2018-11-11 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob2997cf7b486e4c2464a67c9ac6c76761eea8b682
1 2018-11-11  Richard Biener  <rguenther@suse.de>
3         * tree-vrp.h (class value_range_base): New base class for
4         value_range containing all but the m_equiv member.
5         (dump_value_range_base): Add.
6         (range_includes_zero_p): Work on value_range_base.
7         * tree-vrp.c (value_range_base::set): Split out base handling
8         from...
9         (value_range::set): this.
10         (value_range::set_equiv): New.
11         (value_range_base::value_range_base): New constructors.
12         (value_range_base::check): Split out base handling from...
13         (value_range::check): this.
14         (value_range::equal_p): Refactor in terms of
15         ignore_equivs_equal_p which is now member of the base.
16         (value_range_base::set_undefined): New.
17         (value_range_base::set_varying): Likewise.
18         (value_range_base::dump):Split out base handling from...
19         (value_range::dump): this.
20         (value_range_base::set_and_canonicalize): Split out base handling
21         from...
22         (value_range::set_and_canonicalize): this.
23         (value_range_base::union_): New.
24         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
25         for m_vr.
26         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
27         instead of value_range everywhere.
28         (ipcp_vr_lattice::print): Use dump_value_range_base.
29         (ipcp_vr_lattice::meet_with): Adjust.
30         (ipcp_vr_lattice::meet_with_1): Likewise.
31         (ipa_vr_operation_and_type_effects): Likewise.
32         (propagate_vr_across_jump_function): Likewise.
33         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
34         (ipa_get_value_range): Likewise.
35         (ipa_set_jfunc_vr): Likewise.
36         (ipa_compute_jump_functions_for_edge): Likewise.
38 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
40         PR middle-end/65703
41         * doc/invoke.texi (Optimize Options): Add @opindex entries
42         for the positive forms of -fno-xxx and -mno-xxx options
43         that were lacking them.
45 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
47         * combine.c (make_more_copies): Only make an intermediate copy if the
48         dest of a move is a pseudo.
50 2018-11-09  Maya Rashish  <coypu@sdf.org>
52         PR target/87221
53         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
54         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
56 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
58         PR driver/41179
59         PR middle-end/65703
60         * doc/invoke.texi (Optimize Options): Clarify default behavior
61         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
63 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
65         PR tree-optimization/87940
66         * expr.c (string_constant): Don't strip NOPS in subexpressions.
67         Fold PLUS_EXPR correctly.
69 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
71         PR target/87762
72         * config/s390/s390.c (s390_safe_relative_long_p): New function.
73         (annotate_constant_pool_refs): Skip insns which support
74         relative addressing.
75         (annotate_constant_pool_refs_1): New helper function.
76         (find_constant_pool_ref): Skip insns which support relative
77         addression.
78         (find_constant_pool_ref_1): New helper function.
79         (replace_constant_pool_ref): Skip insns which support
80         relative addressing.
81         (replace_constant_pool_ref_1): New helper function.
82         (s390_mainpool_start): Adapt to the new signature.
83         (s390_mainpool_finish): Likewise.
84         (s390_chunkify_start): Likewise.
85         (s390_chunkify_finish): Likewise.
86         (pass_s390_early_mach::execute): Likewise.
87         (s390_prologue_plus_offset): Likewise.
88         (s390_emit_prologue): Likewise.
89         (s390_emit_epilogue): Likewise.
91 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
93         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
94         but unsupported lastprivate with conditional modifier.
96 2018-11-09  Jeff Law  <law@redhat.com>
98         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
99         unused argument better.  Add gcc_unreachable to silence warning.
101 2018-11-09  Martin Sebor  <msebor@redhat.com>
103         PR middle-end/81824
104         * attribs.c (has_attribute): New helper function.
105         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
106         * attribs.h (decls_mismatched_attributes): Declare.
107         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
108         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
109         * common.opt (-Wattribute-alias): Take an argument.
110         (-Wno-attribute-alias): New option.
111         * doc/extend.texi (Common Function Attributes): Document copy.
112         (Common Variable Attributes): Same.
113         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
114         (-Wattribute-alias): Document new option argument.
116 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
118         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
119         matching criteria.  Remove unused array initializer.
121 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
122             Jinsong Ji  <jji@us.ibm.com>
124         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
125         constraints by introducing a new temporary.
126         (_mm_cvtss_si64): Likewise.
128 2018-11-09  Martin Liska  <mliska@suse.cz>
130         * common.opt: Add -fipa-stack-alignment flag.
131         * doc/invoke.texi: Document it.
132         * final.c (rest_of_clean_state): Guard stack
133         shrinking with flag.
135 2018-11-09  Martin Liska  <mliska@suse.cz>
137         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
138         to ...
139         (ipa_discover_variable_flags): ... this.
140         * common.opt: Come up with new flag -fipa-reference-addressable.
141         * doc/invoke.texi: Document it.
142         * ipa-reference.c (propagate): Call the renamed fn.
143         * ipa-visibility.c (whole_program_function_and_variable_visibility):
144         Likewise.
145         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
146         ...
147         (ipa_discover_variable_flags): ... this.  Discover
148         non-addressable variables only with the newly added flag.
149         * opts.c: Enable the newly added flag with -O1 and higher
150         optimization level.
152 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
154         * json.cc (selftest::test_writing_literals): Fix comment.
156 2018-11-09  Martin Liska  <mliska@suse.cz>
158         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
159         string to a stack buffer.
160         (aarch64_parse_cpu): Likewise.
161         (aarch64_parse_tune): Likewise.
163 2018-11-09  Richard Biener  <rguenther@suse.de>
165         PR tree-optimization/87953
166         * tree-vect-loop.c (vectorizable_reduction): For analysis
167         always pass ops[0] to vectorizable_condition.
169 2018-11-09  Stafford Horne  <shorne@gmail.com>
170             Richard Henderson  <rth@twiddle.net>
171             Joel Sherrill  <joel@rtems.org>
173         * common/config/or1k/or1k-common.c: New file.
174         * config/or1k/*: New.
175         * config.gcc (or1k*-*-*): New.
176         * configure.ac (or1k*-*-*): New test for openrisc tls.
177         * configure: Regenerated.
178         * doc/install.texi: Document OpenRISC triplets.
179         * doc/invoke.texi: Document OpenRISC arguments.
180         * doc/md.texi: Document OpenRISC.
182 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
184         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
185         (arm7tdmi-s): Delete CPU.
186         (arm710t): Add aliases for arm720t and arm740t.
187         (arm720t, arm740t): Delete CPUs.
188         (arm920t): Add aliases for arm920, arm922t and arm940t.
189         (arm920, arm922t, arm940t): Delete CPUs.
190         (arm10tdmi): Add alias for arm1020t.
191         (arm1020t): Delete CPU.
192         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
193         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
194         (arm10e): Add aliases for arm1020e and arm1022e.
195         (arm1020e, arm1022e): Delete CPU.
196         * config/arm/arm.md (generic_sched): Remove entries that are now
197         handled by aliases.
198         (generic_vfp): Likewise.
199         * config/arm/arm1020e.md: Simplify tuning selection based on alias
200         changes.
201         * config/arm/arm-tune.md: Regenerated.
202         * config/arm/arm-tables.opt: Regenerated.
204 2018-11-09  Richard Biener  <rguenther@suse.de>
206         PR tree-optimization/87621
207         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
208         op with only phi inputs.
209         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
210         (ch_base::copy_headers): Run CSE on copied loop headers.
211         (pass_ch_vect::process_loop_p): Simplify.
213 2018-11-09  Alexandre Oliva <oliva@adacore.com>
215         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
216         for non-w64 x86_64 biarch.
218 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
220         PR rtl-optimization/86438
221         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
222         of in_b for the compare if in_b is SET_DEST.
224         PR target/87793
225         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
226         non-toplevel UNSPEC.
228 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
230         * tree-vrp.c (value_range::check): Do not access internals
231         directly.
232         (value_range::singleton_p): Same.
233         (value_range::type): Same.
234         (vrp_finalize): Use value_range API.
236 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
238         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
240 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
242         * vr-values.c (vr_values::get_value_range): Use value_range API
243         instead of piecing together ranges.
244         (vr_values::update_value_range): Same.
246 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
248         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
249         of performing ad-hoc calculations.
250         * tree-ssanames.c (set_range_info): New overloaded function
251         accepting value_range &.
252         (get_range_info): Same.
253         * tree-ssanames.h (set_range_info_raw): Remove.
254         (set_range_info): New prototype.
255         (get_range_info): Same.
256         * tree-vrp.h (value_range::null_p): Rename to zero_p.
257         * tree-vrp.c (value_range::null_p): Same.
259 2018-11-09  Jan Hubicka  <jh@suse.cz>
261         * tree.c (fld_type_variant_equal_p): Test user align flag.
262         (flt_type_variant): Copy user align flag.
263         (fld_incomplete_type_of): Clear it.
265 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
267         * config/arm/neon.md (div<mode>3): New pattern.
269 2018-11-08  Andi Kleen  <ak@linux.intel.com>
271         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
272         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
273         (ix86_handle_option): Handle OPT_mptwrite.
274         * config/i386/cpuid.h (bit_PTWRITE): Add.
275         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
276         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
277         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
278         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
279         (ix86_option_override_internal): Handle PTA_PTWRITE.
280         (ix86_valid_target_attribute_inner_p): Define ptwrite.
281         (def_builtin2): Force UINT64 to be 64bit only.
282         * config/i386/i386.h (TARGET_PTWRITE): Add.
283         (TARGET_PTWRITE_P): Add.
284         (PTA_PTWRITE): Add.
285         * config/i386/i386.md: Define ptwrite.
286         * config/i386/i386.opt: Add -mptwrite.
287         * config/i386/immintrin.h (_ptwrite64): Add.
288         (_ptwrite32): Add
289         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
290         * doc/invoke.texi: Document -mptwrite.
292 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
294         PR rtl-optimization/87600
295         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
296         * lra-constraints.c (process_alt_operands): Skip illegal hard
297         register usage.  Prefer reloading non hard register operands.
299 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
301         PR other/36572
302         * doc/invoke.texi (Optimize Options): Clarify default behavior
303         for -fno-sched-interblock and -fno-sched-spec.
305 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
307         * collect2.c (linker_select):  Add USE_LLD_LD.
308         (ld_suffixes): Add ld.lld.
309         (main): Handle -fuse-ld=lld.
310         * common.opt (-fuse-ld=lld): New option.
311         * doc/invoke.texi (-fuse-ld=lld): Document.
312         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
314 2018-11-08  Paul Koning  <ni1d@arrl.net>
316         * config/pdp11/constraints.md: Add "Z" series constraints for use
317         with pre-dec and post-inc addressing.
318         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
319         (pdp11_expand_operands): Add int argument (word count).
320         (pdp11_sp_frame_offset): Delete.
321         (pdp11_cmp_length): New function.
322         (pushpop_regeq): New function.
323         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
324         Add hook.
325         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
326         frame layout.
327         (pdp11_initial_elimination_offset): Ditto.
328         (pdp11_expand_operands): Add word count argument.  Bugfixes.
329         (output_move_multiple): Change how pointer adjustment is done.
330         (pdp11_gen_int_label): Correct format.
331         (output_ascii): Ditto.
332         (pdp11_asm_output_var): Add code for DEC assembler case.
333         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
334         value.
335         (legitimate_const_double_p): Ditto.
336         (pdp11_register_move_cost): Adjust for new register classes.
337         (pdp11_regno_reg_class): Ditto.
338         (expand_block_move): Delete.
339         (pushpop_regeq): New function.
340         (pdp11_legitimate_address_p): Bugfix in check for constant
341         offset.
342         (pdp11_sp_frame_offset): Delete.
343         (pdp11_reg_save_size): New helper function for new frame layout.
344         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
345         (pdp11_expand_shift): Bugfix in check for constant shift count.
346         (pdp11_shift_length): Ditto.
347         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
348         (pdp11_cmp_length): New function.
349         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
350         some compile options.
351         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
352         (CALL_USED_REGISTERS): Ditto.
353         (ELIMINABLE_REGS): Ditto.
354         (REGISTER_NAMES): Ditto.
355         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
356         constraints.
357         (REG_CLASS_NAMES): Ditto.
358         (REG_CLASS_CONTENTS): Ditto.  Also remove
359         HARD_FRAME_POINTER_REGNUM.
360         (CPU_REG_CLASS): New macro.
361         (CLASS_MAX_NREGS): Adjust for new register classes.
362         (FUNCTION_PROFILER): Make no-op.
363         (may_call_alloca): Remove unused declaration.
364         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
365         (ASM_OUTPUT_SKIP): Fix format.
366         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
367         (HARD_FRAME_POINTER_REGNUM): Remove.
368         (return): Delete.
369         (*rts): Rename.  Remove epilogue related checks.
370         (cmpsi, cmpdi): New insn.
371         (cbranch<mode>4): Change to apply to SI and DI modes as well.
372         (mov<mode>): Change constraints to enforce that push/pop
373         destination cannot use the same register as source.
374         (*mov<mode><cc_cc>): Ditto.
375         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
376         at assembly output rather than as RTL expander.
377         (zero_extendqihi2): Bugfix in check for same registers.
378         (adddi3_nocc): Bugfix in check for constant operand.
379         (addsi3_nocc): Ditto.
380         (subdi3_nocc): Ditto.
381         (subsi3_nocc): Ditto.
382         (negdi2_nocc): Copy input to pdp11_expand_operands.
383         (negsi2_nocc): Ditto.
384         (bswap2_nocc): Ditto.
385         * config/pdp11/pdp11.opt (mlra): Fix documentation.
386         * config/pdp11/t-pdp11: Use -Os.
388 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
390         * config/arm/parsecpu.awk (/alias/): New parsing rule.
391         (/begin cpu/): Check that the cpu name hasn't been previously defined.
392         (gen_comm_data): Print out CPU alias tables.
393         (check_cpu): Match aliases when checking the CPU name.
394         * config/arm/arm-protos.h (cpu_alias): New structure.
395         (cpu_option): Add entry for aliases.
396         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
397         strongarm1100 and strongarm1110.
398         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
399         (config/arm/arm-generic.md): Remove redundant references to
400         strongarm110, strongarm1100 and strongarm1110.
401         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
402         Scan aliases for additional hints.
403         (arm_parse_cpu_option_name): Also match a cpu name against the list
404         of aliases.
405         * config/arm/arm-tables.opt: Regenerated.
406         * config/arm/arm-tune.md: Regenerated.
408 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
410         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
411         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
412         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
413         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
414         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
415         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
416         * gengtype.c (open_base_files): Add omp-general.h.
417         * gimple.c (gimple_build_omp_critical):
418         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
419         gimple_omp_taskgroup_set_clauses.
420         (gimple_build_omp_atomic_load): Add mo argument, call
421         gimple_omp_atomic_set_memory_order.
422         (gimple_build_omp_atomic_store): Likewise.
423         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
424         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
425         instead of GSS_OMP.
426         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
427         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
428         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
429         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
430         different value for GF_OMP_ATOMIC_NEED_VALUE.
431         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
432         comments.
433         (struct gimple_statement_omp_single_layout): And remove here.
434         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
435         than gimple_statement_omp_single_layout.
436         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
437         GIMPLE_OMP_TEAMS.
438         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
439         (gimple_omp_subcode): Formatting fix.
440         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
441         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
442         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
443         gimple_omp_teams_host, gimple_omp_teams_set_host,
444         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
445         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
446         gimple_omp_taskgroup_set_clauses): New inline functions.
447         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
448         (gimple_build_omp_atomic_store): Likewise.
449         (gimple_omp_atomic_seq_cst_p): Remove.
450         (gimple_omp_atomic_memory_order): New function.
451         (gimple_omp_atomic_set_seq_cst): Remove.
452         (gimple_omp_atomic_set_memory_order): New function.
453         (gimple_build_omp_taskgroup): Add clauses argument.
454         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
455         (dump_gimple_omp_task): Print taskwait with depend clauses.
456         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
457         dump_omp_atomic_memory_order.
458         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
459         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
460         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
461         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
462         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
463         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
464         ORT_UNTIED_TASKLOOP enumerators.
465         (enum gimplify_defaultmap_kind): New.
466         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
467         target_map_pointers_as_0len_arrays members, add defaultmap.
468         (new_omp_context): Initialize defaultmap member.
469         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
470         (maybe_fold_stmt): Don't fold even in host teams regions.
471         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
472         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
473         ctx->omp_firstprivatize_variable.
474         (omp_add_variable): Don't add private/firstprivate for VLAs in
475         ORT_TASKGROUP.
476         (omp_default_clause): Print "taskloop" rather than "task" if
477         ORT_*TASKLOOP.
478         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
479         Handle new defaultmap clause kinds.
480         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
481         iterator to be lastprivate or private.  Fix up diagnostics if linear
482         is used on collapse>1 simd iterator.
483         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
484         (gimplify_omp_depend): New function.
