen_rtx_REG,
29586         start_sequence and end_sequence and instead assert that no rtl is
29587         generated.
29589 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29591         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
29592         order of HIGH and CONST checks.
29594 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29596         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
29597         if the destination isn't an SSA_NAME.
29599 2018-01-09  Richard Biener  <rguenther@suse.de>
29601         PR tree-optimization/83668
29602         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
29603         move prologue...
29604         (canonicalize_loop_form): ... here, renamed from ...
29605         (canonicalize_loop_closed_ssa_form): ... this and amended to
29606         swap successor edges for loop exit blocks to make us use
29607         the RPO order we need for initial schedule generation.
29609 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
29611         PR tree-optimization/64811
29612         * match.pd: When optimizing comparisons with Inf, avoid
29613         introducing or losing exceptions from comparisons with NaN.
29615 2018-01-09  Martin Liska  <mliska@suse.cz>
29617         PR sanitizer/82517
29618         * asan.c (shadow_mem_size): Add gcc_assert.
29620 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
29622         Don't save registers in main().
29624         PR target/83738
29625         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
29626         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
29627         * config/avr/avr.c (avr_set_current_function): Don't error if
29628         naked, OS_task or OS_main are specified at the same time.
29629         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
29630         OS_main.
29631         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
29632         attribute.
29633         * common/config/avr/avr-common.c (avr_option_optimization_table):
29634         Switch on -mmain-is-OS_task for optimizing compilations.
29636 2018-01-09  Richard Biener  <rguenther@suse.de>
29638         PR tree-optimization/83572
29639         * graphite.c: Include cfganal.h.
29640         (graphite_transform_loops): Connect infinite loops to exit
29641         and remove fake edges at the end.
29643 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29645         * ipa-inline.c (edge_badness): Revert accidental checkin.
29647 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29649         PR ipa/80763
29650         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
29651         symbols; not inline clones.
29653 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
29655         PR target/83507
29656         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
29657         hard registers.  Formatting fixes.
29659         PR preprocessor/83722
29660         * gcc.c (try_generate_repro): Pass
29661         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
29662         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
29663         do_report_bug.
29665 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
29666             Kito Cheng  <kito.cheng@gmail.com>
29668         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
29669         (riscv_leaf_function_p): Delete.
29670         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
29672 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29674         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
29675         function.
29676         (do_ifelse): New function.
29677         (do_isel): New function.
29678         (do_sub3): New function.
29679         (do_add3): New function.
29680         (do_load_mask_compare): New function.
29681         (do_overlap_load_compare): New function.
29682         (expand_compare_loop): New function.
29683         (expand_block_compare): Call expand_compare_loop() when appropriate.
29684         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
29685         option description.
29686         (-mblock-compare-inline-loop-limit): New option.
29688 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29690         PR target/83677
29691         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
29692         Reverse order of second and third operands in first alternative.
29693         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
29694         of first and second elements in UNSPEC_VPERMR vector.
29695         (altivec_expand_vec_perm_le): Likewise.
29697 2018-01-08  Jeff Law  <law@redhat.com>
29699         PR rtl-optimizatin/81308
29700         * tree-switch-conversion.c (cfg_altered): New file scoped static.
29701         (process_switch): If group_case_labels makes a change, then set
29702         cfg_altered.
29703         (pass_convert_switch::execute): If a switch is converted, then
29704         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
29706         PR rtl-optimization/81308
29707         * recog.c (split_all_insns): Conditionally cleanup the CFG after
29708         splitting insns.
29710 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
29712         PR target/83663 - Revert r255946
29713         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
29714         generation for cases where splatting a value is not useful.
29715         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
29716         across a vec_duplicate and a paradoxical subreg forming a vector
29717         mode to a vec_concat.
29719 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29721         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
29722         -march=armv8.3-a variants.
29723         * config/arm/t-multilib: Likewise.
29724         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
29726 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29728         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
29729         to generate rtl.
29730         (cceq_ior_compare_complement): Give it a name so I can use it, and
29731         change boolean_or_operator predicate to boolean_operator so it can
29732         be used to generate a crand.
29733         (eqne): New code iterator.
29734         (bd/bd_neg): New code_attrs.
29735         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
29736         a single define_insn.
29737         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
29738         decrement (bdnzt/bdnzf/bdzt/bdzf).
29739         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
29740         with the new names of the branch decrement patterns, and added the
29741         names of the branch decrement conditional patterns.
29743 2018-01-08  Richard Biener  <rguenther@suse.de>
29745         PR tree-optimization/83563
29746         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
29747         cache.
29749 2018-01-08  Richard Biener  <rguenther@suse.de>
29751         PR middle-end/83713
29752         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
29754 2018-01-08  Richard Biener  <rguenther@suse.de>
29756         PR tree-optimization/83685
29757         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
29758         references to abnormals.
29760 2018-01-08  Richard Biener  <rguenther@suse.de>
29762         PR lto/83719
29763         * dwarf2out.c (output_indirect_strings): Handle empty
29764         skeleton_debug_str_hash.
29765         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
29767 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29769         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
29770         (emit_store_direct): Likewise.
29771         (arc_trampoline_adjust_address): Likewise.
29772         (arc_asm_trampoline_template): New function.
29773         (arc_initialize_trampoline): Use asm_trampoline_template.
29774         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
29775         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
29776         * config/arc/arc.md (flush_icache): Delete pattern.
29778 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29780         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
29781         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
29782         munaligned-access.
29784 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29786         PR target/83681
29787         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
29788         by not USED_FOR_TARGET.
29789         (make_pass_resolve_sw_modes): Likewise.
29791 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29793         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
29794         USED_FOR_TARGET.
29796 2018-01-08  Richard Biener  <rguenther@suse.de>
29798         PR middle-end/83580
29799         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
29801 2018-01-08  Richard Biener  <rguenther@suse.de>
29803         PR middle-end/83517
29804         * match.pd ((t * 2) / 2) -> t): Add missing :c.
29806 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
29808         PR middle-end/81897
29809         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
29810         basic blocks with a small number of successors.
29811         (convert_control_dep_chain_into_preds): Improve handling of
29812         forwarder blocks.
29813         (dump_predicates): Split apart into...
29814         (dump_pred_chain): ...here...
29815         (dump_pred_info): ...and here.
29816         (can_one_predicate_be_invalidated_p): Add debugging printfs.
29817         (can_chain_union_be_invalidated_p): Improve check for invalidation
29818         of paths.
29819         (uninit_uses_cannot_happen): Avoid unnecessary if
29820         convert_control_dep_chain_into_preds yielded nothing.
29822 2018-01-06  Martin Sebor  <msebor@redhat.com>
29824         PR tree-optimization/83640
29825         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
29826         subtracting negative offset from size.
29827         (builtin_access::overlap): Adjust offset bounds of the access to fall
29828         within the size of the object if possible.
29830 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
29832         PR rtl-optimization/83699
29833         * expmed.c (extract_bit_field_1): Restrict the vector usage of
29834         extract_bit_field_as_subreg to cases in which the extracted
29835         value is also a vector.
29837         * lra-constraints.c (process_alt_operands): Test for the equivalence
29838         substitutions when detecting a possible reload cycle.
29840 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
29842         PR debug/83480
29843         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
29844         by default if flag_selective_schedling{,2}.  Formatting fixes.
29846         PR rtl-optimization/83682
29847         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
29848         if it has non-VECTOR_MODE element mode.
29849         (vec_duplicate_p): Likewise.
29851         PR middle-end/83694
29852         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
29853         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
29855 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29857         PR target/83604
29858         * config/i386/i386-builtin.def
29859         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
29860         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
29861         Require also OPTION_MASK_ISA_AVX512F in addition to
29862         OPTION_MASK_ISA_GFNI.
29863         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
29864         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
29865         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
29866         to OPTION_MASK_ISA_GFNI.
29867         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
29868         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
29869         OPTION_MASK_ISA_AVX512BW.
29870         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
29871         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
29872         addition to OPTION_MASK_ISA_GFNI.
29873         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
29874         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
29875         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
29876         to OPTION_MASK_ISA_GFNI.
29877         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
29878         a requirement for all ISAs rather than any of them with a few
29879         exceptions.
29880         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
29881         processing.
29882         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
29883         bitmasks to be enabled with 3 exceptions, instead of requiring any
29884         enabled ISA with lots of exceptions.
29885         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
29886         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
29887         Change avx512bw in isa attribute to avx512f.
29888         * config/i386/sgxintrin.h: Add license boilerplate.