485         (gimplify_scan_omp_clauses): Add shared clause on parallel for
486         combined parallel master taskloop{, simd} if taskloop has
487         firstprivate, lastprivate or reduction clause.  Handle
488         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
489         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
490         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
491         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
492         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
493         lastprivate.
494         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
495         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
496         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
497         OMP_CLAUSE_{TASK,IN}_REDUCTION.
498         (gimplify_omp_task): Handle taskwait with depend clauses.
499         (gimplify_omp_for): Add shared clause on parallel for combined
500         parallel master taskloop{, simd} if taskloop has firstprivate,
501         lastprivate or reduction clause.  Use ORT_TASKLOOP or
502         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
503         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
504         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
505         __for_end and __for_range temporaries on OMP_PARALLEL for
506         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
507         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
508         sandwiched in between two taskloops.
509         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
510         instead of ctx->omp_firstprivatize_variable.
511         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
512         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
513         host teams, use gimplify_and_return_first etc. for body like
514         for target or target data constructs, and at the end call
515         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
516         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
517         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
518         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
519         gimple_omp_atomic_set_seq_cst calls.
520         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
521         case, handle taskgroup clauses.
522         * lto-streamer-out.c (hash_tree): Handle
523         OMP_CLAUSE_{TASK,IN}_REDUCTION.
524         * Makefile.in (GTFILES): Add omp-general.h.
525         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
526         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
527         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
528         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
529         BUILT_IN_GOMP_LOOP_DOACROSS_START,
530         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
531         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
532         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
533         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
534         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
535         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
536         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
537         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
538         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
539         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
540         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
541         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
542         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
543         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
544         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
545         * omp-expand.c (workshare_safe_to_combine_p): Return false for
546         non-worksharing loops.
547         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
548         (determine_parallel_type): Don't combine parallel with worksharing
549         which has _reductemp_ clause.
550         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
551         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
552         if there is nonmonotonic modifier or if there is no modifier and no
553         ordered clause.  For dynamic and guided schedule without monotonic
554         and nonmonotonic modifier, default to nonmonotonic.
555         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
556         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
557         task reductions.
558         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
559         there are any reduction clauses.
560         (expand_taskwait_call): New function.
561         (expand_teams_call): New function.
562         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
563         expand_teams_call for it.  Formatting fix.  Handle taskwait with
564         depend clauses.
565         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
566         of worksharing loops with task reductions.
567         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
568         expansion of worksharing loops with task reductions.
569         (expand_omp_sections): Handle expansion of sections with task
570         reductions.
571         (expand_omp_synch): For host teams call expand_omp_taskreg.
572         (omp_memory_order_to_memmodel): New function.
573         (expand_omp_atomic_load, expand_omp_atomic_store,
574         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
575         instead of gimple_omp_atomic_seq_cst_p.
576         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
577         depend clauses as a standalone directive.
578         * omp-general.c (enum omp_requires): New variable.
579         (omp_extract_for_data): Initialize have_reductemp member.  Allow
580         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
581         GT_EXPR loops depending on incr sign.  Formatting fixes.
582         * omp-general.h (struct omp_for_data): Add have_reductemp member.
583         (enum omp_requires): New enum.
584         (omp_requires_mask): Declare.
585         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
586         Fix comment typos.
587         * omp-low.c (struct omp_context): Add task_reductions and
588         task_reduction_map fields.
589         (is_host_teams_ctx): New function.
590         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
591         (use_pointer_for_field): Use is_global_var instead of
592         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
593         in outer contexts.
594         (build_outer_var_ref): Ignore taskgroup outer contexts.
595         (delete_omp_context): Release task_reductions and task_reduction_map.
596         (scan_sharing_clauses): Don't add any fields for reduction clause on
597         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
598         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
599         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
600         Handle OMP_CLAUSE_NONTEMPORAL.
601         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
602         needed.
603         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
604         clauses with task modifier.
605         (scan_omp_task): Handle taskwait with depend clauses.
606         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
607         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
608         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
609         (scan_omp_for): Fix comment formatting.
610         (scan_omp_teams): Handle host teams constructs.
611         (check_omp_nesting_restrictions): Allow teams with no outer
612         OpenMP context.  Adjust diagnostics for teams strictly nested into
613         some explicit OpenMP construct other than target.  Allow OpenMP atomics
614         inside of simd regions.
615         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
616         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
617         taskreg_nesting_level while scanning host teams construct.
618         (task_reduction_read): New function.
619         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
620         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
621         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
622         second argument create_tmp_var if it is NULL.  Don't ignore shared
623         clauses in is_host_teams_ctx contexts.  Handle
624         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
625         clauses.
626         (lower_reduction_clauses): Ignore reduction clauses with task
627         modifier.  Remove second argument create_tmp_var if it is NULL.
628         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
629         (lower_send_clauses): Ignore reduction clauses with task modifier.
630         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
631         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
632         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
633         rely that it is the last stmt in body so far.  Ignore outer taskgroup
634         contexts.
635         (omp_task_reductions_find_first, omp_task_reduction_iterate,
636         lower_omp_task_reductions): New functions.
637         (lower_omp_sections): Handle reduction clauses with taskgroup
638         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
639         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
640         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
641         modifiers.
642         (lower_omp_taskgroup): Handle taskgroup reductions.
643         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
644         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
645         (lower_depend_clauses): If there are any
646         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
647         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
648         seen, assume lowering is done already and return early.  Set kind
649         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
650         (lower_omp_taskreg): Handle reduction clauses with task modifier on
651         parallel construct.  Handle reduction clause on taskloop construct.
652         Handle taskwait with depend clauses.
653         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
654         for host teams constructs.
655         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
656         nontemporal and _reductemp_ clause entries.
657         (omp_clause_code_name): Likewise.
658         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
659         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
660         * tree-core.h (enum omp_clause_code): Add
661         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
662         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
663         (struct tree_base): Add omp_atomic_memory_order field into union.
664         Remove OMP_ATOMIC_SEQ_CST comment.
665         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
666         and OMP_CLAUSE_DEPEND_DEPOBJ.
667         (struct tree_omp_clause): Add subcode.defaultmap_kind.
668         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
669         OpenMP constructs with body and clauses operands.
670         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
671         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
672         (OMP_TASKGROUP_CLAUSES): Define.
673         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
674         OMP_CLAUSE__LOOPTEMP_.
675         (OMP_ATOMIC_SEQ_CST): Remove.
676         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
677         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
678         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
679         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
680         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
681         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
682         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
683         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
684         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
685         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
686         Define.
687         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
688         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
689         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
690         (convert_local_omp_clauses): Likewise.  Remove useless test.
691         * tree-parloops.c (create_call_for_reduction_1): Pass
692         OMP_MEMORY_ORDER_RELAXED as new argument to
693         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
694         * tree-pretty-print.c (dump_omp_iterators): New function.
695         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
696         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
697         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
698         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
699         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
700         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
701         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
702         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
703         (dump_omp_atomic_memory_order): New function.
704         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
705         taskwait with depend clauses.
706         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
707         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
708         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
709         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
710         write_ts_omp_clause_tree_pointers): Likewise.
712 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
714         PR ipa/86395
715         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
716         "-missed", "-note", and "-all" sub-options.
717         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
718         API.
719         (can_early_inline_edge_p): Likewise.
720         (want_early_inline_function_p): Likewise.
721         (want_inline_self_recursive_call_p): Likewise.
722         (recursive_inlining): Likewise.
723         (inline_small_functions): Likewise.
724         (flatten_function): Likewise.
725         (ipa_inline): Likewise.
726         (inline_always_inline_functions): Likewise.
727         (early_inline_small_functions): Likewise.
728         (early_inliner): Likewise.
729         * tree-inline.c (expand_call_inline): Likewise.
731 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
733         * pretty-print.c (pp_format): Handle %f.
734         (selftest::test_pp_format): Add test of %f.
735         * pretty-print.h (pp_double): New macro.
737 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
739         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
740         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
741         (dump_pretty_printer::decode_format): Implement "%C" for
742         cgraph_node *.
743         (selftest::test_capture_of_dump_calls): Rename "where" to
744         "stmt_loc".  Convert test_decl to a function decl and set its
745         location.  Add a symbol_table_test RAII instance and a
746         cgraph_node, using it to test "%C" and dump_symtab_node.
748 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
750         PR middle-end/87916
751         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
753 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
755         * cgraph.c: Include "selftest.h".
756         (saved_symtab): New variable.
757         (selftest::symbol_table_test::symbol_table_test): New ctor.
758         (selftest::symbol_table_test::~symbol_table_test): New dtor.
759         (selftest::test_symbol_table_test): New test.
760         (selftest::cgraph_c_tests): New.
761         * cgraph.h (saved_symtab): New decl.
762         (selftest::symbol_table_test): New class.
763         * selftest-run-tests.c (selftest::run_tests): Call
764         selftest::cgraph_c_tests.
765         * selftest.h (selftest::cgraph_c_tests): New decl.
767 2018-11-08  Richard Biener  <rguenther@suse.de>
769         * tree-data-ref.h (lambda_int): New typedef.
770         (lambda_vector_gcd): Adjust.
771         (lambda_vector_new): Likewise.
772         (lambda_matrix_new): Likewise.
773         * tree-data-ref.c  (print_lambda_vector): Adjust.
775 2018-11-08  Richard Biener  <rguenther@suse.de>
777         PR tree-optimization/87929
778         * tree-complex.c (expand_complex_comparison): Clean EH.
780 2018-11-08  Martin Liska  <mliska@suse.cz>
782         * doc/extend.texi: Reword.
783         * predict.c (expr_expected_value_1): Likewise.
785 2018-11-08  Richard Biener  <rguenther@suse.de>
787         PR tree-optimization/87913
788         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
789         of extreme values to ordered comparisons.
791 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
793         PR middle-end/42726
794         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
796 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
798         * doc/invoke.texi: Remove leading dash from @opindex entries
799         throughout the file.
801 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
803         PR driver/80828
804         * doc/invoke.texi (Option Summary): Add -e and --entry.
805         (Link Options): Likewise.
807 2018-11-07  Nathan Sidwell  <nathan@acm.org>
809         PR 87926
810         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
811         --disable-checking bootstrap.
813 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
815         * configure: Regenerated.
817 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
819         PR c/87691
820         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
821         to the mode of the widest field iff the widest field has mode class
822         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
823         reference.
825 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
827         * common.opt: Add -Wattribute-warning.
828         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
829         * expr.c (expand_expr_real_1): Add new attribute to warning_at
830         call to allow user configure behavior of "warning" attribute.
832 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
834         * target.def: Put @: after every vs., e.g., and i.e. where it is
835         followed by whitespace.
836         * doc/extend.texi: Ditto.
837         * doc/fragments.texi: Ditto.
838         * doc/gimple.texi: Ditto.
839         * doc/implement-c.texi: Ditto.
840         * doc/install.texi: Ditto.
841         * doc/invoke.texi: Ditto.
842         * doc/md.texi: Ditto.
843         * doc/plugins.texi: Ditto.
844         * doc/rtl.texi: Ditto.
845         * doc/sourcebuild.texi: Ditto.
846         * doc/tm.texi.in: Ditto.
847         * doc/ux.texi: Ditto.
848         * doc/tm.texi: Regenerate.
850 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
852         * config/arm/arm-cpus.in (ares): New entry.
853         * config/arm/arm-tables.opt: Regenerate.
854         * config/arm/arm-tune.md: Likewise.
855         * doc/invoke.texi (ARM Options): Document ares.
857 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
859         * config/aarch64/aarch64-cores.def (ares): Define.
860         * config/aarch64/aarch64-tune.md: Regenerate.
861         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
863 2018-11-07  Jan Hubicka  <jh@suse.cz>
865         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
866         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
867         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
868         TYPE_DECL.
870 2018-11-07  Richard Biener  <rguenther@suse.de>
872         PR tree-optimization/87914
873         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
874         of nested cycles.
875         (vectorizable_reduction): Handle shifts and rotates by dispatching
876         to vectorizable_shift.
877         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
878         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
879         def cases.
880         (vectorizable_shift): Export and handle being called as
881         vect_nested_cycle.
882         (vect_analyze_stmt): Call vectorizable_shift after
883         vectorizable_reduction.
884         * tree-vectorizer.h (vectorizable_shift): Declare.
886 2018-11-07  Jan Hubicka  <jh@suse.cz>
888         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
889         than const decls in TREE_VALUE of enum.
890         (dump_type_inheritance_graph): Improve duplicate dumping.
891         (free_enum_values): New.
892         (build_type_inheritance_graph): Use it.
893         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
894         which are not main variants or not ODR types.
895         (verify_type_variant): Expect variants to have no TYPE_VALUES.
897 2018-11-07  Richard Biener  <rguenther@suse.de>
899         * ipa-inline.c (want_inline_small_function_p): Compute
900         big_speedup_p lazily and last.
902 2018-11-07  Jan Hubicka  <jh@suse.cz>
904         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
905         building incomplete variant of complete type.
906         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
907         variant of complete type.
909 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
911         * config/mips/mips.c: Fix typo in documentation of
912         mips_loongson_ext2_prefetch_cookie.
913         (mips_option_override): fix brain twister logical.
914         * config/mips/mips.h: Fix typo in documentation of
915         ISA_HAS_CTZ_CTO and define pattern.
916         * config/mips/mips.md (prefetch): Hoist EXT2 above
917         the 2EF/EXT block.
918         (prefetch_indexed): Hoist EXT2 above the EXT block.
920 2018-11-07  Jan Hubicka  <jh@suse.cz>
922         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
923         return and parameter types of function and method types.
924         (free_lang_data_in_cgraph): Update.
926 2018-11-07  Martin Liska  <mliska@suse.cz>
928         PR rtl-optimization/87868
929         * postreload-gcse.c (eliminate_partially_redundant_load): Set
930         threshold to max_count if we would overflow.
931         * profile-count.h: Make max_count a public constant.
933 2018-11-07  Martin Liska  <mliska@suse.cz>
935         * mem-stats.h: Fix GNU coding style.
937 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
939         * config/mips/gs264e.md: New.
940         * config/mips/mips-cpus.def: Define gs264e.
941         * config/mips/mips-tables.opt: Regenerate.
942         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
943         gs264e.
944         (mips_issue_rate): Add support for gs264e.
945         (mips_multipass_dfa_lookahead): Likewise.
946         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
947         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
948         (MIPS_ASE_MSA_SPEC): New.
949         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
950         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
951         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
952         * config/mips/mips.md: Include gs264e.md.
953         (processor): Add gs264e.
954         * config/mips/mips.opt (MSA): Use Mask instead of Var.
955         * doc/invoke.texi: Add gs264e to supported architectures.
957 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
959         * config/mips/gs464e.md: New.
960         * config/mips/mips-cpus.def: Define gs464e.
961         * config/mips/mips-tables.opt: Regenerate.
962         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
963         gs464e.
964         (mips_issue_rate): Add support for gs464e.
965         (mips_multipass_dfa_lookahead): Likewise.
966         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
967         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
968         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
969         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
970         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
971         * config/mips/mips.md: Include gs464e.md.
972         (processor): Add gs464e.
973         * doc/invoke.texi: Add gs464e to supported architectures.
975 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
977         * config/mips/loongson3a.md: Rename to ...
978         * config/mips/gs464.md: ... here.
979         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
980         as an alias of gs464 processor.
981         * config/mips/mips-tables.opt: Regenerate.
982         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
983         instead of PROCESSOR_LOONGSON_3A.
984         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
985         TUNE_LOONGSON_3A.
986         (mips_option_override): Enable MMI and EXT for gs464.
987         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
988         Rename TUNE_LOONGSON_3A to TUNE_GS464.
989         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
990         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
991         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
992         TARGET_LOONGSON_3A.
993         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
994         (processor): Add gs464;
995         * doc/invoke.texi: Add gs464 to supported architectures.
997 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
999         * config/mips/mips-protos.h
1000         (mips_loongson_ext2_prefetch_cookie): New prototype.
1001         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1002         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1003         TARGET_LOONGSON_EXT2 is true.
1004         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1005         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1006         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1007         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1008         TARGET_LOONGSON_EXT2.
1009         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1010         (define_insn "ctz<mode>2"): New insn pattern.
1011         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1012         (define_insn "prefetch_indexed_<mode>"): Include
1013         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1014         * config/mips/mips.opt (-mloongson-ext2): Add option.
1015         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1017 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1019         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1020         __mips_loongson_ext.
1021         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1022         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1023         -mloongson-ext.
1024         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1025         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1026         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1027         instead of TARGET_LOONGSON_3A.
1028         * config/mips/mips.opt (-mloongson-ext): Add option.
1029         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1031 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1033         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1034         * config/mips/loongson.md: Move to ...
1035         * config/mips/loongson-mmi.md: here; Adjustment.
1036         * config/mips/loongson.h: Move to ...
1037         State as deprecated. Include loongson-mmiintrin.h for back
1038         compatibility and warning.