29889         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
29890         to __AVX512F__ and __AVX512VL to __AVX512VL__.
29891         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
29892         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
29893         defined.
29894         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
29895         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
29896         temporarily sse2 rather than sse if not enabled already.
29898         PR target/83604
29899         * config/i386/sse.md (VI248_VLBW): Rename to ...
29900         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
29901         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
29902         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
29903         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
29904         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
29905         mode iterator instead of VI248_VLBW.
29907 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
29909         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
29910         (record_modified): Skip clobbers; add debug output.
29911         (param_change_prob): Use sreal frequencies.
29913 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29915         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
29916         punt for user-aligned variables.
29918 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29920         * tree-chrec.c (chrec_contains_symbols): Return true for
29921         POLY_INT_CST.
29923 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
29925         PR target/82439
29926         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
29927         of (x|y) == x for BICS pattern.
29929 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29931         PR tree-optimization/83605
29932         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
29933         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
29934         can throw.
29936 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29938         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
29939         * config/epiphany/rtems.h: New file.
29941 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29942             Uros Bizjak  <ubizjak@gmail.com>
29944         PR target/83554
29945         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
29946         QIreg_operand instead of register_operand predicate.
29947         * config/i386/i386.c (ix86_rop_should_change_byte_p,
29948         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
29949         comments instead of -fmitigate[-_]rop.
29951 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29953         PR bootstrap/81926
29954         * cgraphunit.c (symbol_table::compile): Switch to text_section
29955         before calling assembly_start debug hook.
29956         * run-rtl-passes.c (run_rtl_passes): Likewise.
29957         Include output.h.
29959 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29961         * tree-vrp.c (extract_range_from_binary_expr_1): Check
29962         range_int_cst_p rather than !symbolic_range_p before calling
29963         extract_range_from_multiplicative_op_1.
29965 2018-01-04  Jeff Law  <law@redhat.com>
29967         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
29968         redundant test in assertion.
29970 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29972         * doc/rtl.texi: Document machine_mode wrapper classes.
29974 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29976         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
29977         using tree_to_uhwi.
29979 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29981         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
29982         the VEC_PERM_EXPR fold to fail.
29984 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29986         PR debug/83585
29987         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
29988         to switched_sections.
29990 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29992         PR target/83680
29993         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
29994         test for d.testing.
29996 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
29998         PR target/83387
29999         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30000         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30002 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30004         PR debug/83666
30005         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30006         is BLKmode and bitpos not zero or mode change is needed.
30008 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30010         PR target/83675
30011         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30012         TARGET_VIS2.
30014 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30016         PR target/83628
30017         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30018         instead of MULT rtx.  Update all corresponding splitters.
30019         (*saddl_se): Ditto.
30020         (*ssub<modesuffix>): Ditto.
30021         (*ssubl_se): Ditto.
30022         (*cmp_sadd_di): Update split patterns.
30023         (*cmp_sadd_si): Ditto.
30024         (*cmp_sadd_sidi): Ditto.
30025         (*cmp_ssub_di): Ditto.
30026         (*cmp_ssub_si): Ditto.
30027         (*cmp_ssub_sidi): Ditto.
30028         * config/alpha/predicates.md (const23_operand): New predicate.
30029         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30030         Look for ASHIFT, not MULT inner operand.
30031         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30033 2018-01-04  Martin Liska  <mliska@suse.cz>
30035         PR gcov-profile/83669
30036         * gcov.c (output_intermediate_file): Add version to intermediate
30037         gcov file.
30038         * doc/gcov.texi: Document new field 'version' in intermediate
30039         file format. Fix location of '-k' option of gcov command.
30041 2018-01-04  Martin Liska  <mliska@suse.cz>
30043         PR ipa/82352
30044         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30046 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30048         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30050 2018-01-03  Martin Sebor  <msebor@redhat.com>
30052         PR tree-optimization/83655
30053         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30054         checking calls with invalid arguments.
30056 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30058         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30059         (vectorizable_mask_load_store): Delete.
30060         (vectorizable_call): Return false for masked loads and stores.
30061         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30062         instead of gimple_assign_rhs1.
30063         (vectorizable_load): Handle IFN_MASK_LOAD.
30064         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30066 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30068         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30069         split out from..,
30070         (vectorizable_mask_load_store): ...here.
30071         (vectorizable_load): ...and here.
30073 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30075         * tree-vect-stmts.c (vect_build_all_ones_mask)
30076         (vect_build_zero_merge_argument): New functions, split out from...
30077         (vectorizable_load): ...here.
30079 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30081         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30082         split out from...
30083         (vectorizable_mask_load_store): ...here.
30084         (vectorizable_store): ...and here.
30086 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30088         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30089         split out from...
30090         (vectorizable_mask_load_store): ...here.
30092 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30094         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30095         (vect_model_store_cost): Take a vec_load_store_type instead of a
30096         vect_def_type.
30097         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30098         (vect_model_store_cost): Take a vec_load_store_type instead of a
30099         vect_def_type.
30100         (vectorizable_mask_load_store): Update accordingly.
30101         (vectorizable_store): Likewise.
30102         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30106         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30107         IFN_MASK_LOAD calls here rather than...
30108         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30111             Alan Hayward  <alan.hayward@arm.com>
30112             David Sherwood  <david.sherwood@arm.com>
30114         * expmed.c (extract_bit_field_1): For vector extracts,
30115         fall back to extract_bit_field_as_subreg if vec_extract
30116         isn't available.
30118 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30119             Alan Hayward  <alan.hayward@arm.com>
30120             David Sherwood  <david.sherwood@arm.com>
30122         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30123         they are variable or constant sized.
30124         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30125         slots for constant-sized data.
30127 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30128             Alan Hayward  <alan.hayward@arm.com>
30129             David Sherwood  <david.sherwood@arm.com>
30131         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30132         handling COND_EXPRs with boolean comparisons, try to find a better
30133         basis for the mask type than the boolean itself.
30135 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30137         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30138         is calculated and how it can be overridden.
30139         * genmodes.c (max_bitsize_mode_any_mode): New variable.
30140         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30141         if defined.
30142         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30143         if nonzero.
30145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30146             Alan Hayward  <alan.hayward@arm.com>
30147             David Sherwood  <david.sherwood@arm.com>
30149         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30150         Remove the mode argument.
30151         (aarch64_simd_valid_immediate): Remove the mode and inverse
30152         arguments.
30153         * config/aarch64/iterators.md (bitsize): New iterator.
30154         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30155         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30156         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30157         aarch64_simd_valid_immediate.
30158         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30159         (aarch64_reg_or_bic_imm): Likewise.
30160         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30161         with an insn_type enum and msl with a modifier_type enum.
30162         Replace element_width with a scalar_mode.  Change the shift
30163         to unsigned int.  Add constructors for scalar_float_mode and
30164         scalar_int_mode elements.
30165         (aarch64_vect_float_const_representable_p): Delete.
30166         (aarch64_can_const_movi_rtx_p)
30167         (aarch64_simd_scalar_immediate_valid_for_move)
30168         (aarch64_simd_make_constant): Update call to
30169         aarch64_simd_valid_immediate.
30170         (aarch64_advsimd_valid_immediate_hs): New function.
30171         (aarch64_advsimd_valid_immediate): Likewise.
30172         (aarch64_simd_valid_immediate): Remove mode and inverse
30173         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
30174         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30175         and aarch64_float_const_representable_p on the result.
30176         (aarch64_output_simd_mov_immediate): Remove mode argument.
30177         Update call to aarch64_simd_valid_immediate and use of
30178         simd_immediate_info.
30179         (aarch64_output_scalar_simd_mov_immediate): Update call
30180         accordingly.
30182 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30183             Alan Hayward  <alan.hayward@arm.com>
30184             David Sherwood  <david.sherwood@arm.com>
30186         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30187         (mode_nunits): Likewise CONST_MODE_NUNITS.
30188         * machmode.def (ADJUST_NUNITS): Document.
30189         * genmodes.c (mode_data::need_nunits_adj): New field.
30190         (blank_mode): Update accordingly.
30191         (adj_nunits): New variable.
30192         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30193         parameter.
30194         (emit_mode_size_inline): Set need_bytesize_adj for all modes
30195         listed in adj_nunits.
30196         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30197         listed in adj_nunits.  Don't emit case statements for such modes.
30198         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30199         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
30200         nothing if adj_nunits is nonnull.
30201         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30202         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30203         (emit_mode_fbit): Update use of print_maybe_const_decl.
30204         (emit_move_size): Likewise.  Treat the array as non-const
30205         if adj_nunits.