1039         * config/mips/loongson-mmiintrin.h: ... here.
1040         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1041         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1042         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1043         (mips_option_override): Make sure MMI use hard float;
1044         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1045         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1046         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1047         TARGET_LOONGSON_VECTORS.
1048         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1049         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1050         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1051         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1052         -mloongson-mmi.
1053         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1054         TARGET_LOONGSON_VECTORS.
1055         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1056         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1057         (Loongson MMI patterns): Include loongson-mmi.md instead of
1058         loongson.md.
1059         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1060         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1062 2018-11-07  Richard Biener  <rguenther@suse.de>
1064         PR lto/87906
1065         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1066         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1068 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
1070         PR rtl-optimization/87874
1071         * lra.c (lra_substitute_pseudo): Do not create a subreg for
1072         const wide ints.
1074 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
1076         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1077         if not in indexed or indirect form.
1078         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1079         (bswap<mode>2_store): Ditto.
1081 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
1083         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1084         the UNSPEC.
1086 2018-11-06  Richard Biener  <rguenther@suse.de>
1088         PR tree-optimization/86850
1089         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1090         instead of src.m_vec.
1092 2018-11-06  Jan Hubicka  <jh@suse.cz>
1094         * tree.c (fld_simplified_type_name): Break out form ...
1095         (free_lang_data_in_type): ... here.
1096         (fld_type_variant_equal_p): Use it.
1098 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1100         * config/default-d.c: Include memmodel.h.
1102         * config/sol2-d.c: New file.
1103         * config/t-sol2 (sol2-d.o): New rule.
1104         * config.gcc <*-*-solaris2*>: Set d_target_objs,
1105         target_has_targetdm.
1107 2018-11-06  Jan Hubicka  <jh@suse.cz>
1109         * tree.c (fld_type_variant): Also copy alignment; be sure that
1110         new variant is equal.
1112 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
1114         PR target/87762
1115         * config/s390/s390.md: Add relative_long attribute.
1117 2018-11-06  Jan Hubicka  <jh@suse.cz>
1119         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1120         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1121         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1122         stream TYPE_NEEDS_CONSTRUCTING.
1123         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1124         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1126 2018-11-06  Richard Biener  <rguenther@suse.de>
1128         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1129         dump-scope ...
1130         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1132 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1134         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1135         REG_CLASS_CONTENTS[GEN_REGS].
1136         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1138 2018-11-06  Jan Hubicka  <jh@suse.cz>
1140          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1142 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1144         PR sanitizer/80953
1145         * config/sol2.h (ASAN_CC1_SPEC): Define.
1146         (LD_WHOLE_ARCHIVE_OPTION): Define.
1147         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1148         (ASAN_REJECT_SPEC): Provide default.
1149         (LIBASAN_EARLY_SPEC): Define.
1150         (LIBTSAN_EARLY_SPEC): Define.
1151         (LIBLSAN_EARLY_SPEC): Define.
1152         * config/i386/sol2.h (CC1_SPEC): Redefine.
1153         (ASAN_REJECT_SPEC): Define.
1155         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1156         (TARGET_ASAN_SHADOW_OFFSET): Define.
1157         (sparc_asan_shadow_offset): New function.
1158         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1159         (ASAN_REJECT_SPEC): Define.
1161 2018-11-06  Jan Hubicka  <jh@suse.cz>
1163         * tree.c (fld_type_variant): Copy canonical type.
1164         (fld_incomplete_type_of): Check that canonical types looks sane;
1165         copy canonical type.
1166         (verify_type): Accept when incomplete type has complete canonical type.
1168 2018-11-06  Jan Hubicka  <jh@suse.cz>
1170         * tree.c (free_lang_data): Reset overwite_assembler_name,
1171         print_xnode, print_decl, print_type and print_identifier of
1172         langhooks.
1174 2018-11-06  Richard Biener  <rguenther@suse.de>
1176         PR tree-optimization/87889
1177         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1178         Do nothing if old and new arg are the same
1180 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
1182         PR target/87723
1183         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1184         attributes for operands 3 and 4.
1186 2018-11-06  Richard Biener  <rguenther@suse.de>
1188         PR middle-end/18041
1189         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1190         matching bitfield insertion.
1192 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
1194         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
1195         (reg_next_debug_use): New.
1196         (attempt_change): Propagate adjusted expression into affected
1197         debug insns.
1198         (merge_in_block): Track uses in debug insns.
1199         (pass_inc_dec::execute): Allocate and release
1200         reg_next_debug_use.
1202 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1204         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1205         (_mm512_fixupimm_round_pd): Update parameters and builtin.
1206         (_mm512_maskz_fixupimm_round_pd): Ditto.
1207         (_mm512_fixupimm_round_ps): Ditto.
1208         (_mm512_maskz_fixupimm_round_ps): Ditto.
1209         (_mm_fixupimm_round_sd): Ditto.
1210         (_mm_maskz_fixupimm_round_sd): Ditto.
1211         (_mm_fixupimm_round_ss): Ditto.
1212         (_mm_maskz_fixupimm_round_ss): Ditto.
1213         (_mm512_fixupimm_pd): Ditto.
1214         (_mm512_maskz_fixupimm_pd): Ditto.
1215         (_mm512_fixupimm_ps): Ditto.
1216         (_mm512_maskz_fixupimm_ps): Ditto.
1217         (_mm_fixupimm_sd): Ditto.
1218         (_mm_maskz_fixupimm_sd): Ditto.
1219         (_mm_fixupimm_ss): Ditto.
1220         (_mm_maskz_fixupimm_ss): Ditto.
1221         (_mm512_mask_fixupimm_round_pd): Update builtin.
1222         (_mm512_mask_fixupimm_round_ps): Ditto.
1223         (_mm_mask_fixupimm_round_sd): Ditto.
1224         (_mm_mask_fixupimm_round_ss): Ditto.
1225         (_mm512_mask_fixupimm_pd): Ditto.
1226         (_mm512_mask_fixupimm_ps): Ditto.
1227         (_mm_mask_fixupimm_sd): Ditto.
1228         (_mm_mask_fixupimm_ss): Ditto.
1229         * config/i386/avx512vlintrin.h:
1230         (_mm256_fixupimm_pd): Update parameters and builtin.
1231         (_mm256_maskz_fixupimm_pd): Ditto.
1232         (_mm256_fixupimm_ps): Ditto.
1233         (_mm256_maskz_fixupimm_ps): Ditto.
1234         (_mm_fixupimm_pd): Ditto.
1235         (_mm_maskz_fixupimm_pd): Ditto.
1236         (_mm_fixupimm_ps): Ditto.
1237         (_mm_maskz_fixupimm_ps): Ditto.
1238         (_mm256_mask_fixupimm_pd): Update builtin.
1239         (_mm256_mask_fixupimm_ps): Ditto.
1240         (_mm_mask_fixupimm_pd): Ditto.
1241         (_mm_mask_fixupimm_ps): Ditto.
1242         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1243         * config/i386/i386-builtin.def: Update builtin definitions.
1244         * config/i386/i386.c: Handle new builtin types and remove useless ones.
1245         * config/i386/sse.md: Update VFIXUPIMM* patterns.
1246         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1247         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1248         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1249         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1250         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1251         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1252         * config/i386/subst.md:
1253         (round_saeonly_sd_mask_operand4): Add new subst_attr.
1254         (round_saeonly_sd_mask_op4): Ditto.
1255         (round_saeonly_expand_operand5): Ditto.
1256         (round_saeonly_expand): Update.
1258 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
1260         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1262 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
1264         PR rtl-optimization/87871
1265         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1267 2018-11-05  Paul Koning  <ni1d@arrl.net>
1269         * doc/sourcebuild.texi (target attributes): Document new "inf"
1270         effective target keyword.
1272 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1274         * config/s390/s390.c (s390_register_move_cost): Increase costs for
1275         moves involving the CC reg.
1277 2018-11-05  Richard Biener  <rguenther@suse.de>
1279         PR tree-optimization/87873
1280         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1281         argument.
1282         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1283         * tree-vect-loop.c (vect_transform_loop): When splitting the
1284         loop exit also create forwarder PHIs for constants.
1285         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1286         Handle constant to_arg, add extra checking we match up the correct
1287         PHIs.
1289 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1291         * config/s390/s390.md: QImode and HImode for load on condition.
1293 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1295         * config/s390/predicates.md: Fix typo.
1296         * config/s390/s390.md: Allow immediates for load on condition.
1298 2018-11-05  Martin Liska  <mliska@suse.cz>
1300         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1301         * bitmap.h (struct bitmap_usage): Likewise.
1302         * ggc-common.c (SCALE): Remove.
1303         (LABEL): Likewise.
1304         (struct ggc_usage): Use SIZE_AMOUNT. And update
1305         compare method.
1306         * ggc-page.c (SCALE): Remove.
1307         (STAT_LABEL): Remove.
1308         (ggc_print_statistics): Use SIZE_AMOUNT.
1309         * gimple.h (SCALE): Remove.
1310         (LABEL): Likewise.
1311         * input.c (ONE_K): Remove.
1312         (ONE_M): Likewise.
1313         (SCALE): Likewise.
1314         (STAT_LABEL): Likewise.
1315         (FORMAT_AMOUNT): Likewise.
1316         (dump_line_table_statistics): Use SIZE_AMOUNT.
1317         * mem-stats.h (struct mem_usage): Likewise.
1318         * rtl.c (dump_rtx_statistics): Likewise.
1319         (rtx_alloc_counts): Change type to size_t.
1320         (rtx_alloc_sizes): Likewise.
1321         (rtx_count_cmp): New.
1322         (dump_rtx_statistics): Sort first based on counts.
1323         * tree.c (tree_nodes_cmp): New.
1324         (tree_codes_cmp): New.
1325         (dump_tree_statistics): Sort first based on counts.
1326         * system.h (ONE_K): New.
1327         (ONE_M): Likewise.
1328         (SIZE_SCALE): Likewise.
1329         (SIZE_LABEL): Likewise.
1330         (SIZE_AMOUNT): Likewise.
1331         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
1332         * tree-dfa.c (dump_dfa_stats): Likewise.
1333         * tree-phinodes.c (phinodes_print_statistics): Likewise.
1334         * tree-ssanames.c (ssanames_print_statistics): Likewise.
1335         * tree.c (dump_tree_statistics): Likewise.
1336         * vec.c (struct vec_usage): Likewise.
1337         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
1338         trigger a -Werror=format-overflow with
1339         --enable-gather-detailed-stats.
1341 2018-11-05  Martin Liska  <mliska@suse.cz>
1343         * mem-stats.h (mem_alloc_description::release_instance_overhead):
1344         Return T *.
1345         * vec.c (struct vec_usage): Register m_element_size.
1346         (vec_prefix::register_overhead): New arguments: elements and
1347         element_size.
1348         (vec_prefix::release_overhead): Subtract elements.
1349         * vec.h (struct vec_prefix): Change signature.
1350         (va_heap::reserve): Pass proper arguments.
1351         (va_heap::release): Likewise.
1353 2018-11-05  Martin Liska  <mliska@suse.cz>
1355         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
1356         style.
1357         * vec.c: Likewise.
1359 2018-11-05  Richard Biener  <rguenther@suse.de>
1361         * tree-scalar-evolution.h (final_value_replacement_loop): Update
1362         prototype.
1363         * tree-scalar-evolution.c (final_value_replacement_loop): Return
1364         whether anything was done.
1365         (scev_const_prop): Remove constant propagation part, fold
1366         remains into ...
1367         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
1368         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
1369         conditionally.
1371 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
1373         PR tree-optimization/87859
1374         * gimple-ssa-store-merging.c (struct merged_store_group): Add
1375         only_constants and first_nonmergeable_order members.
1376         (merged_store_group::merged_store_group): Initialize them.
1377         (merged_store_group::do_merge): Clear only_constants member if
1378         adding something other than INTEGER_CST store.
1379         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1380         stores with order >= first_nonmergeable_order.  Use
1381         merged_store->only_constants instead of always recomputing it.
1382         Set merged_store->first_nonmergeable_order if we've skipped any
1383         stores.  Attempt to merge overlapping INTEGER_CST stores that
1384         we would otherwise skip.
1386         PR sanitizer/87837
1387         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
1389 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
1391         PR target/87853
1392         * config/i386/emmintrin.h (__v16qs): New to cope with option
1393         -funsigned-char.
1394         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
1395         (_mm_cmplt_epi8): Likewise.
1396         (_mm_cmpgt_epi8): Likewise.
1398 2018-11-05  Richard Biener  <rguenther@suse.de>
1400         PR rtl-optimization/87852
1401         * fwprop.c (use_killed_between): Only consider single-defs of the
1402         use whose definition statement dominates the use.
1404 2018-11-05  Martin Liska  <mliska@suse.cz>
1406         PR web/87829
1407         * doc/invoke.texi: Remove options that are
1408         not disabled with -Os.
1410 2018-11-05  Martin Liska  <mliska@suse.cz>
1412         PR c/87811
1413         * doc/extend.texi: Update constrain about the last argument
1414         of __builtin_expect_with_probability.
1416 2018-11-05  Martin Liska  <mliska@suse.cz>
1418         PR c/87811
1419         * predict.c (expr_expected_value_1): Verify
1420         that last argument is a real constants and emit
1421         error.
1423 2018-11-05  Martin Liska  <mliska@suse.cz>
1425         PR gcov-profile/77698
1426         * ipa-profile.c (ipa_profile): Adjust hotness threshold
1427         only in LTO mode.
1429 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1431         PR tree-optimization/86572
1432         * builtins.c (c_strlen): Handle negative offsets in a safe way.
1434 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1436         PR tree-optimization/87672
1437         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
1438         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
1440 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
1442         PR middle-end/58372
1443         * cfgexpand.c (pass_expand::execute): Move the call to
1444         finish_eh_generation in front of the call to expand_stack_alignment.
1446 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1448         * common/config/i386/i386-common.c (processor_alias_table): Add
1449         znver2 entry.
1450         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
1451         (case ${target}): Add znver2.
1452         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
1453         -march=native recognize znver2 processors.
1454         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
1455         * config/i386/i386.c (m_znver2): New definition.
1456         (m_ZNVER): New definition.
1457         (m_AMD_MULTIPLE): Includes m_znver2.
1458         (processor_cost_table): Add znver2 entry.
1459         (processor_target_table): Add znver2 entry.
1460         (get_builtin_code_for_version): Set priority for
1461         PROCESSOR_ZNVER2.
1462         (processor_model): Add M_AMDFAM17H_ZNVER2.
1463         (arch_names_table): Ditto.
1464         (ix86_reassociation_width): Include znver2.
1465         * config/i386/i386.h (TARGET_znver2): New definition.
1466         (struct ix86_size_cost): Add TARGET_ZNVER2.
1467         (enum processor_type): Add PROCESSOR_ZNVER2.
1468         * config/i386/i386.md (define_attr "cpu"): Add znver2.
1469         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
1470         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
1471         (ix86_adjust_cost): Add znver2.
1472         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
1473         * gcc/doc/extend.texi: Add details about znver2.
1474         * gcc/doc/invoke.texi: Add details about znver2.
1476 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
1478         PR target/87079
1480         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
1481         pattern.
1483 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1485         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
1486         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
1488 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
1490         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
1491         shrink-wrapping when -mtrack-speculation.
1493 2018-11-02  Richard Biener  <rguenther@suse.de>
1495         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
1496         (pop_cost_one_pair): Do not free pair.
1497         (pop_best_coalesce): Likewise.
1498         (create_coalesce_list): Initialize obstack.
1499         (delete_coalesce_list): Free obstack.
1500         (find_coalesce_pair): Obstack-allocate coalesce pairs.
1501         (add_cost_one_coalesce): Likewise.
1502         (struct live_track): Remove bitmap pointer indirections.
1503         (new_live_track): Adjust.
1504         (delete_live_track): Likewise.
1505         (live_track_remove_partition): Likewise.
1506         (live_track_add_partition): Likewise.
1507         (live_track_live_p): Likewise.
1508         (live_track_process_def): Likewise.
1509         (live_track_clear_base_vars): Likewise.
1511 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1513         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
1514         * configure: Regenerate.
1515         * config.in: Regenerate.
1516         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
1517         HAVE_AS_REGISTER_PSEUDO_OP guard.
1518         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
1519         !HAVE_AS_REGISTER_PSEUDO_OP support.
1520         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
1521         guard.
1523 2018-11-02  Richard Biener  <rguenther@suse.de>
1525         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
1526         with is_type_die.
1528 2018-11-02  Richard Biener  <rguenther@suse.de>
1530         PR tree-optimization/87776
1531         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
1532         executable when iterating but running into rpo-vn-max-loop-depth
1533         and not eliding the iteration.
1535 2018-11-30  Jan Hubicka  <jh@suse.cz>
1537         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
1538         TYPE_DECL.
1540 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
1542         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
1543         prototype.
1544         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
1545         Combine with rs6000_address_for_fpconvert.
1546         (rs6000_address_for_fpconvert): Combine with
1547         rs6000_force_indexed_or_indirect_mem.
1548         (rs6000_expand_vector_init): Change function call from
1549         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1550         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
1551         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1552         (floatsi<mode>2_lfiwax_mem): Ditto.
1553         (floatunssi<mode>2_lfiwzx): Ditto.
1554         (floatunssi<mode>2_lfiwzx_mem): Ditto.
1555         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
1556         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
1557         (fix_trunc<mode>si2_stfiwx): Ditto.
1558         (fixuns_trunc<mode>si2_stfiwx): Ditto.
1559         (float_<mode>si2_hw): Ditto.
1560         (floatuns_<mode>si2_hw): Ditto.
1561         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
1562         (vsx_splat_<mode>): Ditto.
1564 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
1566         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
1567         at end of assembler input text.
1568         * configure: Regenerate.
1570 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
1572         PR tree-optimization/87826
1573         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
1574         negative or larger or equal to type's precision.
1576 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
1578         * opts.c (default_options_table): Do not enable
1579         OPT_fdelayed_branch at -Og.
1580         * doc/invoke.texi (-fdelayed-branch): Document it.
1582 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1584         * optabs-libfuncs.c (build_libfunc_function_visibility):
1585         New, split out from...
1586         (build_libfunc_function): ... here.
1587         (init_one_libfunc_visibility): New, split out from ...
1588         (init_one_libfunc): ... here.
1590         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
1591         scratch register need not be early-clobber.  Document the reason
1592         why we cannot use ST<OP>.
1594 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
1596         PR bootstrap/82856
1597         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
1598         line for second argument of AC_DEFINE_UNQUOTED.
1599         * doc/install.texi (Tools/packages necessary for modifying GCC):
1600         Update to autoconf 2.69 and automake 1.15.1.
1601         * aclocal.m4, config.in, configure: Regenerate.
1603 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
1605         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
1606         initialization.
1608 2018-10-31  Martin Liska  <mliska@suse.cz>
1610         PR driver/83193
1611         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
1612         Add new argument invalid_extension.
1613         (aarch64_get_all_extension_candidates): New function.
1614         (aarch64_rewrite_selected_cpu): Add NULL to function call.
1615         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
1616         new argument.
1617         (aarch64_get_all_extension_candidates): New function.
1618         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
1619         argument invalid_extension.
1620         (aarch64_parse_cpu): Likewise.
1621         (aarch64_print_hint_for_extensions): New function.
1622         (aarch64_validate_mcpu): Provide hint about invalid extension.
1623         (aarch64_validate_march): Likewise.
1624         (aarch64_handle_attr_arch): Pass new argument.
1625         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
1626         (aarch64_handle_attr_isa_flags): Likewise.
1628 2018-10-31  Richard Biener  <rguenther@suse.de>
1630         PR middle-end/70359
1631         PR middle-end/86270
1632         * tree-outof-ssa.c (insert_backedge_copies): Restrict
1633         copy generation to useful cases.  Place the copy before
1634         the definition of the backedge value when possible.
1636 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1638         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
1639         * config/arc/arc.c (arc_active_insn): New function.
1640         (check_store_cacheline_hazard): Likewise.
1641         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
1642         (arc_override_options): Disable delay slot scheduler for older
1643         A7.
1644         (arc_store_addr_hazard_p): New implementation, old one renamed to
1645         ...
1646         (arc_store_addr_hazard_internal_p): Renamed.
1647         (arc_reorg): Don't combine into brcc instructions which are part
1648         of hardware hazard solution.
1649         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
1650         (tune_arc700): Likewise.
1651         * config/arc/arc.opt (arc7xx): New tune value.
1652         * config/arc/arc700.md: Improve A7 scheduler.
1654 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1656         * config/arc/arc.c (arc_override_options): Remove
1657         TARGET_COMPACT_CASESI.
1658         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
1659         (CASE_VECTOR_MODE): Likewise.
1660         (CASE_VECTOR_PC_RELATIVE): Likewise.
1661         (CASE_VECTOR_SHORTEN_MODE): Likewise.
1662         (CASE_VECTOR_SHORTEN_MODE1): Delete.
1663         (ADDR_VEC_ALIGN): Update.
1664         (ASM_OUTPUT_CASE_LABEL): Undefine.
1665         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
1666         (TARGET_BI_BIH): Define.
1667         (DEFAULT_BRANCH_INDEX): Likewise.
1668         * config/arc/arc.md (casesi): Rework to accept BI/BIH
1669         instructions, remove compact_casesi use case.
1670         (casesi_compact_jump): Remove.
1671         (casesi_dispatch): New pattern.
1672         * config/arc/arc.opt: Add mbranch-index option. Deprecate
1673         compact_casesi option.
1674         * doc/invoke.texi: Document mbranch-index option.
1676 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1678         * config/arc/arc.c (arc_get_tp): Remove function.
1679         (arc_emit_call_tls_get_addr): Likewise.
1680         (arc_call_tls_get_addr): New function.
1681         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
1682         * config/arc/arc.md (tls_load_tp_soft): Remove.
1683         (tls_gd_get_addr): Likewise.
1685 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1687         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
1688         (mulsi3_600_lib): Remove pattern.
1689         (umulsi3_highpart_600_lib_le): Likewise.
1690         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
1691         (umulsidi3): Remove call to umulsidi3_600_lib.
1692         (umulsidi3_600_lib): Remove pattern.
1693         (peephole2): Remove peephole using the above deprecated patterns.
1695 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
1697         PR target/87374
1698         * config/arm/arm.c (arm_option_check_internal): Disable the combined
1699         use of -mslow-flash-data and -mword-relocations.
1700         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
1701         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
1702         flag_pic.
1703         * doc/invoke.texi (-mword-relocations): Mention conflict with
1704         -mslow-flash-data.
1705         (-mslow-flash-data): Reciprocally.
1707 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1709         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
1710         16-byte modes held in GP registers to use an even regno.
1712         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
1713         (aarch64_atomic_ldop_supported_p): Remove.
1714         (aarch64_gen_atomic_ldop): Remove.
1715         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
1716         Fully expand LSE operations here.
1717         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
1718         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
1719         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
1720         and use ATOMIC_LDOP instead; use register_operand for the input;
1721         drop the split and emit insns directly.
1722         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
1723         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
1724         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
1726         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
1727         (aarch64_gen_atomic_ldop): Don't call it.
1728         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
1729         Use aarch64_reg_or_zero.
1730         (aarch64_atomic_exchange<ALLI>): Likewise.
1731         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
1732         operand 0; use aarch64_reg_or_zero for input; merge ...
1733         (@aarch64_atomic_swp<ALLI>): ... this and remove.
1735         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
1736         (aarch64_split_compare_and_swap): Use it.
1737         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
1738         test oldval against the proper predicate.
1739         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
1740         Use nonmemory_operand for expected.
1741         (cas_short_expected_pred): New.
1742         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
1743         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
1744         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
1745         (aarch64_plushi_operand): New.
1747         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
1748         Force oldval into the rval register for TARGET_LSE; emit the compare
1749         during initial expansion so that it may be deleted if unused.
1750         (aarch64_gen_atomic_cas): Remove.
1751         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
1752         Change =&r to +r for operand 0; use match_dup for operand 2;
1753         remove is_weak and mod_f operands as unused.  Drop the split
1754         and merge with...
1755         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
1756         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
1757         (@aarch64_atomic_cas<GPI>): Similarly.
1759 2018-10-31  Richard Biener  <rguenther@suse.de>
1761         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
1762         using ABSU_EXPR.
1764 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1766         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
1767         * config/aarch64/aarch64.md: Include saphira.md
1768         * config/aarch64/saphira.md: New file for pipeline description.
1770 2018-10-30  Martin Sebor  <msebor@redhat.com>
1772         PR middle-end/87041
1773         * gimple-ssa-sprintf.c (format_directive): Use %G to include
1774         inlining context.
1775         (sprintf_dom_walker::compute_format_length):
1776         Avoid setting POSUNDER4K here.
1777         (get_destination_size): Handle null argument values.
1778         (get_user_idx_format): New function.
1779         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
1780         functions, including user-defined with attribute format printf.
1781         Use %G to include inlining context.
1782         Set POSUNDER4K here.
1784 2018-10-30  Jan Hubicka  <jh@suse.cz>
1786         * params.def (lto-partitions): Bump from 32 to 128.
1788 2018-10-30  Jan Hubicka  <jh@suse.cz>
1790         * tree.c
1791         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
1792         head in file.
1793         (free_lang_data_in_type): Forward declare.
1794         (fld_type_variant_equal_p): New function.
1795         (fld_type_variant): New function
1796         (fld_incomplete_types): New hash.
1797         (fld_incomplete_type_of): New function
1798         (fld_simplfied-type): New function.
1799         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
1800         (free_lang_data): Allocate and free fld_incomplete_type; update call
1801         of free_lang_data_in_decl.
1803 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
1805         * gcov.c (output_lines): Remove duplicate line.
1807 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
1809         * config/rs6000/rs6000.md (bswapdi2): Force address into register
1810         if not in indexed or indirect form.
1811         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
1812         (bswapdi2_store): Ditto.
1813         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
1814         helper function.
1815         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
1816         Prototype for helper function.
1818 2018-10-30  Martin Sebor  <msebor@redhat.com>
1820         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
1821         (target, pragma GCC optimize, pragma GCC target): Ditto.
1823 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
1825         * doc/extend.texi: Fix prototype and description of
1826         __builtin_expect_with_probability.
1828 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1830         * cgraph.h (clone_function_name_1): Replaced by new
1831           clone_function_name_numbered that takes name as string; for
1832           privatize_symbol_name_1 use only.
1833           (clone_function_name): Renamed to
1834           clone_function_name_numbered to be explicit about numbering.
1835           (clone_function_name): New two-argument function that does
1836           not number its output.
1837           (clone_function_name): New three-argument function that
1838           takes a number to append to its output.
1839         * cgraphclones.c (duplicate_thunk_for_node):
1840           (clone_function_name_1): Renamed.
1841           (clone_function_name_numbered): Two new functions.
1842           (clone_function_name): Improved documentation.
1843           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
1844         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
1845         * final.c (final_scan_insn_1): Use the new clone_function_name
1846           without numbering.
1847         * multiple_target.c (create_dispatcher_calls): Ditto.
1848           (create_target_clone): Ditto.
1849         * omp-expand.c (grid_expand_target_grid_body): Ditto.
1850         * omp-low.c (create_omp_child_function_name): Ditto.
1851         * omp-simd-clone.c (simd_clone_create): Ditto.
1852         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
1853           new clone_function_name without numbering.
1855 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
1857         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
1858         Assert that the allocation size is not zero.
1860 2018-10-30  Richard Biener  <rguenther@suse.de>
1862         PR tree-optimization/87800
1863         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
1864         non-induction or reduction PHIs.
1866 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
1868         * config/aarch64/falkor-tag-collision-avoidance.c
1869         (execute_tag_collision_avoidance): Call df_note_add_problem.
1871 2018-10-30  Martin Liska  <mliska@suse.cz>
1873         * doc/extend.texi: Fix typo in documentation
1874         of __builtin_expect_with_probability.
1876 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
1878         PR c++/87721
1879         * input.c (get_substring_ranges_for_loc): Detect if
1880         linemap_resolve_location gives us a NULL map, and reject
1881         this case.
1883 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
1885         * config.gcc (xstormy16-*-elf): Set tm_d_file.
1887 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
1888             Martin Sebor  <msebor@redhat.com>
1889             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1891         * doc/ux.texi (Quoting): New subsection, adapted from material at
1892         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
1893         MartinSebor and ManuelLopezIbanez.
1894         (Fix-it hints): Note that fix-it hints shouldn't be marked for
1895         translation.
1897 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
1899         PR middle-end/87469
1900         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
1901         max value.
1903 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
1905         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
1907 2018-10-29  Paul Koning  <ni1d@arrl.net>
1909         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
1911 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
1913         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
1914         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
1915         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
1916         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
1917         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
1918         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
1919         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
1920         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
1921         Change 'vector' to '__vector'.
1922         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
1923         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
1924         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
1925         _mm_avg_pu16): Likewise.  And, whitespace corrections.
1927 2018-10-29  Richard Biener  <rguenther@suse.de>
1929         PR tree-optimization/87785
1930         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
1931         internal defs.
1933 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1935         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
1937 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1939         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
1940         ports configuration macro, defaults to "gnu".
1941         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
1942         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
1944 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1946         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
1948 2018-10-29  Richard Biener  <rguenther@suse.de>
1950         PR tree-optimization/87790
1951         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
1952         (vect_make_slp_decision): Adjust.
1953         (vect_slp_analyze_bb_1): Likewise.
1954         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
1955         edges.
1957 2018-10-29  Richard Biener  <rguenther@suse.de>
1959         PR tree-optimization/87785
1960         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
1961         and processing.
1962         (vect_build_slp_tree): Likewise.
1963         (vect_gather_slp_loads): New function.
1964         (vect_analyze_slp_instance): Gather loads separately from the
1965         SLP tree build.
1967 2018-10-29  Martin Liska  <mliska@suse.cz>
1969         * Makefile.in: Make dependency to json.o.
1970         * doc/gcov.texi: Document new JSON format, remove
1971         old intermediate format documentation.
1972         * gcov.c (struct function_info): Come up with m_name and
1973         m_demangled_name.
1974         (function_info::function_info): Initialize it.
1975         (function_info::~function_info): Release it.
1976         (main): Rename flag_intermediate_format to flag_json_format.
1977         (print_usage): Describe --json-format.
1978         (process_args): Set flag_json_format.
1979         (output_intermediate_line): Remove.
1980         (output_intermediate_json_line): Likewise.
1981         (get_gcov_intermediate_filename): Return new extension
1982         ".gcov.json.gz".
1983         (output_intermediate_file): Implement JSON emission.
1984         (output_json_intermediate_file): Implement JSON emission.
1985         (generate_results): Use ::get_name for function name.
1986         Handle JSON output file.
1987         (read_graph_file): Use ::get_name instead of cplus_demangle.
1988         (read_count_file): Likewise.
1989         (solve_flow_graph): Likewise.
1990         (add_line_counts): Likewise.
1991         (accumulate_line_counts): Use new flag_json_format.
1992         (output_function_details): Use ::get_name instead of cplus_demangle.
1993         (output_lines): Likewise.
1994         * json.cc (test_writing_literals): Add new tests.
1995         * json.h (class literal): Add new boolean constructor.
1997 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
1999         PR rtl-optimization/87701
2000         PR rtl-optimization/87780
2001         * combine.c (make_more_copies): Rewrite.
2003 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2005         * doc/generic.texi (ABSU_EXPR): Document.
2006         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2007         (absu(absu(X)) -> absu(X)): Likewise.
2008         (absu(-X) -> absu(X)): Likewise.
2009         (absu(X)  where X is nonnegative -> X): Likewise.
2011 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2013         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2014         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2015         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2016         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2017         (s-tm-texi): Also check timestamp on d-target.def.
2018         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2019         (build/genhooks.o): Also depend on D_TARGET_DEF.
2020         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2021         variables.
2022         * config/aarch64/aarch64-d.c: New file.
2023         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2024         Define.
2025         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2026         prototype.
2027         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2028         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2029         * config/arm/arm-d.c: New file.
2030         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2031         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2032         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2033         * config/arm/t-arm (arm-d.o): New rule.
2034         * config/default-d.c: New file.
2035         * config/glibc-d.c: New file.
2036         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2037         * config/i386/i386-d.c: New file.
2038         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2039         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2040         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2041         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2042         * config/i386/t-i386 (i386-d.o): New rule.
2043         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2044         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2045         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2046         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2047         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2048         * config/mips/mips-d.c: New file.
2049         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2050         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2051         * config/mips/t-mips (mips-d.o): New rule.
2052         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2053         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2054         * config/powerpcspe/powerpcspe-d.c: New file.
2055         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2056         New prototype.
2057         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2058         Support GNU D by using 0 as the language type.
2059         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2060         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2061         * config/riscv/riscv-d.c: New file.
2062         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2063         prototype.
2064         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2065         * config/riscv/t-riscv (riscv-d.o): New rule.
2066         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2067         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2068         * config/rs6000/rs6000-d.c: New file.
2069         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2070         prototype.
2071         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2072         Support GNU D by using 0 as the language type.
2073         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2074         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2075         * config/s390/s390-d.c: New file.
2076         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2077         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2078         * config/s390/t-s390 (s390-d.o): New rule.
2079         * config/sparc/sparc-d.c: New file.
2080         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2081         prototype.
2082         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2083         * config/sparc/t-sparc (sparc-d.o): New rule.
2084         * config/t-glibc (glibc-d.o): New rule.
2085         * configure: Regenerated.
2086         * configure.ac (tm_d_file): New variable.
2087         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2088         * doc/contrib.texi (Contributors): Add self for the D frontend.