30206         (emit_mode_adjustments): Handle adj_nunits.
30208 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30210         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30211         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30212         (VECTOR_MODES): Use it.
30213         (make_vector_modes): Take the prefix as an argument.
30215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30216             Alan Hayward  <alan.hayward@arm.com>
30217             David Sherwood  <david.sherwood@arm.com>
30219         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30220         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30221         for MODE_VECTOR_BOOL.
30222         * machmode.def (VECTOR_BOOL_MODE): Document.
30223         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30224         (make_vector_bool_mode): New function.
30225         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30226         MODE_VECTOR_BOOL.
30227         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30228         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30229         Likewise.
30230         * stor-layout.c (int_mode_for_mode): Likewise.
30231         * tree.c (build_vector_type_for_mode): Likewise.
30232         * varasm.c (output_constant_pool_2): Likewise.
30233         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30234         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30235         for MODE_VECTOR_BOOL.
30236         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30237         of mode class checks.
30238         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30239         instead of a list of mode class checks.
30240         (expand_vector_scalar_condition): Likewise.
30241         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30243 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30244             Alan Hayward  <alan.hayward@arm.com>
30245             David Sherwood  <david.sherwood@arm.com>
30247         * machmode.h (mode_size): Change from unsigned short to
30248         poly_uint16_pod.
30249         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30250         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30251         or if measurement_type is not polynomial.
30252         (fixed_size_mode::includes_p): Check for constant-sized modes.
30253         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30254         return a poly_uint16 rather than an unsigned short.
30255         (emit_mode_size): Change the type of mode_size from unsigned short
30256         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30257         (emit_mode_adjustments): Cope with polynomial vector sizes.
30258         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30259         for GET_MODE_SIZE.
30260         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30261         for GET_MODE_SIZE.
30262         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30263         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30264         * caller-save.c (setup_save_areas): Likewise.
30265         (replace_reg_with_saved_mem): Likewise.
30266         * calls.c (emit_library_call_value_1): Likewise.
30267         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30268         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30269         (gen_lowpart_for_combine): Likewise.
30270         * convert.c (convert_to_integer_1): Likewise.
30271         * cse.c (equiv_constant, cse_insn): Likewise.
30272         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30273         (cselib_subst_to_values): Likewise.
30274         * dce.c (word_dce_process_block): Likewise.
30275         * df-problems.c (df_word_lr_mark_ref): Likewise.
30276         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30277         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30278         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30279         (rtl_for_decl_location): Likewise.
30280         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30281         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30282         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30283         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30284         (expand_expr_real_1): Likewise.
30285         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30286         (pad_below): Likewise.
30287         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30288         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30289         * ira.c (get_subreg_tracking_sizes): Likewise.
30290         * ira-build.c (ira_create_allocno_objects): Likewise.
30291         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30292         (ira_sort_regnos_for_alter_reg): Likewise.
30293         * ira-costs.c (record_operand_costs): Likewise.
30294         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30295         (resolve_simple_move): Likewise.
30296         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30297         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30298         (lra_constraints): Likewise.
30299         (CONST_POOL_OK_P): Reject variable-sized modes.
30300         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30301         (add_pseudo_to_slot, lra_spill): Likewise.
30302         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30303         * optabs-query.c (get_best_extraction_insn): Likewise.
30304         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30305         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30306         (expand_mult_highpart, valid_multiword_target_p): Likewise.
30307         * recog.c (offsettable_address_addr_space_p): Likewise.
30308         * regcprop.c (maybe_mode_change): Likewise.
30309         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30310         * regrename.c (build_def_use): Likewise.
30311         * regstat.c (dump_reg_info): Likewise.
30312         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30313         (find_reloads, find_reloads_subreg_address): Likewise.
30314         * reload1.c (eliminate_regs_1): Likewise.
30315         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30316         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30317         (simplify_binary_operation_1, simplify_subreg): Likewise.
30318         * targhooks.c (default_function_arg_padding): Likewise.
30319         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30320         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30321         (verify_gimple_assign_ternary): Likewise.
30322         * tree-inline.c (estimate_move_cost): Likewise.
30323         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30324         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30325         (get_address_cost_ainc): Likewise.
30326         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30327         (vect_supportable_dr_alignment): Likewise.
30328         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30329         (vectorizable_reduction): Likewise.
30330         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30331         (vectorizable_operation, vectorizable_load): Likewise.
30332         * tree.c (build_same_sized_truth_vector_type): Likewise.
30333         * valtrack.c (cleanup_auto_inc_dec): Likewise.
30334         * var-tracking.c (emit_note_insn_var_location): Likewise.
30335         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30336         (ADDR_VEC_ALIGN): Likewise.
30338 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30339             Alan Hayward  <alan.hayward@arm.com>
30340             David Sherwood  <david.sherwood@arm.com>
30342         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30343         unsigned short.
30344         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30345         or if measurement_type is polynomial.
30346         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30347         * combine.c (make_extraction): Likewise.
30348         * dse.c (find_shift_sequence): Likewise.
30349         * dwarf2out.c (mem_loc_descriptor): Likewise.
30350         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30351         (extract_bit_field, extract_low_bits): Likewise.
30352         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30353         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30354         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30355         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30356         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30357         * reload.c (find_reloads): Likewise.
30358         * reload1.c (alter_reg): Likewise.
30359         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30360         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30361         * tree-if-conv.c (predicate_mem_writes): Likewise.
30362         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30363         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30364         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30365         * valtrack.c (dead_debug_insert_temp): Likewise.
30366         * varasm.c (mergeable_constant_section): Likewise.
30367         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30369 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30370             Alan Hayward  <alan.hayward@arm.com>
30371             David Sherwood  <david.sherwood@arm.com>
30373         * expr.c (expand_assignment): Cope with polynomial mode sizes
30374         when assigning to a CONCAT.
30376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30377             Alan Hayward  <alan.hayward@arm.com>
30378             David Sherwood  <david.sherwood@arm.com>
30380         * machmode.h (mode_precision): Change from unsigned short to
30381         poly_uint16_pod.
30382         (mode_to_precision): Return a poly_uint16 rather than an unsigned
30383         short.
30384         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30385         or if measurement_type is not polynomial.
30386         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
30387         in which the mode is already known to be a scalar_int_mode.
30388         * genmodes.c (emit_mode_precision): Change the type of mode_precision
30389         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
30390         initializer.
30391         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30392         for GET_MODE_PRECISION.
30393         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30394         for GET_MODE_PRECISION.
30395         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30396         as polynomial.
30397         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30398         (expand_field_assignment, make_extraction): Likewise.
30399         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30400         (get_last_value): Likewise.
30401         * convert.c (convert_to_integer_1): Likewise.
30402         * cse.c (cse_insn): Likewise.
30403         * expr.c (expand_expr_real_1): Likewise.
30404         * lra-constraints.c (simplify_operand_subreg): Likewise.
30405         * optabs-query.c (can_atomic_load_p): Likewise.
30406         * optabs.c (expand_atomic_load): Likewise.
30407         (expand_atomic_store): Likewise.
30408         * ree.c (combine_reaching_defs): Likewise.
30409         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30410         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30411         * tree.h (type_has_mode_precision_p): Likewise.
30412         * ubsan.c (instrument_si_overflow): Likewise.
30414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30415             Alan Hayward  <alan.hayward@arm.com>
30416             David Sherwood  <david.sherwood@arm.com>
30418         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30419         polynomial numbers of units.
30420         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30421         (valid_vector_subparts_p): New function.
30422         (build_vector_type): Remove temporary shim and take the number
30423         of units as a poly_uint64 rather than an int.
30424         (build_opaque_vector_type): Take the number of units as a
30425         poly_uint64 rather than an int.
30426         * tree.c (build_vector_from_ctor): Handle polynomial
30427         TYPE_VECTOR_SUBPARTS.
30428         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30429         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30430         (build_vector_from_val): If the number of units is variable,
30431         use build_vec_duplicate_cst for constant operands and
30432         VEC_DUPLICATE_EXPR otherwise.
30433         (make_vector_type): Remove temporary is_constant ().
30434         (build_vector_type, build_opaque_vector_type): Take the number of
30435         units as a poly_uint64 rather than an int.
30436         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30437         VECTOR_CST_NELTS.
30438         * cfgexpand.c (expand_debug_expr): Likewise.
30439         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30440         (store_constructor, expand_expr_real_1): Likewise.
30441         (const_scalar_mask_from_tree): Likewise.
30442         * fold-const-call.c (fold_const_reduction): Likewise.
30443         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30444         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30445         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30446         (fold_relational_const): Likewise.
30447         (native_interpret_vector): Likewise.  Change the size from an
30448         int to an unsigned int.
30449         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30450         TYPE_VECTOR_SUBPARTS.
30451         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30452         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30453         duplicating a non-constant operand into a variable-length vector.
30454         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30455         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30456         * ipa-icf.c (sem_variable::equals): Likewise.
30457         * match.pd: Likewise.
30458         * omp-simd-clone.c (simd_clone_subparts): Likewise.
30459         * print-tree.c (print_node): Likewise.
30460         * stor-layout.c (layout_type): Likewise.
30461         * targhooks.c (default_builtin_vectorization_cost): Likewise.
30462         * tree-cfg.c (verify_gimple_comparison): Likewise.
30463         (verify_gimple_assign_binary): Likewise.
30464         (verify_gimple_assign_ternary): Likewise.
30465         (verify_gimple_assign_single): Likewise.
30466         * tree-pretty-print.c (dump_generic_node): Likewise.
30467         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30468         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30469         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30470         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30471         (vect_shift_permute_load_chain): Likewise.
30472         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30473         (expand_vector_condition, optimize_vector_constructor): Likewise.
30474         (lower_vec_perm, get_compute_type): Likewise.
30475         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30476         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30477         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30478         (vect_recog_mask_conversion_pattern): Likewise.
30479         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30480         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30481         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30482         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30483         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30484         (vectorizable_shift, vectorizable_operation, vectorizable_store)
30485         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30486         (supportable_widening_operation): Likewise.
30487         (supportable_narrowing_operation): Likewise.
30488         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30489         Likewise.
30490         * varasm.c (output_constant): Likewise.
30492 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30493             Alan Hayward  <alan.hayward@arm.com>
30494             David Sherwood  <david.sherwood@arm.com>
30496         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30497         so that both the length == 3 and length != 3 cases set up their
30498         own permute vectors.  Add comments explaining why we know the
30499         number of elements is constant.
30500         (vect_permute_load_chain): Likewise.
30502 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30503             Alan Hayward  <alan.hayward@arm.com>
30504             David Sherwood  <david.sherwood@arm.com>
30506         * machmode.h (mode_nunits): Change from unsigned char to
30507         poly_uint16_pod.
30508         (ONLY_FIXED_SIZE_MODES): New macro.
30509         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30510         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30511         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30512         New typedefs.
30513         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30514         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30515         or if measurement_type is not polynomial.
30516         * genmodes.c (ZERO_COEFFS): New macro.
30517         (emit_mode_nunits_inline): Make mode_nunits_inline return a
30518         poly_uint16.
30519         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30520         Use ZERO_COEFFS when emitting initializers.
30521         * data-streamer.h (bp_pack_poly_value): New function.
30522         (bp_unpack_poly_value): Likewise.
30523         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30524         for GET_MODE_NUNITS.
30525         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30526         for GET_MODE_NUNITS.
30527         * tree.c (make_vector_type): Remove temporary shim and make
30528         the real function take the number of units as a poly_uint64
30529         rather than an int.
30530         (build_vector_type_for_mode): Handle polynomial nunits.
30531         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
30532         * emit-rtl.c (const_vec_series_p_1): Likewise.
30533         (gen_rtx_CONST_VECTOR): Likewise.
30534         * fold-const.c (test_vec_duplicate_folding): Likewise.
30535         * genrecog.c (validate_pattern): Likewise.
30536         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
30537         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30538         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
30539         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
30540         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
30541         * rtlanal.c (subreg_get_info): Likewise.
30542         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30543         (vect_grouped_load_supported): Likewise.
30544         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
30545         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
30546         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
30547         (simplify_const_unary_operation, simplify_binary_operation_1)
30548         (simplify_const_binary_operation, simplify_ternary_operation)
30549         (test_vector_ops_duplicate, test_vector_ops): Likewise.
30550         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
30551         instead of CONST_VECTOR_NUNITS.
30552         * varasm.c (output_constant_pool_2): Likewise.
30553         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
30554         explicit-encoded elements in the XVEC for variable-length vectors.
30556 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30558         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
30560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30561             Alan Hayward  <alan.hayward@arm.com>
30562             David Sherwood  <david.sherwood@arm.com>
30564         * coretypes.h (fixed_size_mode): Declare.
30565         (fixed_size_mode_pod): New typedef.
30566         * builtins.h (target_builtins::x_apply_args_mode)
30567         (target_builtins::x_apply_result_mode): Change type to
30568         fixed_size_mode_pod.
30569         * builtins.c (apply_args_size, apply_result_size, result_vector)
30570         (expand_builtin_apply_args_1, expand_builtin_apply)
30571         (expand_builtin_return): Update accordingly.
30573 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30575         * cse.c (hash_rtx_cb): Hash only the encoded elements.
30576         * cselib.c (cselib_hash_rtx): Likewise.
30577         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
30578         CONST_VECTOR encoding.
30580 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30581             Jeff Law  <law@redhat.com>
30583         PR target/83641
30584         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
30585         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
30586         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
30587         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
30589         PR target/83641
30590         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
30591         explicitly probe *sp in a noreturn function if there were any callee
30592         register saves or frame pointer is needed.
30594 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30596         PR debug/83621
30597         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
30598         BLKmode for ternary, binary or unary expressions.
30600         PR debug/83645
30601         * var-tracking.c (delete_vta_debug_insn): New inline function.
30602         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
30603         insns from get_insns () to NULL instead of each bb separately.
30604         Use delete_vta_debug_insn.  No longer static.
30605         (vt_debug_insns_local, variable_tracking_main_1): Adjust
30606         delete_vta_debug_insns callers.
30607         * rtl.h (delete_vta_debug_insns): Declare.
30608         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
30609         instead of variable_tracking_main.
30611 2018-01-03  Martin Sebor  <msebor@redhat.com>
30613         PR tree-optimization/83603
30614         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
30615         arguments past the endof the argument list in functions declared
30616         without a prototype.
30617         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
30618         Avoid checking when arguments are null.
30620 2018-01-03  Martin Sebor  <msebor@redhat.com>
30622         PR c/83559
30623         * doc/extend.texi (attribute const): Fix a typo.
30624         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
30625         issuing -Wsuggest-attribute for void functions.
30627 2018-01-03  Martin Sebor  <msebor@redhat.com>
30629         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
30630         offset_int::from instead of wide_int::to_shwi.
30631         (maybe_diag_overlap): Remove assertion.
30632         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
30633         * gimple-ssa-sprintf.c (format_directive): Same.
30634         (parse_directive): Same.
30635         (sprintf_dom_walker::compute_format_length): Same.
30636         (try_substitute_return_value): Same.
30638 2018-01-03  Jeff Law  <law@redhat.com>
30640         PR middle-end/83654
30641         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
30642         non-constant residual for zero at runtime and avoid probing in
30643         that case.  Reorganize code for trailing problem to mirror handling
30644         of the residual.
30646 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30648         PR tree-optimization/83501
30649         * tree-ssa-strlen.c (get_string_cst): New.
30650         (handle_char_store): Call get_string_cst.
30652 2018-01-03  Martin Liska  <mliska@suse.cz>
30654         PR tree-optimization/83593
30655         * tree-ssa-strlen.c: Include tree-cfg.h.
30656         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
30657         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
30658         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
30659         to false.
30660         (strlen_dom_walker::before_dom_children): Call
30661         gimple_purge_dead_eh_edges. Dump tranformation with details
30662         dump flags.
30663         (strlen_dom_walker::before_dom_children): Update call by adding
30664         new argument cleanup_eh.
30665         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
30667 2018-01-03  Martin Liska  <mliska@suse.cz>
30669         PR ipa/83549
30670         * cif-code.def (VARIADIC_THUNK): New enum value.
30671         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
30672         thunks.
30674 2018-01-03  Jan Beulich  <jbeulich@suse.com>
30676         * sse.md (mov<mode>_internal): Tighten condition for when to use
30677         vmovdqu<ssescalarsize> for TI and OI modes.
30679 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30681         Update copyright years.
30683 2018-01-03  Martin Liska  <mliska@suse.cz>
30685         PR ipa/83594
30686         * ipa-visibility.c (function_and_variable_visibility): Skip
30687         functions with noipa attribure.
30689 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30691         * gcc.c (process_command): Update copyright notice dates.
30692         * gcov-dump.c (print_version): Ditto.
30693         * gcov.c (print_version): Ditto.