2089         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2090         * doc/install.texi (Configuration): Mention libphobos as an option for
2091         --enable-shared.  Mention d as an option for --enable-languages.
2092         (Testing): Mention check-d as a target.
2093         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2094         name suffixes.  Mention d as a -x option.
2095         * doc/sourcebuild.texi (Top Level): Mention libphobos.
2096         * doc/standards.texi (Standards): Add section on D language.
2097         * doc/tm.texi: Regenerated.
2098         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2099         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2100         * dwarf2out.c (is_dlang): New function.
2101         (gen_compile_unit_die): Use DW_LANG_D for D.
2102         (declare_in_namespace): Return module die for D, instead of adding
2103         extra declarations into the namespace.
2104         (gen_namespace_die): Generate DW_TAG_module for D.
2105         (gen_decl_die): Handle CONST_DECLSs for D.
2106         (dwarf2out_decl): Likewise.
2107         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2108         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2109         kinds of aggregates.
2110         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2111         * genhooks.c: Include d/d-target.def.
2113 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
2115         PR target/85669
2116         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2117         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2118         (STACK_DYNAMIC_OFFSET): Likewise.
2120 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
2122         PR target/80024
2123         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2124         error message.
2126 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
2128         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2129         return type and other typos.
2131 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2133         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2134         a shorter sequence with fewer branches.
2135         (emit_final_str_compare_gpr): Ditto.
2137 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2139         * config/rs6000/tmmintrin.h: New file.
2140         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2142 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2144         * config/rs6000/mmintrin.h: Enable 32bit compilation.
2145         * config/rs6000/xmmintrin.h: Likewise.
2147 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
2149         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2151 2018-10-26  Richard Biener  <rguenther@suse.de>
2153         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2154         and wrapper.
2155         (vect_mark_slp_stmts_relevant): Likewise.
2156         (vect_detect_hybrid_slp_stmts): Likewise.
2157         (vect_bb_slp_scalar_cost): Likewise.
2158         (vect_remove_slp_scalar_calls): Likewise.
2160 2018-10-26  Jan Hubicka  <jh@suse.cz>
2162         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2163         (warn_types_mismatch): Fix walk of DECL_NAME.
2164         (odr_types_equivalent_p): Fix overactive assert.
2166 2018-10-26  Richard Biener  <rguenther@suse.de>
2168         PR tree-optimization/87105
2169         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2170         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2171         refcnt.
2172         (vect_create_new_slp_node): Initialize refcnt to one.
2173         (bst_traits): Move.
2174         (scalar_stmts_set_t, bst_fail): Remove.
2175         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2176         (vect_build_slp_tree): Add bst_map argument and lookup
2177         already created SLP nodes.
2178         (vect_print_slp_tree): Handle a SLP graph, print SLP node
2179         addresses.
2180         (vect_slp_rearrange_stmts): Handle a SLP graph.
2181         (vect_analyze_slp_instance): Adjust and free SLP nodes from
2182         the CSE map.  Fix indenting.
2183         (vect_schedule_slp_instance): Add short-cut.
2185 2018-10-26  Martin Liska  <mliska@suse.cz>
2187         PR testsuite/86158
2188         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2189         addr_expr and not with pointers.
2191 2018-10-26  Jan Hubicka  <jh@suse.cz>
2193         * tree.c (free_lang_data_in_type): Only check main variants.
2194         * ipa-devirt.c (warn_odr): Make static.
2195         (types_same_for_odr): Drop strict variant.
2196         (types_odr_comparable): Likewise.
2197         (odr_or_derived_type_p): Look for main variants.
2198         (odr_name_hasher::equal): Cleanup comment.
2199         (odr_subtypes_equivalent): Add warn and warned arguments; check main
2200         variants.
2201         (type_variants_equivalent_p): break out from ...
2202         (odr_types_equivalent): ... here; go for main variants where needed.
2203         (warn_odr): ... here; turn static.
2204         (warn_types_mismatch): Compare mangled names of main variants.
2205         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2206         (type_with_linkage_p): Sanity check that we look at main variant.
2207         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2208         * tree.h (types_same_for_odr): Drop strict argument.
2210 2018-10-26  Richard Biener  <rguenther@suse.de>
2212         PR tree-optimization/87746
2213         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2214         Simplify and fix WRT strided store groups with size not
2215         equal to step in element count.
2216         (vect_analyze_group_access_1): Dump the whole group.
2218 2018-10-25  Carl Love  <cel@us.ibm.com>
2220         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2221         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2222         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
2223         precicion entry for each overloaded builtin.
2224         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2225         VSCEDPUO): Rename overloaded name.
2226         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2227         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2228         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2229         define_expand for xscmpexqp instruction.
2230         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2232 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2233             Jinsong Ji <jji@us.ibm.com>
2235         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2236         function with vec_sl.
2237         (_mm_slli_epi32): Likewise.
2238         (_mm_slli_epi64): Likewise.
2239         (_mm_srai_epi16): Replace deprecated function with vec_sra.
2240         (_mm_srai_epi32): Likewise.
2241         (_mm_srli_epi16): Replace deprecated function with vec_sr.
2242         (_mm_srli_epi32): Likewise.
2243         (_mm_srli_epi64): Likewise.
2244         (_mm_sll_epi16): Replace deprecated function with vec_sl.
2245         (_mm_sll_epi32): Likewise.
2246         (_mm_sll_epi64): Likewise.
2247         (_mm_sra_epi16): Replace deprecated function with vec_sra.
2248         (_mm_sra_epi32): Likewise.
2249         (_mm_srl_epi16): Replace deprecated function with vec_sr.
2250         (_mm_srl_epi32): Likewise.
2251         (_mm_srl_epi64): Likewise.
2253 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2254             Jinsong Ji <jji@us.ibm.com>
2256         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2257         comparison operators with vec_cmp* for compatibility due to
2258         unfortunate history; clean up formatting and use types more
2259         appropriately.
2260         (_mm_sll_epi32): Likewise.
2261         (_mm_sll_epi64): Likewise.
2262         (_mm_srl_epi16): Likewise.
2263         (_mm_srl_epi32): Likewise.
2264         (_mm_srl_epi64): Likewise.
2266 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2267             Jinsong Ji <jji@us.ibm.com>
2269         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2270         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2271         __vector __bool int.  Use vec_cmpgt in preference to deprecated
2272         function vec_vcmpgtfp.
2273         (_mm_max_ps): Likewise.
2275 2018-10-25  Jeff Law  <law@redhat.com>
2277         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2278         if returning false.
2280 2018-10-25  Martin Sebor  <msebor@redhat.com>
2282         * doc/extend.texi (aligned): Expand attribute description.
2283         (Alignment): Rename section.  Discuss function arguments.
2285 2018-10-25  Jan Hubicka  <jh@suse.cz>
2287         * ipa-devirt.c (main_odr_variant): Remove.
2288         (hash_odr_name, types_same_for_odr, types_odr_comparable,
2289         odr_name_hasher::equal, odr_subtypes_equivalent_p):
2290         Drop use of main_odr_variant.
2291         (add_type_duplicate): Silence confused warnings on integer types.
2292         (get_odr_type): Always look for main variant.
2293         (register_odr_type): Simplify.
2295 2018-10-25  Richard Biener  <rguenther@suse.de>
2297         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2298         Initialize ng to silence error with release checking bootstrap.
2300 2018-10-25  Richard Biener  <rguenther@suse.de>
2302         * tree-if-conv.c: Include tree-ssa-sccvn.h.
2303         (tree_if_conversion): Run CSE on the if-converted loop body.
2305 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2307         * config/s390/constraints.md (ZL): New constraint.
2308         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2309         operands.
2310         * config/s390/s390.md (movdi_larl): Remove.
2311         (movdi_64): Add the LARL alternative.
2313 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2315         PR bootstrap/87747
2316         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
2317         (RTX_CODE_HWINT_P): New macro.
2318         (rtx_code_size): Use RTX_CODE_HWINT_P ().
2320 2018-10-25  Jan Hubicka  <jh@suse.cz>
2322         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
2323         is anonymous.
2325 2018-10-25  Richard Biener  <rguenther@suse.de>
2327         PR tree-optimization/87665
2328         PR tree-optimization/87745
2329         * tree-vectorizer.h (get_earlier_stmt): Remove.
2330         (get_later_stmt): Pick up UID from the original non-pattern stmt.
2332 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
2334         * options.texi (Deprecated): Move list to Var section.
2336 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
2337             Jinsong Ji <jji@us.ibm.com>
2339         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
2340         __vector long to __vector long long.
2341         (_mm_cvtpd_ps): Likewise.
2342         (_mm_cvttpd_epi32): Likewise.
2343         (_mm_cvtpi32_pd): Likewise.
2344         (_mm_unpackhi_epi64): Likewise.
2345         (_mm_unpacklo_epi64): Likewise.
2347 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
2349         PR rtl-optimization/87720
2350         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
2352 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
2354         * gimple-ssa-isolate-paths.c
2355         (find_implicit_erroneous_behavior): Do not change code if the
2356         pass is running for warnings only.
2357         (find_explicit_erroneous_behavior): Likewise.
2359 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
2361         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
2362         Define as rs6000_mangle_decl_assembler_name.
2363         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
2364         long double to IEEE long double, switch the names of the long
2365         double built-in functions to be <func>f128 instead of <func>l.
2367 2018-10-24  Martin Sebor  <msebor@redhat.com>
2369         * doc/extend.texi (nonnull): List no-argument form.  Reference
2370         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
2372 2018-10-24  Richard Biener  <rguenther@suse.de>
2374         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
2376 2018-10-24  Martin Liska  <mliska@suse.cz>
2378         PR tree-optimization/84436
2379         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
2380         Remove.
2381         (switch_conversion::contains_linear_function_p): New.
2382         (switch_conversion::build_one_array): Support linear
2383         transformation on input.
2384         * tree-switch-conversion.h (struct switch_conversion): Add
2385         contains_linear_function_p declaration.
2387 2018-10-24  Richard Biener  <rguenther@suse.de>
2389         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
2390         if its argument is CONSTANT_CLASS_P.
2392 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2394         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
2395         it is wrong for forward declarations.
2397 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2399         * config/s390/s390.c (s390_check_qrst_address): Add the missing
2400         SYMBOL_REF_P () check.
2402 2018-10-24  Richard Biener  <rguenther@suse.de>
2404         PR tree-optimization/87105
2405         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
2406         dump classification.
2407         (vect_analyze_data_ref_accesses): Handle duplicate loads and
2408         stores by splitting the affected group after the fact.
2409         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
2410         fail the SLP build because of size constraints.
2412 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2414         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
2415         * configure: Regenerate.
2416         * config.in: Regenerate.
2417         * varasm.c (mergeable_string_section): Use readonly_data_section
2418         if linker doesn't support SHF_MERGE with alignment > 8.
2419         (mergeable_constant_section): Likewise.
2421 2018-10-24  Richard Biener  <rguenther@suse.de>
2423         PR tree-optimization/84013
2424         * tree-ssa-structalias.c (struct msdi_data): New struct for
2425         marshalling data to walk_stmt_load_store_ops.
2426         (maybe_set_dependence_info): Refactor as callback for
2427         walk_stmt_load_store_ops.
2428         (compute_dependence_clique): Set restrict info on all stmt kinds.
2430 2018-10-24  Martin Liska  <mliska@suse.cz>
2432         * cgraph.c (cgraph_node::dump):
2433         Remove reduntant dumps and make tp_first_run dump more compact.
2435 2018-10-24  Richard Biener  <rguenther@suse.de>
2437         PR tree-optimization/87665
2438         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2439         to reflect reality.
2441 2018-10-12  Jeff Law  <law@redhat.com>
2443         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
2444         for H8/S.
2446 2018-10-23  Richard Biener  <rguenther@suse.de>
2448         * tree-vrp.c (add_assert_info): Guard dump_printf with
2449         dump_enabled_p.
2450         * gimple-ssa-evrp-analyze.c
2451         (evrp_range_analyzer::record_ranges_from_incoming_edge):
2452         Use value_range::ignore_equivs_equal_p.
2454 2018-10-23  Richard Biener  <rguenther@suse.de>
2456         PR tree-optimization/87105
2457         PR tree-optimization/87608
2458         * passes.def (pass_all_early_optimizations): Add early phi-opt
2459         after dce.
2460         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
2461         addition to debug stmts.
2462         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
2463         and abs replacement early.
2464         * tree-cfg.c (gimple_empty_block_p): Likewise.
2466 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
2468         PR target/86383
2469         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
2470         specified to configure.
2471         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
2473 2018-10-23  Richard Biener  <rguenther@suse.de>
2475         PR tree-optimization/87700
2476         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2478 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
2480         PR target/87674
2481         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
2482         second argument from __mmask16 to __mmask8.
2483         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
2484         _mm_mask_packs_epi32): Likewise.
2485         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
2486         Likewise.
2487         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
2489 2018-10-23  Richard Biener  <rguenther@suse.de>
2491         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
2493 2018-10-23  Richard Biener  <rguenther@suse.de>
2495         PR tree-optimization/86144
2496         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
2497         over simd attribute.
2499 2018-10-23  Richard Biener  <rguenther@suse.de>
2501         PR tree-optimization/87693
2502         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
2503         the case we do not find the taken edge.
2505 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
2506             Jinsong Ji  <jji@us.ibm.com>
2508         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
2509         (_mm_store_pd): Use unaligned vector type for pointer cast.
2510         (_mm_maskmoveu_si128): Likewise.
2511         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
2512         (_mm_store_ps): Use unaligned vector type for pointer cast.
2514 2018-10-22  Paul Koning  <ni1d@arrl.net>
2516         * symtab.c (symtab_node::increase_alignment): Correct max
2517         alignment check.
2519 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
2521         PR tree-optimization/87633
2522         * match.pd: Do not generate unordered integer comparisons.
2524 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2526         PR rtl-optimization/87600
2527         * combine.c: Add include of expr.h.
2528         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
2529         register to a pseudo.
2530         (make_more_copies): New function, add a copy to a new pseudo after
2531         the moves from hard registers into pseudos.
2532         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
2533         later.  Call make_more_copies.
2535 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
2537         * lra-constraints.c (process_alt_operands): New local array,
2538         matching_early_clobber.  Check matching_early_clobber before
2539         decrementing reject, and set matching_early_clobber after.
2541 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2543         PR target/87598
2544         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
2545         call output_addr_const and hope for the best.
2547 2018-10-22  Richard Biener  <rguenther@suse.de>
2549         * gimple-ssa-evrp-analyze.c
2550         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
2551         smarter about what ranges to use.
2552         * tree-vrp.c (add_assert_info): Dump here.
2553         (register_edge_assert_for_2): Instead of here at multiple but
2554         not all places.
2556         * gcc.dg/tree-ssa/evrp12.c: New testcase.
2557         * gcc.dg/predict-6.c: Adjust.
2558         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
2559         * gcc.dg/tree-ssa/vrp02.c: Likewise.
2560         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
2562 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
2563         Richard Biener  <rguenther@suse.de>
2565         * bitmap.h: Update data structure documentation, including a
2566         description of bitmap views as either linked-lists or splay trees.
2567         (struct bitmap_element_def): Update comments for splay tree bitmaps.
2568         (struct bitmap_head_def): Likewise.
2569         (bitmap_list_view, bitmap_tree_view): New prototypes.
2570         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
2571         tree_form fields.
2572         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
2573         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
2574         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
2575         released bitmap element here.
2576         (bitmap_element_free): Remove.
2577         (bitmap_elt_clear_from): Work on splay tree bitmaps.
2578         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
2579         this function similar ones such that linked-list bitmap implementation
2580         functions are grouped.
2581         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
2582         and moved for grouping.
2583         (bitmap_list_insert_element_after): Renamed from
2584         bitmap_elt_insert_after, and moved for grouping.
2585         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
2586         (bitmap_tree_link_left, bitmap_tree_link_right,
2587         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
2588         bitmap_tree_link_element, bitmap_tree_unlink_element,
2589         bitmap_tree_find_element): New functions for splay-tree bitmap
2590         implementation.
2591         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
2592         Renamed and moved, see above entries.
2593         (bitmap_tree_listify_from): New function to convert part of a splay
2594         tree bitmap to a linked-list bitmap.
2595         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
2596         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
2597         (bitmap_find_bit): Remove.
2598         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
2599         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
2600         Handle splay tree bitmaps.
2601         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
2602         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
2603         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
2604         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
2605         bitmap_intersect_compl_p, bitmap_ior_and_compl,
2606         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
2607         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
2608         corresponding changes to use linked-list specific bitmap_element
2609         manipulation functions as applicable for efficiency.
2610         (bitmap_tree_to_vec): New function.
2611         (debug_bitmap_elt_file): New function split out from ...
2612         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
2613         (bitmap_print): Likewise.
2615         PR tree-optimization/63155
2616         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
2617         SSA edge worklists.
2618         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
2619         in tree-view.