30694         * gcov-tool.c (print_version): Ditto.
30695         * gengtype.c (create_file): Ditto.
30696         * doc/cpp.texi: Bump @copying's copyright year.
30697         * doc/cppinternals.texi: Ditto.
30698         * doc/gcc.texi: Ditto.
30699         * doc/gccint.texi: Ditto.
30700         * doc/gcov.texi: Ditto.
30701         * doc/install.texi: Ditto.
30702         * doc/invoke.texi: Ditto.
30704 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30706         * vector-builder.h (vector_builder::m_full_nelts): Change from
30707         unsigned int to poly_uint64.
30708         (vector_builder::full_nelts): Update prototype accordingly.
30709         (vector_builder::new_vector): Likewise.
30710         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
30711         (vector_builder::operator ==): Likewise.
30712         (vector_builder::finalize): Likewise.
30713         * int-vector-builder.h (int_vector_builder::int_vector_builder):
30714         Take the number of elements as a poly_uint64 rather than an
30715         unsigned int.
30716         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
30717         from unsigned int to poly_uint64.
30718         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
30719         (vec_perm_indices::new_vector): Likewise.
30720         (vec_perm_indices::length): Likewise.
30721         (vec_perm_indices::nelts_per_input): Likewise.
30722         (vec_perm_indices::input_nelts): Likewise.
30723         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
30724         number of elements per input as a poly_uint64 rather than an
30725         unsigned int.  Use the original encoding for variable-length
30726         vectors, rather than clamping each individual element.
30727         For the second and subsequent elements in each pattern,
30728         clamp the step and base before clamping their sum.
30729         (vec_perm_indices::series_p): Handle polynomial element counts.
30730         (vec_perm_indices::all_in_range_p): Likewise.
30731         (vec_perm_indices_to_tree): Likewise.
30732         (vec_perm_indices_to_rtx): Likewise.
30733         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
30734         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
30735         (tree_vector_builder::new_binary_operation): Handle polynomial
30736         element counts.  Return false if we need to know the number
30737         of elements at compile time.
30738         * fold-const.c (fold_vec_perm): Punt if the number of elements
30739         isn't known at compile time.
30741 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30743         * vec-perm-indices.h (vec_perm_builder): Change element type
30744         from HOST_WIDE_INT to poly_int64.
30745         (vec_perm_indices::element_type): Update accordingly.
30746         (vec_perm_indices::clamp): Handle polynomial element_types.
30747         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30748         (vec_perm_indices::all_in_range_p): Likewise.
30749         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
30750         than shwi trees.
30751         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
30752         polynomial vec_perm_indices element types.
30753         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
30754         * fold-const.c (fold_vec_perm): Likewise.
30755         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
30756         * tree-vect-generic.c (lower_vec_perm): Likewise.
30757         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30758         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
30759         element type to HOST_WIDE_INT.
30761 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30762             Alan Hayward  <alan.hayward@arm.com>
30763             David Sherwood  <david.sherwood@arm.com>
30765         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
30766         rather than an int.  Use plus_constant.
30767         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
30768         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
30770 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30771             Alan Hayward  <alan.hayward@arm.com>
30772             David Sherwood  <david.sherwood@arm.com>
30774         * calls.c (emit_call_1, expand_call): Change struct_value_size from
30775         a HOST_WIDE_INT to a poly_int64.
30777 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30778             Alan Hayward  <alan.hayward@arm.com>
30779             David Sherwood  <david.sherwood@arm.com>
30781         * calls.c (load_register_parameters): Cope with polynomial
30782         mode sizes.  Require a constant size for BLKmode parameters
30783         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
30784         forces a parameter to be padded at the lsb end in order to
30785         fill a complete number of words, require the parameter size
30786         to be ordered wrt UNITS_PER_WORD.
30788 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30789             Alan Hayward  <alan.hayward@arm.com>
30790             David Sherwood  <david.sherwood@arm.com>
30792         * reload1.c (spill_stack_slot_width): Change element type
30793         from unsigned int to poly_uint64_pod.
30794         (alter_reg): Treat mode sizes as polynomial.
30796 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30797             Alan Hayward  <alan.hayward@arm.com>
30798             David Sherwood  <david.sherwood@arm.com>
30800         * reload.c (complex_word_subreg_p): New function.
30801         (reload_inner_reg_of_subreg, push_reload): Use it.
30803 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30804             Alan Hayward  <alan.hayward@arm.com>
30805             David Sherwood  <david.sherwood@arm.com>
30807         * lra-constraints.c (process_alt_operands): Reject matched
30808         operands whose sizes aren't ordered.
30809         (match_reload): Refer to this check here.
30811 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30812             Alan Hayward  <alan.hayward@arm.com>
30813             David Sherwood  <david.sherwood@arm.com>
30815         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
30816         that the mode size is in the set {1, 2, 4, 8, 16}.
30818 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30819             Alan Hayward  <alan.hayward@arm.com>
30820             David Sherwood  <david.sherwood@arm.com>
30822         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
30823         Use plus_constant instead of gen_rtx_PLUS.
30825 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30826             Alan Hayward  <alan.hayward@arm.com>
30827             David Sherwood  <david.sherwood@arm.com>
30829         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
30830         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
30831         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
30832         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
30833         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
30834         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
30835         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
30836         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
30837         * config/i386/i386.c (ix86_push_rounding): ...this new function.
30838         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
30839         a poly_int64.
30840         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
30841         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
30842         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
30843         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
30844         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
30845         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
30846         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
30847         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
30848         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
30849         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
30850         function.
30851         * expr.c (emit_move_resolve_push): Treat the input and result
30852         of PUSH_ROUNDING as a poly_int64.
30853         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
30854         (emit_push_insn): Likewise.
30855         * lra-eliminations.c (mark_not_eliminable): Likewise.
30856         * recog.c (push_operand): Likewise.
30857         * reload1.c (elimination_effects): Likewise.
30858         * rtlanal.c (nonzero_bits1): Likewise.
30859         * calls.c (store_one_arg): Likewise.  Require the padding to be
30860         known at compile time.
30862 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30863             Alan Hayward  <alan.hayward@arm.com>
30864             David Sherwood  <david.sherwood@arm.com>
30866         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
30867         Use plus_constant instead of gen_rtx_PLUS.
30869 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30870             Alan Hayward  <alan.hayward@arm.com>
30871             David Sherwood  <david.sherwood@arm.com>
30873         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
30874         rather than an int.
30876 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30877             Alan Hayward  <alan.hayward@arm.com>
30878             David Sherwood  <david.sherwood@arm.com>
30880         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
30881         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
30882         via stack temporaries.  Treat the mode size as polynomial too.
30884 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30885             Alan Hayward  <alan.hayward@arm.com>
30886             David Sherwood  <david.sherwood@arm.com>
30888         * expr.c (expand_expr_real_2): When handling conversions involving
30889         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
30890         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
30891         as a poly_uint64 too.
30893 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30894             Alan Hayward  <alan.hayward@arm.com>
30895             David Sherwood  <david.sherwood@arm.com>
30897         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
30899 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30900             Alan Hayward  <alan.hayward@arm.com>
30901             David Sherwood  <david.sherwood@arm.com>
30903         * combine.c (can_change_dest_mode): Handle polynomial
30904         REGMODE_NATURAL_SIZE.
30905         * expmed.c (store_bit_field_1): Likewise.
30906         * expr.c (store_constructor): Likewise.
30907         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
30908         and polynomial REGMODE_NATURAL_SIZE.
30909         (gen_lowpart_common): Likewise.
30910         * reginfo.c (record_subregs_of_mode): Likewise.
30911         * rtlanal.c (read_modify_subreg_p): Likewise.
30913 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30914             Alan Hayward  <alan.hayward@arm.com>
30915             David Sherwood  <david.sherwood@arm.com>
30917         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
30918         numbers of elements.
30920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30921             Alan Hayward  <alan.hayward@arm.com>
30922             David Sherwood  <david.sherwood@arm.com>
30924         * match.pd: Cope with polynomial numbers of vector elements.
30926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30927             Alan Hayward  <alan.hayward@arm.com>
30928             David Sherwood  <david.sherwood@arm.com>
30930         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
30931         in a POINTER_PLUS_EXPR.
30933 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30934             Alan Hayward  <alan.hayward@arm.com>
30935             David Sherwood  <david.sherwood@arm.com>
30937         * omp-simd-clone.c (simd_clone_subparts): New function.
30938         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
30939         (ipa_simd_modify_function_body): Likewise.
30941 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30942             Alan Hayward  <alan.hayward@arm.com>
30943             David Sherwood  <david.sherwood@arm.com>
30945         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
30946         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
30947         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
30948         (expand_vector_condition, vector_element): Likewise.
30949         (subparts_gt): New function.
30950         (get_compute_type): Use subparts_gt.
30951         (count_type_subparts): Delete.
30952         (expand_vector_operations_1): Use subparts_gt instead of
30953         count_type_subparts.
30955 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30956             Alan Hayward  <alan.hayward@arm.com>
30957             David Sherwood  <david.sherwood@arm.com>
30959         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
30960         (vect_compile_time_alias): ...this new function.  Do the calculation
30961         on poly_ints rather than trees.
30962         (vect_prune_runtime_alias_test_list): Update call accordingly.
30964 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30965             Alan Hayward  <alan.hayward@arm.com>
30966             David Sherwood  <david.sherwood@arm.com>
30968         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
30969         numbers of units.
30970         (vect_schedule_slp_instance): Likewise.
30972 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30973             Alan Hayward  <alan.hayward@arm.com>
30974             David Sherwood  <david.sherwood@arm.com>
30976         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
30977         constant and extern definitions for variable-length vectors.
30978         (vect_get_constant_vectors): Note that the number of units
30979         is known to be constant.
30981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30982             Alan Hayward  <alan.hayward@arm.com>
30983             David Sherwood  <david.sherwood@arm.com>
30985         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
30986         of units as polynomial.  Choose between WIDE and NARROW based
30987         on multiple_p.
30989 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30990             Alan Hayward  <alan.hayward@arm.com>
30991             David Sherwood  <david.sherwood@arm.com>
30993         * tree-vect-stmts.c (simd_clone_subparts): New function.
30994         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
30996 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30997             Alan Hayward  <alan.hayward@arm.com>
30998             David Sherwood  <david.sherwood@arm.com>
31000         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31001         vectors as polynomial.  Use build_index_vector for
31002         IFN_GOMP_SIMD_LANE.
31004 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31005             Alan Hayward  <alan.hayward@arm.com>
31006             David Sherwood  <david.sherwood@arm.com>
31008         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31009         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31010         for variable-length vectors.
31011         (vectorizable_mask_load_store): Treat the number of units as
31012         polynomial, asserting that it is constant if the condition has
31013         already been enforced.
31014         (vectorizable_store, vectorizable_load): Likewise.
31016 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31017             Alan Hayward  <alan.hayward@arm.com>
31018             David Sherwood  <david.sherwood@arm.com>
31020         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31021         of units as polynomial.  Punt if we can't tell at compile time
31022         which vector contains the final result.
31024 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31025             Alan Hayward  <alan.hayward@arm.com>
31026             David Sherwood  <david.sherwood@arm.com>
31028         * tree-vect-loop.c (vectorizable_induction): Treat the number
31029         of units as polynomial.  Punt on SLP inductions.  Use an integer
31030         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31031         cast of such a series for variable-length floating-point
31032         reductions.
31034 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31035             Alan Hayward  <alan.hayward@arm.com>
31036             David Sherwood  <david.sherwood@arm.com>
31038         * tree.h (build_index_vector): Declare.
31039         * tree.c (build_index_vector): New function.
31040         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31041         of units as polynomial, forcibly converting it to a constant if
31042         vectorizable_reduction has already enforced the condition.
31043         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31044         to create a {1,2,3,...} vector.
31045         (vectorizable_reduction): Treat the number of units as polynomial.
31046         Choose vectype_in based on the largest scalar element size rather
31047         than the smallest number of units.  Enforce the restrictions
31048         relied on above.
31050 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31051             Alan Hayward  <alan.hayward@arm.com>
31052             David Sherwood  <david.sherwood@arm.com>
31054         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31055         number of units as polynomial.
31057 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31058             Alan Hayward  <alan.hayward@arm.com>
31059             David Sherwood  <david.sherwood@arm.com>
31061         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31062         * target.def (autovectorize_vector_sizes): Return the vector sizes
31063         by pointer, using vector_sizes rather than a bitmask.
31064         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31065         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31066         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31067         Likewise.
31068         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31069         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31070         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31071         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31072         * omp-general.c (omp_max_vf): Likewise.
31073         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31074         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31075         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31076         * tree-vect-slp.c (vect_slp_bb): Likewise.
31077         * doc/tm.texi: Regenerate.
31078         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31079         to a poly_uint64.
31080         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31081         the vector size as a poly_uint64 rather than an unsigned int.
31082         (current_vector_size): Change from an unsigned int to a poly_uint64.
31083         (get_vectype_for_scalar_type): Update accordingly.
31084         * tree.h (build_truth_vector_type): Take the size and number of
31085         units as a poly_uint64 rather than an unsigned int.
31086         (build_vector_type): Add a temporary overload that takes
31087         the number of units as a poly_uint64 rather than an unsigned int.
31088         * tree.c (make_vector_type): Likewise.
31089         (build_truth_vector_type): Take the number of units as a poly_uint64
31090         rather than an unsigned int.
31092 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31093             Alan Hayward  <alan.hayward@arm.com>
31094             David Sherwood  <david.sherwood@arm.com>
31096         * target.def (get_mask_mode): Take the number of units and length
31097         as poly_uint64s rather than unsigned ints.
31098         * targhooks.h (default_get_mask_mode): Update accordingly.
31099         * targhooks.c (default_get_mask_mode): Likewise.
31100         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31101         * doc/tm.texi: Regenerate.
31103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31104             Alan Hayward  <alan.hayward@arm.com>
31105             David Sherwood  <david.sherwood@arm.com>
31107         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31108         * omp-general.c (omp_max_vf): Likewise.
31109         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31110         (expand_omp_simd): Handle polynomial safelen.
31111         * omp-low.c (omplow_simd_context): Add a default constructor.
31112         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31113         (lower_rec_simd_input_clauses): Update accordingly.
31114         (lower_rec_input_clauses): Likewise.
31116 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31117             Alan Hayward  <alan.hayward@arm.com>
31118             David Sherwood  <david.sherwood@arm.com>
31120         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31121         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31122         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31123         (vect_analyze_slp_cost): Likewise.
31124         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31125         (vect_model_load_cost): Likewise.
31127 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31128             Alan Hayward  <alan.hayward@arm.com>
31129             David Sherwood  <david.sherwood@arm.com>
31131         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31132         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31133         from an unsigned int * to a poly_uint64_pod *.
31134         (calculate_unrolling_factor): New function.
31135         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
31137 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31138             Alan Hayward  <alan.hayward@arm.com>
31139             David Sherwood  <david.sherwood@arm.com>
31141         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31142         from an unsigned int to a poly_uint64.
31143         (_loop_vec_info::slp_unrolling_factor): Likewise.
31144         (_loop_vec_info::vectorization_factor): Change from an int
31145         to a poly_uint64.
31146         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31147         (vect_get_num_vectors): New function.
31148         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31149         (vect_get_num_copies): Use vect_get_num_vectors.
31150         (vect_analyze_data_ref_dependences): Change max_vf from an int *
31151         to an unsigned int *.
31152         (vect_analyze_data_refs): Change min_vf from an int * to a
31153         poly_uint64 *.
31154         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31155         than an unsigned HOST_WIDE_INT.
31156         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31157         (vect_analyze_data_ref_dependence): Change max_vf from an int *
31158         to an unsigned int *.
31159         (vect_analyze_data_ref_dependences): Likewise.
31160         (vect_compute_data_ref_alignment): Handle polynomial vf.
31161         (vect_enhance_data_refs_alignment): Likewise.
31162         (vect_prune_runtime_alias_test_list): Likewise.
31163         (vect_shift_permute_load_chain): Likewise.
31164         (vect_supportable_dr_alignment): Likewise.
31165         (dependence_distance_ge_vf): Take the vectorization factor as a
31166         poly_uint64 rather than an unsigned HOST_WIDE_INT.
31167         (vect_analyze_data_refs): Change min_vf from an int * to a
31168         poly_uint64 *.
31169         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31170         vfm1 as a poly_uint64 rather than an int.  Make the same change
31171         for the returned bound_scalar.
31172         (vect_gen_vector_loop_niters): Handle polynomial vf.
31173         (vect_do_peeling): Likewise.  Update call to
31174         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31175         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31176         be constant.
31177         * tree-vect-loop.c (vect_determine_vectorization_factor)
31178         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31179         (vect_get_known_peeling_cost): Likewise.