2621 2018-10-22  Martin Liska  <mliska@suse.cz>
2623         PR tree-optimization/87686
2624         Revert
2625         2018-08-29  Martin Liska  <mliska@suse.cz>
2627         * tree-switch-conversion.c (switch_conversion::expand):
2628         Strenghten assumption about gswitch statements.
2630 2018-10-22  Martin Liska  <mliska@suse.cz>
2632         * ipa-icf.c (sem_item::compare_attributes): Remove.
2633         (sem_item::compare_referenced_symbol_properties): Use
2634         attribute_list_equal instead.
2635         (sem_function::equals_wpa): Likewise.
2636         * ipa-icf.h: Remove compare_attributes.
2638 2018-10-22  Richard Biener  <rguenther@suse.de>
2640         PR middle-end/87682
2641         * mem-stats.h (mem_usage::operator==): Fix pasto.
2643 2018-10-22  Richard Biener  <rguenther@suse.de>
2645         PR tree-optimization/87640
2646         * tree-vrp.c (set_value_range_with_overflow): Decompose
2647         incomplete result.
2648         (extract_range_from_binary_expr_1): Adjust.
2650 2018-10-22  Martin Jambor  <mjambor@suse.cz>
2652         * tree-eh.h (stmt_could_throw_p): Add function parameter.
2653         (stmt_can_throw_external): Likewise.
2654         (stmt_can_throw_internal): Likewise.
2655         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
2656         (lower_eh_constructs_2): Likewise.
2657         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
2658         (stmt_can_throw_external): Likewise.
2659         (stmt_can_throw_internal): Likewise.
2660         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
2661         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
2662         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
2663         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
2664         (pass_lower_eh_dispatch::execute): Pass cfun to
2665         stmt_can_throw_external.
2666         (cleanup_empty_eh): Likewise.
2667         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
2668         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
2669         stmt_can_throw_external instead of pushing it to cfun.
2670         (symbol_table::create_edge): Likewise.
2671         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
2672         stmt_can_throw_internal.
2673         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
2674         to stmt_could_throw_p.
2675         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
2676         stmt_can_throw_internal.
2677         (pass_store_merging::execute): Likewise.
2678         * gimple-ssa-strength-reduction.c
2679         (find_candidates_dom_walker::before_dom_children): Pass cfun to
2680         stmt_could_throw_p.
2681         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
2682         stmt_can_throw_internal.
2683         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
2684         to stmt_can_throw_external.
2685         (check_stmt): Pass cfun to stmt_could_throw_p.
2686         (check_stmt): Pass cfun to stmt_can_throw_external.
2687         (pass_nothrow::execute): Likewise.
2688         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
2689         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
2690         stmt_can_throw_internal.
2691         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
2692         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
2693         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
2694         * tree-complex.c (expand_complex_libcall): Pass cfun to
2695         stmt_could_throw_p and to stmt_can_throw_internal.
2696         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
2697         * tree-inline.c (copy_edges_for_bb): Likewise.
2698         (maybe_move_debug_stmts_to_successors): Likewise.
2699         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
2700         stmt_could_throw_p.
2701         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
2702         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
2703         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
2704         stmt_can_throw_internal.
2705         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2706         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
2707         stmt_could_throw_p.
2708         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
2709         stmt_can_throw_internal.
2710         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
2711         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
2712         stmt_could_throw_p.
2713         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
2714         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
2715         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2716         (convert_mult_to_fma_1): Likewise.
2717         (convert_to_divmod): Likewise.
2718         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
2719         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
2720         * tree-ssa-propagate.c
2721         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
2722         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
2723         (maybe_optimize_range_tests): Likewise.
2724         (linearize_expr_tree): Likewise.
2725         (reassociate_bb): Likewise.
2726         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2727         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
2728         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
2729         (handle_char_store): Likewise.
2730         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
2731         stmt_can_throw_internal.
2732         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
2733         stmt_could_throw_p.
2734         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2735         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
2736         (vectorizable_simd_clone_call): Likewise.
2737         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
2738         (gimple_stringop_fixed_value): Likewise.
2740 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
2742         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
2743         literal pool references.
2744         (s390_check_qrst_address): Adapt to the new behavior of
2745         s390_loadrelative_operand_p ().
2747 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2749         PR target/72782
2750         * config/i386/sse.md (*andnot<mode>3_bcst): New.
2752 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2754         PR target/72782
2755         * config/i386/sse.md (*<code><mode>3_bcst): New.
2757 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2759         PR target/72782
2760         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
2761         V4DI, V16SI and V8DI.
2762         (*sub<mode>3<mask_name>_bcst): New.
2763         (*add<mode>3<mask_name>_bcst): Likewise.
2765 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
2766             Jinsong Ji  <jji@us.ibm.com>
2768         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
2769         __m64 with __vector unsigned long long for compatibility.
2770         (_mm_movemask_epi8): Likewise.
2771         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
2772         (_mm_cvttps_pi32): Likewise.
2773         (_mm_cvtpi32_ps): Likewise.
2774         (_mm_cvtps_pi16): Likewise.
2775         (_mm_loadh_pi): Likewise.
2776         (_mm_storeh_pi): Likewise.
2777         (_mm_movehl_ps): Likewise.
2778         (_mm_movelh_ps): Likewise.
2779         (_mm_loadl_pi): Likewise.
2780         (_mm_storel_pi): Likewise.
2781         (_mm_movemask_ps): Likewise.
2782         (_mm_shuffle_pi16): Likewise.
2784 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2786         PR target/72782
2787         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
2788         __builtin_ia32_vfnmsubpd512_mask.
2789         (_mm512_mask_fnmsub_round_pd): Likewise.
2790         (_mm512_fnmsub_pd): Likewise.
2791         (_mm512_mask_fnmsub_pd): Likewise.
2792         (_mm512_maskz_fnmsub_round_pd): Use
2793         __builtin_ia32_vfnmsubpd512_maskz.
2794         (_mm512_maskz_fnmsub_pd): Likewise.
2795         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
2796         (_mm512_mask_fnmsub_round_ps): Likewise.
2797         (_mm512_fnmsub_ps): Likewise.
2798         (_mm512_mask_fnmsub_ps): Likewise.
2799         (_mm512_maskz_fnmsub_round_ps): Use
2800         __builtin_ia32_vfnmsubps512_maskz.
2801         (_mm512_maskz_fnmsub_ps): Likewise.
2802         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
2803         __builtin_ia32_vfnmsubpd256_mask.
2804         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
2805         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2806         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
2807         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2808         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2809         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
2810         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
2811         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
2812         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
2813         __builtin_ia32_vfnmsubpd.
2814         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
2815         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
2816         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
2817         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
2818         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
2819         * config/i386/i386-builtin.def: Add
2820         __builtin_ia32_vfnmsubpd256_mask,
2821         __builtin_ia32_vfnmsubpd256_maskz,
2822         __builtin_ia32_vfnmsubpd128_mask,
2823         __builtin_ia32_vfnmsubpd128_maskz,
2824         __builtin_ia32_vfnmsubps256_mask,
2825         __builtin_ia32_vfnmsubps256_maskz,
2826         __builtin_ia32_vfnmsubps128_mask,
2827         __builtin_ia32_vfnmsubps128_maskz,
2828         __builtin_ia32_vfnmsubpd512_mask,
2829         __builtin_ia32_vfnmsubpd512_maskz,
2830         __builtin_ia32_vfnmsubps512_mask,
2831         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
2832         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
2833         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
2834         __builtin_ia32_vfnmsubpd256.
2835         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
2836         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
2837         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
2838         Likewise.
2839         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
2840         Likewise.
2841         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
2842         Likewise.
2843         (fmai_vmfnmsub_<mode><round_name>): Likewise.
2845 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2847         PR target/72782
2848         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
2849         __builtin_ia32_vfnmaddpd512_mask.
2850         (_mm512_mask_fnmadd_round_pd): Likewise.
2851         (_mm512_fnmadd_pd): Likewise.
2852         (_mm512_mask_fnmadd_pd): Likewise.
2853         (_mm512_maskz_fnmadd_round_pd): Use
2854         __builtin_ia32_vfnmaddpd512_maskz.
2855         (_mm512_maskz_fnmadd_pd): Likewise.
2856         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
2857         (_mm512_mask_fnmadd_round_ps): Likewise.
2858         (_mm512_fnmadd_ps): Likewise.
2859         (_mm512_mask_fnmadd_ps): Likewise.
2860         (_mm512_maskz_fnmadd_round_ps): Use
2861         __builtin_ia32_vfnmaddps512_maskz.
2862         (_mm512_maskz_fnmadd_ps): Likewise.
2863         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
2864         __builtin_ia32_vfnmaddpd256_mask.
2865         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
2866         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
2867         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
2868         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2869         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2870         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
2871         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
2872         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
2873         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
2874         __builtin_ia32_vfnmaddpd.
2875         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
2876         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
2877         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
2878         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
2879         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
2880         * config/i386/i386-builtin.def: Add
2881         __builtin_ia32_vfnmaddpd256_mask,
2882         __builtin_ia32_vfnmaddpd256_maskz,
2883         __builtin_ia32_vfnmaddpd128_mask,
2884         __builtin_ia32_vfnmaddpd128_maskz,
2885         __builtin_ia32_vfnmaddps256_mask,
2886         __builtin_ia32_vfnmaddps256_maskz,
2887         __builtin_ia32_vfnmaddps128_mask,
2888         __builtin_ia32_vfnmaddps128_maskz,
2889         __builtin_ia32_vfnmaddpd512_mask,
2890         __builtin_ia32_vfnmaddpd512_maskz,
2891         __builtin_ia32_vfnmaddps512_mask,
2892         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
2893         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
2894         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
2895         __builtin_ia32_vfnmaddpd256.
2896         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
2897         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
2898         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
2899         Likewise.
2900         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
2901         Likewise.
2902         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
2903         Likewise.
2904         (fmai_vmfnmadd_<mode><round_name>): Likewise.
2906 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2908         PR target/72782
2909         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
2910         __builtin_ia32_vfmsubpd512_mask.
2911         (_mm512_mask_fmsub_round_pd): Likewise.
2912         (_mm512_fmsub_pd): Likewise.
2913         (_mm512_mask_fmsub_pd): Likewise.
2914         (_mm512_maskz_fmsub_round_pd): Use
2915         __builtin_ia32_vfmsubpd512_maskz.
2916         (_mm512_maskz_fmsub_pd): Likewise.
2917         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
2918         (_mm512_mask_fmsub_round_ps): Likewise.
2919         (_mm512_fmsub_ps): Likewise.
2920         (_mm512_mask_fmsub_ps): Likewise.
2921         (_mm512_maskz_fmsub_round_ps): Use
2922         __builtin_ia32_vfmsubps512_maskz.
2923         (_mm512_maskz_fmsub_ps): Likewise.
2924         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
2925         __builtin_ia32_vfmsubpd256_mask.
2926         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
2927         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2928         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
2929         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2930         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2931         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
2932         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
2933         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
2934         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
2935         __builtin_ia32_vfmsubpd.
2936         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
2937         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
2938         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
2939         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
2940         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
2941         * config/i386/i386-builtin.def: Add
2942         __builtin_ia32_vfmsubpd256_mask,
2943         __builtin_ia32_vfmsubpd256_maskz,
2944         __builtin_ia32_vfmsubpd128_mask,
2945         __builtin_ia32_vfmsubpd128_maskz,
2946         __builtin_ia32_vfmsubps256_mask,
2947         __builtin_ia32_vfmsubps256_maskz,
2948         __builtin_ia32_vfmsubps128_mask,
2949         __builtin_ia32_vfmsubps128_maskz,
2950         __builtin_ia32_vfmsubpd512_mask,
2951         __builtin_ia32_vfmsubpd512_maskz,
2952         __builtin_ia32_vfmsubps512_mask,
2953         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
2954         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
2955         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
2956         __builtin_ia32_vfmsubpd256.
2957         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
2958         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
2959         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
2960         Likewise.
2961         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
2962         Likewise.
2963         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
2964         Likewise.
2965         (fmai_vmfmsub_<mode><round_name>): Likewise.
2967 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2969         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
2970         Remove plus.  Renamed to ...
2971         (*sub<mode>3<mask_name>_bcst): This.
2972         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
2973         (*add<mode>3<mask_name>_bcst): This.
2975 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2977         PR target/72782
2978         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
2980 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2982         PR target/87662
2983         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
2984         (_mm_or_epi32): Likewise.
2985         (_mm256_xor_epi32): Likewise.
2986         (_mm_xor_epi32): Likewise.
2987         (_mm256_or_epi64): Likewise.
2988         (_mm_or_epi64): Likewise.
2989         (_mm256_xor_epi64): Likewise.
2990         (_mm_xor_epi64): Likewise.
2992 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
2994         PR target/72782
2995         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
2997 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
2999         PR middle-end/87647
3000         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3002 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3004         * doc/ux.texi: Move @section directly after @node.
3006 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3008         PR middle-end/85488
3009         PR middle-end/87649
3010         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3011         depend closely nested inside of loop with ordered clause with
3012         a parameter.
3014 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3016         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3017         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3018         * doc/ux.texi: New file.
3020 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3022         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3023         be the first CR field allocated.
3025 2018-10-19  Richard Biener  <rguenther@suse.de>
3027         PR target/87657
3028         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3029         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3031 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3033         PR target/72782
3034         * config/i386/sse.md
3035         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3036         (*add<mode>3<mask_name>_bcst_2): Likewise.
3038 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3040         * config/i386/sse.md
3041         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3042         Replace nonimmediate_operand with register_operand.
3043         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3044         Likewise.
3045         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3046         Likewise.
3048 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3050         PR rtl-optimization/87596
3051         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3052         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3053         for instructions in FROM..TO range.
3055 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
3057         * cfgexpand.c (expand_one_var): Use specific wording in error message
3058         for non-local frame variables.
3059         * stor-layout.c (layout_decl): Do not issue a warning for them.
3061 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
3063         * haifa-sched.c (priority): Add force_recompute parameter.
3064         (apply_replacement): Call priority () with force_recompute = true.
3065         (restore_pattern): Likewise.
3067 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3069         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3070         HOST_BITS_PER_WIDE_INT.
3071         (test_vector_ops_duplicate): Likewise.
3073 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3075         PR target/72782
3076         * config/i386/sse.md (VF_AVX512): New.
3077         (avx512bcst): Likewise.
3078         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3079         Likewise.
3080         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3081         Likewise.
3082         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3083         Likewise.
3085 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
3087         * doc/invoke.texi (-dumpversion): Improve grammar.
3088         (-dumpfullversion): Make more consistent with -dumpversion.
3090 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
3092         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3093         Set JUMP_LABEL to the jump insn.
3094         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3095         Predict emitted jump and add label to jump insn.
3097 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
3099         PR tree-optimization/87562
3100         * input.c (get_substring_ranges_for_loc): Use
3101         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3102         getting the linemap for the endpoint.  Verify that it's either
3103         in the same linemap as the start point's spelling location, or
3104         at least in the same file.
3106 2018-10-18  Richard Biener  <rguenther@suse.de>
3108         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3109         feed width-specific load/store costs through ix86_vec_cost.
3110         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3111         (k8_cost): Likewise.
3112         (bdver_cost): Likewise.
3113         (znver1_cost): Likewise.
3114         (btver1_cost): Likewise.
3115         (btver2_cost): Likewise.
3117 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3119         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3120         to simplify subreg of vec_merge.
3122 2018-10-18  Richard Biener  <rguenther@suse.de>
3124         * config/i386/i386.c: Fix costing of vector FMA.
3126 2018-10-18  Richard Biener  <rguenther@suse.de>
3128         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3129         and argument.
3130         (ix86_builtin_vectorization_cost): For vec_construct properly
3131         cost insertion into SSE regs.
3132         (...): Adjust calls to ix86_vec_cost.
3134 2018-10-18  Richard Biener  <rguenther@suse.de>
3136         PR middle-end/87087
3137         Revert
3138         2018-02-07  Richard Biener  <rguenther@suse.de>
3140         PR tree-optimization/84204
3141         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3142         this place.
3144 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
3146         PR target/87537
3147         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3148         of vec_duplicate.
3149         (test_vector_ops_duplicate): Add test for a scalar subreg of a
3150         VEC_MERGE of a VEC_DUPLICATE.
3152 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3154         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3155         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3156         * doc/standards.texi (C Language): Document C2X.
3157         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3158         (rl78_option_override): Handle "GNU C2X" language name.
3160 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3162         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3163         Document C17 as published in 2018.
3165 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3167         PR middle-end/87623
3168         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3169         bail out if both sides do not have the same storage order.
3171 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3173         * bitmap.c (bitmap_head::dump): New.
3174         * bitmap.h (bitmap_head): Add dump().
3175         * gimple-ssa-evrp-analyze.c
3176         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3177         (evrp_range_analyzer::set_ssa_range_info): Same.
3178         (evrp_range_analyzer::record_ranges_from_phis): Same.