31180         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31181         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31182         (vect_transform_loop): Likewise.  Use the lowest possible VF when
31183         updating the upper bounds of the loop.
31184         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
31185         rather than an int.
31186         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31187         polynomial unroll factors.
31188         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31189         (vect_make_slp_decision): Likewise.
31190         (vect_supported_load_permutation_p): Likewise, and polynomial
31191         vf too.
31192         (vect_analyze_slp_cost): Handle polynomial vf.
31193         (vect_slp_analyze_node_operations): Likewise.
31194         (vect_slp_analyze_bb_1): Likewise.
31195         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31196         than an unsigned HOST_WIDE_INT.
31197         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31198         (vectorizable_load): Handle polynomial vf.
31199         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31200         a poly_uint64.
31201         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31204             Alan Hayward  <alan.hayward@arm.com>
31205             David Sherwood  <david.sherwood@arm.com>
31207         * match.pd: Handle bit operations involving three constants
31208         and try to fold one pair.
31210 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31212         * tree-vect-loop-manip.c: Include gimple-fold.h.
31213         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31214         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31215         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31216         Add a path that uses a step of VF instead of 1, but disable it
31217         for now.
31218         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31219         and niters_no_overflow parameters.  Update calls to
31220         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31221         Create a new SSA name if the latter choses to use a ste other
31222         than zero, and return it via niters_vector_mult_vf_var.
31223         * tree-vect-loop.c (vect_transform_loop): Update calls to
31224         vect_do_peeling, vect_gen_vector_loop_niters and
31225         slpeel_make_loop_iterate_ntimes.
31226         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31227         (vect_gen_vector_loop_niters): Update declarations after above changes.
31229 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31231         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31232         128-bit round to integer instructions.
31233         (ceil<mode>2): Likewise.
31234         (btrunc<mode>2): Likewise.
31235         (round<mode>2): Likewise.
31237 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31239         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31240         unaligned VSX load/store on P8/P9.
31241         (expand_block_clear): Allow the use of unaligned VSX
31242         load/store on P8/P9.
31244 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31246         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31247         New function.
31248         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31249         swap associated with both a load and a store.
31251 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31253         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31254         * config/riscv/riscv.md (clear_cache): Use it.
31256 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31258         * web.c: Remove out-of-date comment.
31260 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31262         * expr.c (fixup_args_size_notes): Check that any existing
31263         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31264         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31265         (emit_single_push_insn): ...here.
31267 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31269         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31270         (const_vector_encoded_nelts): New function.
31271         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31272         (const_vector_int_elt, const_vector_elt): Declare.
31273         * emit-rtl.c (const_vector_int_elt_1): New function.
31274         (const_vector_elt): Likewise.
31275         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31276         of CONST_VECTOR_ELT.
31278 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31280         * expr.c: Include rtx-vector-builder.h.
31281         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31282         directly on the tree encoding.
31283         (const_vector_from_tree): Likewise.
31284         * optabs.c: Include rtx-vector-builder.h.
31285         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31286         sequence of "u" values.
31287         * vec-perm-indices.c: Include rtx-vector-builder.h.
31288         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31289         directly on the vec_perm_indices encoding.
31291 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31293         * doc/rtl.texi (const_vector): Describe new encoding scheme.
31294         * Makefile.in (OBJS): Add rtx-vector-builder.o.
31295         * rtx-vector-builder.h: New file.
31296         * rtx-vector-builder.c: Likewise.
31297         * rtl.h (rtx_def::u2): Add a const_vector field.
31298         (CONST_VECTOR_NPATTERNS): New macro.
31299         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31300         (CONST_VECTOR_DUPLICATE_P): Likewise.
31301         (CONST_VECTOR_STEPPED_P): Likewise.
31302         (CONST_VECTOR_ENCODED_ELT): Likewise.
31303         (const_vec_duplicate_p): Check for a duplicated vector encoding.
31304         (unwrap_const_vec_duplicate): Likewise.
31305         (const_vec_series_p): Check for a non-duplicated vector encoding.
31306         Say that the function only returns true for integer vectors.
31307         * emit-rtl.c: Include rtx-vector-builder.h.
31308         (gen_const_vec_duplicate_1): Delete.
31309         (gen_const_vector): Call gen_const_vec_duplicate instead of
31310         gen_const_vec_duplicate_1.
31311         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31312         (gen_const_vec_duplicate): Use rtx_vector_builder.
31313         (gen_const_vec_series): Likewise.
31314         (gen_rtx_CONST_VECTOR): Likewise.
31315         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31316         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31317         Build a new vector rather than modifying a CONST_VECTOR in-place.
31318         (handle_special_swappables): Update call accordingly.
31319         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31320         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31321         Build a new vector rather than modifying a CONST_VECTOR in-place.
31322         (handle_special_swappables): Update call accordingly.
31324 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31326         * simplify-rtx.c (simplify_const_binary_operation): Use
31327         CONST_VECTOR_ELT instead of XVECEXP.
31329 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31331         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31332         the selector elements to be different from the data elements
31333         if the selector is a VECTOR_CST.
31334         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31335         ssizetype for the selector.
31337 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31339         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31340         before testing each element individually.
31341         * tree-vect-generic.c (lower_vec_perm): Likewise.
31343 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31345         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31346         * selftest-run-tests.c (selftest::run_tests): Call it.
31347         * vector-builder.h (vector_builder::operator ==): New function.
31348         (vector_builder::operator !=): Likewise.
31349         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31350         (vec_perm_indices::all_from_input_p): New function.
31351         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31352         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31353         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31354         instead of reading the VECTOR_CST directly.  Detect whether both
31355         vector inputs are the same before constructing the vec_perm_indices,
31356         and update the number of inputs argument accordingly.  Use the
31357         utility functions added above.  Only construct sel2 if we need to.
31359 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31361         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31362         the broadcast of the low byte.
31363         (expand_mult_highpart): Use an explicit encoding for the permutes.
31364         * optabs-query.c (can_mult_highpart_p): Likewise.
31365         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31366         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31367         (vectorizable_bswap): Likewise.
31368         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31369         explicit encoding for the power-of-2 permutes.
31370         (vect_permute_store_chain): Likewise.
31371         (vect_grouped_load_supported): Likewise.
31372         (vect_permute_load_chain): Likewise.
31374 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31376         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31377         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31378         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31379         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31380         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31381         (vect_gen_perm_mask_any): Likewise.
31383 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31385         * int-vector-builder.h: New file.
31386         * vec-perm-indices.h: Include int-vector-builder.h.
31387         (vec_perm_indices): Redefine as an int_vector_builder.
31388         (auto_vec_perm_indices): Delete.
31389         (vec_perm_builder): Redefine as a stand-alone class.
31390         (vec_perm_indices::vec_perm_indices): New function.
31391         (vec_perm_indices::clamp): Likewise.
31392         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31393         (vec_perm_indices::new_vector): New function.
31394         (vec_perm_indices::new_expanded_vector): Update for new
31395         vec_perm_indices class.
31396         (vec_perm_indices::rotate_inputs): New function.
31397         (vec_perm_indices::all_in_range_p): Operate directly on the
31398         encoded form, without computing elided elements.
31399         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31400         encoding.  Update for new vec_perm_indices class.
31401         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31402         the given vec_perm_builder.
31403         (expand_vec_perm_var): Update vec_perm_builder constructor.
31404         (expand_mult_highpart): Use vec_perm_builder instead of
31405         auto_vec_perm_indices.
31406         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31407         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
31408         or double series encoding as appropriate.
31409         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31410         vec_perm_indices instead of auto_vec_perm_indices.
31411         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31412         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31413         (vect_permute_store_chain): Likewise.
31414         (vect_grouped_load_supported): Likewise.
31415         (vect_permute_load_chain): Likewise.
31416         (vect_shift_permute_load_chain): Likewise.
31417         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31418         (vect_transform_slp_perm_load): Likewise.
31419         (vect_schedule_slp_instance): Likewise.
31420         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31421         (vectorizable_mask_load_store): Likewise.
31422         (vectorizable_bswap): Likewise.
31423         (vectorizable_store): Likewise.
31424         (vectorizable_load): Likewise.
31425         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31426         vec_perm_indices instead of auto_vec_perm_indices.  Use
31427         tree_to_vec_perm_builder to read the vector from a tree.
31428         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31429         vec_perm_builder instead of a vec_perm_indices.
31430         (have_whole_vector_shift): Use vec_perm_builder and
31431         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
31432         truncation to calc_vec_perm_mask_for_shift.
31433         (vect_create_epilog_for_reduction): Likewise.
31434         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31435         from auto_vec_perm_indices to vec_perm_indices.
31436         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31437         instead of changing individual elements.
31438         (aarch64_vectorize_vec_perm_const): Use new_vector to install
31439         the vector in d.perm.
31440         * config/arm/arm.c (expand_vec_perm_d::perm): Change
31441         from auto_vec_perm_indices to vec_perm_indices.
31442         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31443         instead of changing individual elements.
31444         (arm_vectorize_vec_perm_const): Use new_vector to install
31445         the vector in d.perm.
31446         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31447         Update vec_perm_builder constructor.
31448         (rs6000_expand_interleave): Likewise.
31449         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31450         (rs6000_expand_interleave): Likewise.
31452 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31454         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31455         to qimode could truncate the indices.
31456         * optabs.c (expand_vec_perm_var): Likewise.
31458 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31460         * Makefile.in (OBJS): Add vec-perm-indices.o.
31461         * vec-perm-indices.h: New file.
31462         * vec-perm-indices.c: Likewise.
31463         * target.h (vec_perm_indices): Replace with a forward class
31464         declaration.
31465         (auto_vec_perm_indices): Move to vec-perm-indices.h.
31466         * optabs.h: Include vec-perm-indices.h.
31467         (expand_vec_perm): Delete.
31468         (selector_fits_mode_p, expand_vec_perm_var): Declare.
31469         (expand_vec_perm_const): Declare.
31470         * target.def (vec_perm_const_ok): Replace with...
31471         (vec_perm_const): ...this new hook.
31472         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31473         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31474         * doc/tm.texi: Regenerate.
31475         * optabs.def (vec_perm_const): Delete.
31476         * doc/md.texi (vec_perm_const): Likewise.
31477         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31478         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31479         expand_vec_perm for constant permutation vectors.  Assert that
31480         the mode of variable permutation vectors is the integer equivalent
31481         of the mode that is being permuted.
31482         * optabs-query.h (selector_fits_mode_p): Declare.
31483         * optabs-query.c: Include vec-perm-indices.h.
31484         (selector_fits_mode_p): New function.
31485         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31486         is defined, instead of checking whether the vec_perm_const_optab
31487         exists.  Use targetm.vectorize.vec_perm_const instead of
31488         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
31489         fit in the vector mode before using a variable permute.
31490         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31491         vec_perm_indices instead of an rtx.
31492         (expand_vec_perm): Replace with...
31493         (expand_vec_perm_const): ...this new function.  Take the selector
31494         as a vec_perm_indices rather than an rtx.  Also take the mode of
31495         the selector.  Update call to shift_amt_for_vec_perm_mask.
31496         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31497         Use vec_perm_indices::new_expanded_vector to expand the original
31498         selector into bytes.  Check whether the indices fit in the vector
31499         mode before using a variable permute.
31500         (expand_vec_perm_var): Make global.
31501         (expand_mult_highpart): Use expand_vec_perm_const.
31502         * fold-const.c: Includes vec-perm-indices.h.
31503         * tree-ssa-forwprop.c: Likewise.
31504         * tree-vect-data-refs.c: Likewise.
31505         * tree-vect-generic.c: Likewise.
31506         * tree-vect-loop.c: Likewise.
31507         * tree-vect-slp.c: Likewise.
31508         * tree-vect-stmts.c: Likewise.
31509         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31510         Delete.
31511         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31512         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31513         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31514         (aarch64_vectorize_vec_perm_const): ...this new function.
31515         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31516         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31517         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31518         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31519         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31520         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31521         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31522         into...
31523         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
31524         check for NEON modes.
31525         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
31526         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
31527         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
31528         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
31529         into...
31530         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
31531         the old VEC_PERM_CONST conditions.
31532         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
31533         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
31534         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
31535         (ia64_vectorize_vec_perm_const_ok): Merge into...
31536         (ia64_vectorize_vec_perm_const): ...this new function.
31537         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
31538         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
31539         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
31540         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
31541         * config/mips/mips.c (mips_expand_vec_perm_const)
31542         (mips_vectorize_vec_perm_const_ok): Merge into...
31543         (mips_vectorize_vec_perm_const): ...this new function.
31544         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
31545         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
31546         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
31547         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
31548         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
31549         (rs6000_expand_vec_perm_const): Delete.
31550         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
31551         Delete.
31552         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31553         (altivec_expand_vec_perm_const_le): Take each operand individually.
31554         Operate on constant selectors rather than rtxes.
31555         (altivec_expand_vec_perm_const): Likewise.  Update call to
31556         altivec_expand_vec_perm_const_le.
31557         (rs6000_expand_vec_perm_const): Delete.
31558         (rs6000_vectorize_vec_perm_const_ok): Delete.
31559         (rs6000_vectorize_vec_perm_const): New function.
31560         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31561         an element count and rtx array.
31562         (rs6000_expand_extract_even): Update call accordingly.
31563         (rs6000_expand_interleave): Likewise.
31564         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
31565         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
31566         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
31567         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
31568         (rs6000_expand_vec_perm_const): Delete.
31569         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31570         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31571         (altivec_expand_vec_perm_const_le): Take each operand individually.
31572         Operate on constant selectors rather than rtxes.
31573         (altivec_expand_vec_perm_const): Likewise.  Update call to
31574         altivec_expand_vec_perm_const_le.
31575         (rs6000_expand_vec_perm_const): Delete.
31576         (rs6000_vectorize_vec_perm_const_ok): Delete.
31577         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
31578         reference to the SPE evmerge intructions.
31579         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31580         an element count and rtx array.
31581         (rs6000_expand_extract_even): Update call accordingly.
31582         (rs6000_expand_interleave): Likewise.
31583         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
31584         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
31585         new function.
31586         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31588 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31590         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
31591         vector mode and that that mode matches the mode of the data
31592         being permuted.
31593         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
31594         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
31595         directly using expand_vec_perm_1 when forcing selectors into
31596         registers.
31597         (expand_vec_perm_var): New function, split out from expand_vec_perm.
31599 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31601         * optabs-query.h (can_vec_perm_p): Delete.
31602         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
31603         * optabs-query.c (can_vec_perm_p): Split into...
31604         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
31605         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
31606         particular selector is valid.
31607         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31608         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31609         (vect_grouped_load_supported): Likewise.
31610         (vect_shift_permute_load_chain): Likewise.
31611         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31612         (vect_transform_slp_perm_load): Likewise.
31613         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31614         (vectorizable_bswap): Likewise.
31615         (vect_gen_perm_mask_checked): Likewise.
31616         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
31617         implementations of variable permutation vectors into account
31618         when deciding which selector to use.
31619         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
31620         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
31621         with a false third argument.
31622         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
31623         to test whether the constant selector is valid and can_vec_perm_var_p
31624         to test whether a variable selector is valid.
31626 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31628         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
31629         * optabs-query.c (can_vec_perm_p): Likewise.
31630         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
31631         instead of vec_perm_indices.
31632         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
31633         (vect_gen_perm_mask_checked): Likewise,
31634         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
31635         (vect_gen_perm_mask_checked): Likewise,
31637 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31639         * optabs-query.h (qimode_for_vec_perm): Declare.
31640         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
31641         (qimode_for_vec_perm): ...this new function.
31642         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
31644 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31646         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
31647         does not have a conditional at the top.
31649 2018-01-02  Richard Biener  <rguenther@suse.de>
31651         * ipa-inline.c (big_speedup_p): Fix expression.
31653 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31655         PR target/81616
31656         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
31657         for generic 4->6.
31659 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31661         PR target/81616
31662         Generic tuning.
31663         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
31664         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
31665         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
31666         cond_taken_branch_cost 3->4.
31668 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
31670         PR tree-optimization/83581
31671         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
31672         TODO_cleanup_cfg if any changes have been made.
31674         PR middle-end/83608
31675         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
31676         convert_modes if target mode has the right side, but different mode
31677         class.
31679         PR middle-end/83609
31680         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
31681         last argument when extracting from CONCAT.  If either from_real or
31682         from_imag is NULL, use expansion through memory.  If result is not
31683         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
31684         the parts directly to inner mode, if even that fails, use expansion
31685         through memory.
31687         PR middle-end/83623
31688         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
31689         check for bswap in mode rather than HImode and use that in expand_unop
31690         too.
31692 Copyright (C) 2018 Free Software Foundation, Inc.
31694 Copying and distribution of this file, with or without modification,
31695 are permitted in any medium without royalty provided the copyright
31696 notice and this notice are preserved.