3179         (evrp_range_analyzer::record_ranges_from_stmt): Same.
3180         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3181         * gimple-ssa-sprintf.c (get_int_range): Same.
3182         (format_integer): Same.
3183         (sprintf_dom_walker::handle_gimple_call): Same.
3184         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3185         (ipcp_vr_lattice::top_p): Same.
3186         (ipcp_vr_lattice::bottom_p): Same.
3187         (ipcp_vr_lattice::set_to_bottom): Same.
3188         (ipa_vr_operation_and_type_effects): Same.
3189         (propagate_vr_across_jump_function): Same.
3190         (ipcp_store_vr_results): Same.
3191         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3192         (ipa_print_node_jump_functions_for_edge): Same.
3193         (ipa_get_value_range): Same.
3194         (ipa_compute_jump_functions_for_edge): Same.
3195         (ipa_write_jump_function): Same.
3196         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3197         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3198         Same.
3199         * vr-values.c (set_value_range_to_nonnegative): Same.
3200         (set_value_range_to_truthvalue): Same.
3201         (vr_values::get_value_range): Same.
3202         (vr_values::set_defs_to_varying): Same.
3203         (vr_values::update_value_range): Same.
3204         (symbolic_range_based_on_p): Same.
3205         (vr_values::op_with_boolean_value_range_p): Same.
3206         (vr_values::extract_range_for_var_from_comparison_expr): Same.
3207         (vr_values::extract_range_from_ssa_name): Same.
3208         (vr_values::extract_range_from_binary_expr): Same.
3209         (vr_values::extract_range_from_unary_expr): Same.
3210         (vr_values::extract_range_from_cond_expr): Same.
3211         (vr_values::extract_range_from_comparison): Same.
3212         (vr_values::check_for_binary_op_overflow): Same.
3213         (vr_values::extract_range_basic): Same.
3214         (vr_values::extract_range_from_assignment): Same.
3215         (compare_ranges): Same.
3216         (compare_range_with_value): Same.
3217         (vr_values::adjust_range_with_scev): Same.
3218         (vrp_valueize): Same.
3219         (vrp_valueize_1): Same.
3220         (vr_values::get_vr_for_comparison): Same.
3221         (vr_values::compare_name_with_value): Same.
3222         (vr_values::compare_names): Same.
3223         (vr_values::vrp_evaluate_conditional): Same.
3224         (find_case_label_ranges): Same.
3225         (vr_values::vrp_visit_switch_stmt): Same.
3226         (vr_values::extract_range_from_phi_node): Same.
3227         (vr_values::simplify_div_or_mod_using_ranges): Same.
3228         (vr_values::simplify_bit_ops_using_ranges): Same.
3229         (test_for_singularity): Same.
3230         (range_fits_type_p): Same.
3231         (vr_values::simplify_cond_using_ranges_1): Same.
3232         (vr_values::simplify_switch_using_ranges): Same.
3233         (vr_values::simplify_float_conversion_using_ranges): Same.
3234         (vr_values::two_valued_val_range_p): Same.
3235         (vr_values::add_equivalence): Move to value_range::equiv_add.
3236         * vr-values.h (vr_values::add_equivalence): Remove.
3237         (VR_INITIALIZER): Remove.
3238         * tree-vrp.c (value_range::set): New.
3239         (value_range::equiv_add): New.
3240         (value_range::value_range): New.
3241         (value_range::deep_copy): New.
3242         (value_range::check): New.
3243         (value_range::equal_p): New.
3244         (value_range::ignore_equivs_equal_p): New.
3245         (value_range::operator==): New.
3246         (value_range::operator!=): New.
3247         (value_range::symbolic_p): New.
3248         (value_range::numeric_p): New.
3249         (value_range::set_undefined): New.
3250         (value_range::set_varying): New.
3251         (value_range::may_contain_p): New.
3252         (value_range::equiv_clear): New.
3253         (value_range::singleton_p): New.
3254         (value_range::intersect): New.
3255         (value_range::dump): New.
3256         (value_range::set_and_canonicalize): New.
3257         (set_value_range): Adjust for value_range API.
3258         (set_value_range_to_undefined): Same.
3259         (set_value_range_to_varying): Same.
3260         (set_and_canonicalize_value_range): Same.
3261         (set_value_range_to_nonnull): Same.
3262         (set_value_range_to_null): Same.
3263         (range_is_null): Same.
3264         (range_is_nonnull): Same.
3265         (range_int_cst_p): Same.
3266         (range_int_cst_singleton_p): Same.
3267         (symbolic_range_p): Same.
3268         (range_includes_zero_p): Same.
3269         (value_range_constant_singleton): Same.
3270         (vrp_set_zero_nonzero_bits): Same.
3271         (ranges_from_anti_range): Same.
3272         (extract_range_into_wide_ints): Same.
3273         (extract_range_from_multiplicative_op): Same.
3274         (set_value_range_with_overflow): Same.
3275         (extract_range_from_binary_expr_1): Same.
3276         (extract_range_from_unary_expr): Same.
3277         (dump_value_range): Same.
3278         (debug_value_range): Same.
3279         (vrp_prop::check_array_ref): Same.
3280         (vrp_prop::check_mem_ref): Same.
3281         (vrp_prop::vrp_initialize): Same.
3282         (vrp_prop::visit_stmt): Same.
3283         (intersect_ranges): Same.
3284         (vrp_prop::visit_phi): Same.
3285         (vrp_prop::vrp_finalize): Same.
3286         (determine_value_range_1): Same.
3287         (determine_value_range): Same.
3288         (vrp_intersect_ranges_1): Rename to...
3289         (vrp_intersect_1): this.
3290         (vrp_intersect_ranges): Rename to...
3291         (value_range::intersect_helper): ...this.
3292         (vrp_meet_1): Rename to...
3293         (value_range::union_helper): ...this.
3294         (vrp_meet): Rename to...
3295         (value_range::union_): ...this.
3296         (copy_value_range): Remove.
3297         * tree-vrp.h (struct value_range): Rewrite into a proper class.
3298         (value_range::vrtype): New.
3299         (value_range::type): New.
3300         (value_range::equiv): New.
3301         (value_range::min): New.
3302         (value_range::max): New.
3303         (value_range::varying_p): New.
3304         (value_range::undefined_p): New.
3305         (value_range::null_p): New.
3306         (value_range::equiv_add): New.
3307         (copy_value_range): Remove.
3309 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
3311         * Makefile.in (SELFTEST_TARGETS): New.
3312         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3313         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
3314         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
3315         c/Make-lang.in.
3316         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
3317         (selftest-c++-gdb, selftest-c++-valgrind): Move to
3318         cp/Make-lang.in.
3319         * configure: Regenerate.
3320         * configure.ac (selftest_languages): New.
3322 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3324         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
3325         overflow wraps argument.
3326         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
3327         wide_int_range_multiplicative_op.
3328         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
3329         overflow wraps argument.
3330         (wide_int_range_multiplicative_op): Same.
3331         (wide_int_range_lshift): Same.
3332         (wide_int_range_div): Same.
3333         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
3334         (wide_int_range_lshift): Same.
3335         (wide_int_range_div): Same.
3337 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3339         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
3340         use sign as argument.
3341         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
3342         wide_int_range_shift_undefined_p.
3344 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3346         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
3347         Rename to...
3348         (@despeculate_copy<ALLI_TI:mode>): ... This.
3349         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
3350         switch statement.
3352 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3354         * config.gcc: Obsolete *-*-solaris2.10*.
3355         * doc/install.texi (Specific, *-*-solaris2*): Document it.
3357 2018-10-12  Jeff Law  <law@redhat.com>
3359         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
3360         reg + sym +- const_int addressing modes.
3362 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3364         * common.opt (fdiagnostics-minimum-margin-width=): New option.
3365         * diagnostic-show-locus.c (layout::layout): Apply the minimum
3366         margin width.
3367         (layout::start_annotation_line): Only print up to 3 of the
3368         margin character, to avoid touching the left-hand side.
3369         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
3370         minimum margin width, as set by test_diagnostic_context's ctor.
3371         (selftest::test_fixit_insert_containing_newline): Likewise.
3372         (selftest::test_fixit_insert_containing_newline_2): Likewise.
3373         (selftest::test_line_numbers_multiline_range): Clear
3374         dc.min_margin_width.
3375         * diagnostic.c (diagnostic_initialize): Initialize
3376         min_margin_width.
3377         * diagnostic.h (struct diagnostic_context): Add field
3378         "min_margin_width".
3379         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
3380         * opts.c (common_handle_option): Handle
3381         OPT_fdiagnostics_minimum_margin_width_.
3382         * selftest-diagnostic.c
3383         (selftest::test_diagnostic_context::test_diagnostic_context):
3384         Initialize min_margin_width to 6.
3385         * toplev.c (general_init): Initialize global_dc->min_margin_width.
3387 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3389         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3390         Fix usage of "error_at_rich_loc" in the comment.
3392 2018-10-15  Renlin Li  <renlin.li@arm.com>
3394         PR target/87563
3395         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
3396         if-conversioned loop when it contains ifn with types not
3397         supported by backend.
3398         * internal-fn.c (expand_direct_optab_fn): Add an assert.
3399         (direct_internal_fn_supported_p): New helper function.
3400         * internal-fn.h (direct_internal_fn_supported_p): Declare.
3402 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
3404         PR target/87572
3405         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3406         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3407         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3409 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
3411         PR tree-optimization/87022
3412         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
3413         bits in dist vector rather than the first one.
3415 2018-10-15  Richard Biener  <rguenther@suse.de>
3417         PR middle-end/87610
3418         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
3419         (visit_loadstore): When a used restrict tag escaped verify that
3420         the points-to solution of "other" pointers do not include
3421         escaped.
3422         (compute_dependence_clique): If a used restrict tag escaped
3423         communicated that down to visit_loadstore.
3425 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
3427         * config/s390/s390.c (s390_expand_vec_init): Force vector element
3428         into reg if it isn't a general operand.
3430 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3432         PR target/87599
3433         * config/i386/sse.md (*vec_dupv2di): Add register source to
3434         movddup.
3436 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3438         PR target/87572
3439         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3440         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
3441         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3442         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3444 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
3446         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
3447         (notice_args_size): Set it in the current trace if no insn that can
3448         throw internally has been seen yet.
3449         (connect_traces): When connecting args_size between traces, allow the
3450         incoming values not to match if there is an insn setting it before the
3451         first insn that can throw internally; in that case, force the creation
3452         of a CFI note on this latter insn.
3454 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
3456         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
3457         the base class, because using the injected-class-name was not clearly
3458         specified until DR 176.
3460 2018-10-12  Paul Koning  <ni1d@arrl.net>
3462         * config/pdp11/pdp11.md (doloop_end): New expander.
3463         (doloop_end_insn): renamed from "doloop_end".
3464         (addqi3): New pattern.
3465         (subqi3): New pattern.
3466         * config/pdp11/predicates.md (incdec_operand): New predicate.
3468 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
3470         PR middle-end/81376
3471         * real.c (format_helper::can_represent_integral_type_p): New function
3472         * real.h (format_helper::can_represent_integral_type_p): Ditto.
3473         * match.pd: New pattern.
3475 2018-10-12  Alexandre Oliva <oliva@adacore.com>
3477         * configure.ac: Introduce --enable-large-address-aware
3478         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
3479         * doc/install.texi: Document it.
3480         * configure, config.in: Rebuilt.
3481         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
3482         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
3483         (LINK_SPEC): Insert it.
3484         * config/i386/mingw-w64.h: Likewise.
3486         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
3488 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
3490         PR rtl-optimization/87600
3491         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
3493 2018-10-12  Paul Koning  <ni1d@arrl.net>
3495         * doc/md.texi (doloop_end): Document that the pattern code may
3496         need to check operand mode.
3498 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
3500         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
3501         to zero-extend between int and floating-point registers.
3502         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
3503         ldp into floating-point registers.  Add type and arch attributes.
3504         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
3505         Use f_loads for type attribute.
3507 2018-10-11  Martin Sebor  <msebor@redhat.com>
3509         * doc/extend.texi (attribute packed): Correct typos.
3511 2018-10-11  Martin Sebor  <msebor@redhat.com>
3513         * doc/extend.texi (attribute flatten): Mention interaction with
3514         noinline.
3516 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3518         PR target/87156
3519         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3520         Set new_decl virtual flag to zero.
3522 2018-10-11  Martin Sebor  <msebor@redhat.com>
3524         PR middle-end/87593
3525         * doc/extend.texi (attribute format_arg): Discuss using multiple
3526         attributes on a single function.
3528 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
3530         PR tree-optimization/86829
3531         * match.pd (sin (atan (x))): New simplification rules.
3532         (cos (atan (x))): Likewise.
3533         * real.c (build_sinatan_real): New function.
3534         * real.h (build_sinatan_real): Prototype.
3536 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
3538         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
3539         function.
3540         (fold_mergeeo_helper): New helper function.
3541         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
3542         intrinsics.  Correct some whitespace indentation issues.
3544 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
3546         PR target/87511
3547         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
3548         Use HOST_WIDE_INT_1U for shift.
3550 2018-10-11  Doug Rupp  <rupp@adacore.com>
3551             Olivier Hainque  <hainque@adacore.com>
3553         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
3554         Pass --relax to the linker for RTPs.
3555         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
3557 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
3558             Jan Hubicka  <jh@suse.cz>
3559             Martin Jambor  <mjambor@suse.cz>
3561         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
3562         the same elements are repeated rather than printing all of them.
3563         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
3564         "repeated" elements.
3565         * read-rtl-function.c (test_loading_repeat): New function.
3566         (read_rtl_function_c_tests): Call test_loading_repeat.
3567         * rtl-tests.c (test_dumping_repeat): New function.
3568         (rtl_tests_c_tests): Call test_dumping_repeat.
3570 2018-10-11  Richard Biener  <rguenther@suse.de>
3572         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
3573         bdver?_cost): Unify to ...
3574         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
3575         * config/i386/i386.c (processor_cost_table): Adjust.
3577 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
3579         PR middle-end/87574
3580         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
3581         the thunk when expanding to GIMPLE.
3583 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3585         * varasm.c (mergeable_string_section): Don't try to move zero-length
3586         strings to the merge section.
3588 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
3590         PR target/87573
3591         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
3593 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
3595         PR target/87550
3596         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
3597         to special_args set.
3599 2018-10-10  Richard Biener  <rguenther@suse.de>
3601         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
3602         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
3603         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
3604         and recursing and pattern terminating the recursion on SSE
3605         vector width using ix86_expand_reduc.
3606         (reduc_sminmax_scal_<mode>): Split into part reducing to half
3607         width and recursing and SSE2 vector variant doing the final
3608         reduction with ix86_expand_reduc.
3609         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
3610         with terminating the recursion at AVX level, splitting that
3611         to SSE there.
3613 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
3615         * genmatch.c (error_cb): Rename to...
3616         (diagnostic_cb): ...this, converting int params to enums.
3617         (fatal_at): Update for renaming.
3618         (warning_at): Likewise.
3619         (main): Likewise.
3620         * input.c (selftest::ebcdic_execution_charset::apply):
3621         Update for renaming of...
3622         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
3623         to...
3624         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
3625         converting level and reason to enums.
3626         (class selftest::lexer_error_sink): Rename to...
3627         (class selftest::lexer_test_options): ...this, renaming field
3628         "m_errors" to "m_diagnostics".
3629         (selftest::lexer_test_options::apply): Update for renaming of...
3630         (selftest::lexer_test_options::on_error): ...this, renaming to...
3631         (selftest::lexer_test_options::on_diagnostic): ...this
3632         converting level and reason to enums.
3633         (selftest::test_lexer_string_locations_raw_string_unterminated):
3634         Update for renamings.
3635         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
3636         "reason".
3638 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
3640         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
3641         * config/rs6000/pmmintrin.h: New file.
3643 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
3645         PR tree-optimization/86659
3646         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
3648 2018-10-09  Richard Biener  <rguenther@suse.de>
3650         PR tree-optimization/63155
3651         * tree-ssa-structalias.c: Include tree-ssa.h.
3652         (get_constraint_for_ssa_var): For undefs return nothing_id.
3653         (find_func_aliases): Cleanup PHI handling.
3655 2018-10-09  Richard Biener  <rguenther@suse.de>
3657         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
3658         replacements.
3660 2018-10-09  Martin Liska  <mliska@suse.cz>
3662         * asan.c (asan_emit_stack_protection): If a stack variable
3663         is located in a same file as current function, then emit
3664         line info into variable definition string.
3666 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3668         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
3669         information.
3671 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3673         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
3674         on the thunk.
3676 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3678         PR tree-optimization/86659
3679         * gimple-match.h (struct gimple_match_op): Add reverse field.
3680         (gimple_match_op::set_op): New overloaded method.
3681         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
3682         the REF_REVERSE_STORAGE_ORDER flag on the value.
3683         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
3684         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
3685         
3686 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
3688         PR middle-end/63155
3689         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
3690         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
3692 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
3694         PR target/87517
3695         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
3696         Defined with __builtin_ia32_vfmaddsubpd512_mask.
3698 2018-10-08  Richard Biener  <rguenther@suse.de>
3700         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
3701         cost the same as AVX128 ones.
3703 2018-10-08  Paul Koning  <ni1d@arrl.net>
3705         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
3706         (expand_block_move): New function.
3707         * config/pdp11/pdp11.c (output_block_move): Remove.
3708         (expand_block_move): New function.
3709         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
3710         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
3711         (*movmemhi1): Remove.
3713 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
3715         * config/s390/2827.md: Increase latencies for some FP instructions.
3717 2018-10-08  Richard Biener  <rguenther@suse.de>
3719         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3720         Open a dump scope.
3721         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
3722         * tree-vectorizer.h (dump_stmt_cost): Adjust.
3723         (add_stmt_cost): Dump return value of the hook.
3725 2018-10-08  Richard Biener  <rguenther@suse.de>
3727         PR tree-optimization/63155
3728         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
3729         (ssa_propagation_engine::ssa_propagate): Remove redundant
3730         bitmap bit clearing.
3732 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
3734         PR rtl-optimization/86939
3735         PR rtl-optimization/87479
3736         * ira.h (non_conflicting_reg_copy_p): New prototype.
3737         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
3738         (make_hard_regno_dead): Don't add conflicts for register
3739         ignore_reg_for_conflicts.
3740         (make_object_dead): Likewise.
3741         (non_conflicting_reg_copy_p): New function.
3742         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
3743         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
3744         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
3745         (make_hard_regno_dead): Don't add conflicts for register
3746         ignore_reg_for_conflicts.  Remove special conflict handling of
3747         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
3748         check_pic_pseudo_p and update callers.
3749         (mark_pseudo_dead): Don't add conflicts for register
3750         ignore_reg_for_conflicts.
3751         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
3753 2018-10-05  Andrew Waterman  <andrew@sifive.com>
3754             Jim Wilson  <jimw@sifive.com>
3756         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
3757         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
3758         new pattern using HONOR_SNANS that emits one extra instruction.
3760 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
3762         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
3763         patterns): Merge SI and DI patterns to a GPR pattern.
3764         (unnamed define_insn and define_split for record form of that): Merge
3765         to a single define_insn_and_split pattern.
3767 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
3769         PR c++/56856
3770         * input.c (expand_location_to_spelling_point): Add param "aspect"
3771         and use rather than hardcoding LOCATION_ASPECT_CARET.
3772         (get_substring_ranges_for_loc): Handle the case of a single token
3773         within a macro expansion.
3774         * input.h (expand_location_to_spelling_point): Add "aspect" param,
3775         defaulting to LOCATION_ASPECT_CARET.
3777 2018-10-05  Paul Koning  <ni1d@arrl.net>
3779         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
3780         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
3781         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3782         (pdp11_guard_type): New function.
3784 2018-10-05  Paul Koning  <ni1d@arrl.net>
3786         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
3787         * config/pdp11/pdp11.opt (mfloat32): Remove.
3788         (mfloat64): Remove.
3789         * doc/invoke.texi (pdp11 -mfloat32): Remove:
3790         (pdp11 -mfloat64): Remove.
3792 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3794         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
3795         (*cmp<mode>_cc_i387): Ditto.
3796         (*cmpu<mode>_cc_i387): Ditto.
3797         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3798         * config/i386/i386.c (ix86_expand_fp_compare): Remove
3799         "scratch" argument.
3800         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
3801         Emit x86_sahf_1 pattern.
3802         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
3803         (ix86_expand_carry_flag_compare): Ditto.
3805 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3807         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
3808         to reg_or_0_operand.  Add "C" constraint.
3809         (*cmpxf_cc_i387): Ditto.
3810         (*cmp<mode>_i387): Change operand 2 predicate
3811         to nonimm_or_0_operand.  Add "C" constraint.
3812         (*cmp<mode>_cc_i387): Ditto.
3813         (*cmp<mode>_0_i387): Remove insn pattern.
3814         (*cmp<mode>_0_cc_i387): Ditto.
3816 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3818         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
3819         * config/i386/predicates.md (nonimm_or_0_operand): Rename
3820         from vector_move_operand.  Update all uses.
3822 2018-10-05  Martin Sebor  <msebor@redhat.com>
3824         PR tree-optimization/87490
3825         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
3826         consistently.
3828 2018-10-05  Richard Biener  <rguenther@suse.de>
3830         PR tree-optimization/63155
3831         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
3832         vertical space in dumpfiles.
3833         * tree-ssa-propagate.h
3834         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
3835         * tree-ssa-propagate.c (cfg_blocks_back): New global.
3836         (ssa_edge_worklist_back): Likewise.
3837         (curr_order): Likewise.
3838         (cfg_blocks_get): Remove abstraction.
3839         (cfg_blocks_add): Likewise.
3840         (cfg_blocks_empty_p): Likewise.
3841         (add_ssa_edge): Add to current or next worklist based on
3842         RPO index.
3843         (add_control_edge): Likewise.
3844         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
3845         into ...
3846         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
3847         iteration from CFG and SSA edge worklist so we process
3848         everything in RPO order, prioritizing forward progress
3849         over iteration.
3850         (ssa_prop_init): Allocate new worklists, do not dump
3851         immediate uses.
3852         (ssa_prop_fini): Free new worklists.
3854 2018-10-05  Richard Biener  <rguenther@suse.de>
3856         * tree-core.h (tree_block::abstract_flag): Remove.
3857         (tree_block::block_num): Make full 32bits.
3858         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
3859         * tree.h (BLOCK_ABSTRACT): Remove.
3860         * dwarf2out.c (gen_lexical_block_die): Remove dead code
3861         resulting from BLOCK_ABSTRACT being always false.
3862         (gen_inlined_subroutine_die): Likewise.
3863         (gen_block_die): Likewise.
3864         * tree.c (block_ultimate_origin): Likewise.
3865         * tree-pretty-print.c (dump_block_node): Remove code dealing
3866         with BLOCK_ABSTRACT.
3867         * tree-ssa-live.c (dump_scope_block): Likewise.
3868         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
3869         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
3871 2018-10-05   Richard Biener  <rguenther@suse.de>
3873         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
3874         is asked for initialize mode to the component mode of the
3875         vector type.
3877 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
3879         PR target/87522
3880         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
3881         assembler for -mavx.
3882         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
3884 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
3886         PR target/87509
3887         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
3888         RS6000_BTM_DFP.
3889         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
3890         to be DImode.  When using mffscrn, force the operand to a register.
3892 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
3894         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
3895         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
3896         X87MODEF mode iterator.
3897         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
3898         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
3899         X87MODEF mode iterator.
3901 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
3903         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
3904         -Wno-prio-ctor-dtor.
3906 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3908         * Makefile.in (OBJS): Add opt-problem.o.
3909         * dump-context.h: Include "selftest.h.
3910         (selftest::temp_dump_context): New forward decl.
3911         (class dump_context): Make friend of class
3912         selftest::temp_dump_context.
3913         (dump_context::dump_loc_immediate): New decl.
3914         (class dump_pretty_printer): Move here from dumpfile.c.
3915         (class temp_dump_context): Move to namespace selftest.
3916         (temp_dump_context::temp_dump_context): Add param
3917         "forcibly_enable_dumping".
3918         (selftest::verify_dumped_text):
3919         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
3920         (selftest::verify_item):
3921         (ASSERT_IS_TEXT): Move here from dumpfile.c.
3922         (ASSERT_IS_TREE): Likewise.
3923         (ASSERT_IS_GIMPLE): Likewise.
3924         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
3925         to...
3926         (dump_context::dump_loc_immediate): ...this new function.
3927         (class dump_pretty_printer): Move to dump-context.h.
3928         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
3929         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
3930         (temp_dump_context::temp_dump_context): Move to "selftest"
3931         namespace.  Add param "forcibly_enable_dumping", and use it to
3932         conditionalize the use of m_pp;
3933         (selftest::verify_dumped_text): Make non-static.
3934         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
3935         (selftest::verify_item): Make non-static.
3936         (ASSERT_IS_TEXT): Move to dump-context.h.
3937         (ASSERT_IS_TREE): Likewise.
3938         (ASSERT_IS_GIMPLE): Likewise.
3939         (selftest::test_capture_of_dump_calls): Pass "true" for new
3940         param of temp_dump_context.
3941         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
3942         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
3943         TDF_COMPARE_DEBUG.
3944         * opt-problem.cc: New file.
3945         * opt-problem.h: New file.
3946         * optinfo-emit-json.cc
3947         (selftest::test_building_json_from_dump_calls): Pass "true" for
3948         new param of temp_dump_context.
3949         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
3950         (optinfo::emit_for_opt_problem): New function.
3951         (optinfo::emit): Clarity which emit_item is used.
3952         * optinfo.h (optinfo::get_dump_location): New accessor.
3953         (optinfo::emit_for_opt_problem): New decl.
3954         (optinfo::emit): Make const.
3955         * selftest-run-tests.c (selftest::run_tests): Call
3956         selftest::opt_problem_cc_tests.
3957         * selftest.h (selftest::opt_problem_cc_tests): New decl.
3958         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
3959         bool to opt_result, converting fprintf messages to
3960         opt_result::failure_at calls.  Add "stmt" param for use by the
3961         failure_at calls.
3962         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
3963         (runtime_alias_check_p): Convert return type from bool to
3964         opt_result, converting dump_printf calls to
3965         opt_result::failure_at, using the statement DDR_A for their
3966         location.
3967         (find_data_references_in_stmt): Convert return type from bool to
3968         opt_result, converting "return false" to opt_result::failure_at
3969         with a new message.
3970         * tree-data-ref.h: Include "opt-problem.h".
3971         (dr_analyze_innermost): Convert return type from bool to opt_result,
3972         and add a const gimple * param.
3973         (find_data_references_in_stmt): Convert return type from bool to
3974         opt_result.
3975         (runtime_alias_check_p): Likewise.
3976         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
3977         dr_analyze_innermost.
3978         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
3979         Convert return type from bool to opt_result, adding a message for
3980         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
3981         (vect_analyze_data_ref_dependence): Convert return type from bool
3982         to opt_result.  Change sense of return type from "false"
3983         effectively meaning "no problems" to "false" meaning a problem,
3984         so that "return false" becomes "return opt_result::success".
3985         Convert "return true" calls to opt_result::failure_at, using
3986         the location of statement A rather than vect_location.
3987         (vect_analyze_data_ref_dependences): Convert return type from bool
3988         to opt_result.
3989         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
3990         calls to opt_result::failure_at, using the stmt location rather
3991         than vect_location.
3992         (vect_verify_datarefs_alignment): Convert return type from bool
3993         to opt_result.
3994         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
3995         into multiple more-tightly-scoped copies.
3996         (vect_analyze_data_refs_alignment): Convert return type from bool
3997         to opt_result.
3998         (vect_analyze_data_ref_accesses): Likewise, converting a
3999         "return false" to a "return opt_result::failure_at", adding a
4000         new message.
4001         (vect_prune_runtime_alias_test_list): Convert return type from
4002         bool to opt_result, converting dump_printf_loc to
4003         opt_result::failure_at.  Add a %G to show the pertinent statement,
4004         and use the stmt's location rather than vect_location.
4005         (vect_find_stmt_data_reference): Convert return type from
4006         bool to opt_result, converting dump_printf_loc to
4007         opt_result::failure_at, using stmt's location.
4008         (vect_analyze_data_refs):  Convert return type from bool to
4009         opt_result.  Convert "return false" to "return
4010         opt_result::failure_at", adding messages as needed.
4011         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4012         type from bool to opt_result.
4013         (vect_determine_vf_for_stmt): Likewise.
4014         (vect_determine_vectorization_factor): Likewise, converting
4015         dump_printf_loc to opt_result::failure_at, using location of phi
4016         rather than vect_location.
4017         (vect_analyze_loop_form_1): Convert return type from bool to
4018         opt_result, converting dump_printf_loc calls, retaining the use of
4019         vect_location.
4020         (vect_analyze_loop_form): Convert return type from loop_vec_info
4021         to opt_loop_vec_info.
4022         (vect_analyze_loop_operations): Convert return type from bool to
4023         opt_result, converting dump_printf_loc calls, using the location
4024         of phi/stmt rather than vect_location where available.  Convert
4025         various "return false" to "return opt_result::failure_at" with
4026         "unsupported phi" messages.
4027         (vect_get_datarefs_in_loop): Convert return type from bool to
4028         opt_result.  Add a message for the
4029         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4030         (vect_analyze_loop_2): Convert return type from bool to
4031         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4032         each "goto again;", adding new messages where needed.
4033         Add "unsupported grouped {store|load}" messages.
4034         (vect_analyze_loop): Convert return type from loop_vec_info to
4035         opt_loop_vec_info.
4036         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4037         bool to opt_result.
4038         * tree-vect-stmts.c (process_use): Likewise, converting
4039         dump_printf_loc call and using stmt location, rather than
4040         vect_location.
4041         (vect_mark_stmts_to_be_vectorized): Likeise.
4042         (vect_analyze_stmt): Likewise, adding a %G.
4043         (vect_get_vector_types_for_stmt): Convert return type from bool to
4044         opt_result, converting dump_printf_loc calls and using stmt
4045         location, rather than vect_location.
4046         (vect_get_mask_type_for_stmt): Convert return type from tree to
4047         opt_tree, converting dump_printf_loc calls and using stmt location.
4048         * tree-vectorizer.c: Include "opt-problem.h.
4049         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4050         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4051         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4052         enabled, use it to report at the top level "couldn't vectorize
4053         loop" followed by the problem.
4054         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4055         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4056         to opt_result.
4057         (vect_analyze_stmt): Likewise.
4058         (vect_get_vector_types_for_stmt): Likewise.
4059         (tree vect_get_mask_type_for_stmt): Likewise.
4060         (vect_analyze_data_ref_dependences): Likewise.
4061         (vect_enhance_data_refs_alignment): Likewise.
4062         (vect_analyze_data_refs_alignment): Likewise.
4063         (vect_verify_datarefs_alignment): Likewise.
4064         (vect_analyze_data_ref_accesses): Likewise.
4065         (vect_prune_runtime_alias_test_list): Likewise.
4066         (vect_find_stmt_data_reference): Likewise.
4067         (vect_analyze_data_refs): Likewise.
4068         (vect_analyze_loop): Convert return type from loop_vec_info to
4069         opt_loop_vec_info.
4070         (vect_analyze_loop_form): Likewise.
4071         (vect_analyze_slp): Convert return type from bool to opt_result.
4073 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4075         * doc/invoke.texi (-fopt-info): Document new "internals"
4076         sub-option.
4077         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4078         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4079         MSG_ALL_KINDS.
4080         (optinfo_verbosity_options): Add "internals".
4081         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4082         (dump_context::apply_dump_filter_p): New member function.
4083         (dump_context::dump_loc): Use apply_dump_filter_p rather than
4084         explicitly masking the dump_kind.
4085         (dump_context::begin_scope): Increment the scope depth first.  Use
4086         apply_dump_filter_p rather than explicitly masking the dump_kind.
4087         (dump_context::emit_item): Use apply_dump_filter_p rather than
4088         explicitly masking the dump_kind.
4089         (dump_dec): Likewise.
4090         (dump_hex): Likewise.
4091         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4092         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4093         (opt_info_switch_p): Update handling of default
4094         MSG_OPTIMIZED_LOCATIONS to cope with default of
4095         MSG_PRIORITY_USER_FACING.
4096         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4097         masking the dump_kind.
4098         (selftest::test_capture_of_dump_calls): Update test_dump_context
4099         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4100         than MSG_ALL.  Generalize scope test to be run at all four
4101         combinations of with/without MSG_PRIORITY_USER_FACING and
4102         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4103         for each of the two values.
4104         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4105         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
4106         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4107         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4108         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4109         with MSG_PRIORITY_*.
4110         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4111         dump messages as MSG_PRIORITY_USER_FACING.
4112         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4113         about the interaction with MSG_PRIORITY_*.
4115 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4117         * varasm.c (output_constant): Add new parameter merge_strings.
4118         Make strings properly zero terminated in merge string sections.
4119         (mergeable_string_section): Don't fail if the last char is non-zero.
4120         (assemble_variable_contents): Handle merge string sections.
4121         (assemble_variable): Likewise.
4122         (assemble_constant_contents): Likewise.
4123         (output_constant_def_contents): Likewise.
4124         (output_constructor_array_range,
4125         output_constructor_regular_field): Adjust call to output_constant.
4126         (output_object_block): Adjust call to assemble_constant_contents
4127         and assemble_variable_contents.
4129 2018-10-04  Martin Liska  <mliska@suse.cz>
4131         PR c/87483
4132         * cgraphunit.c (process_function_and_variable_attributes):
4133         Warn about a function with alias attribute and a body.
4135 2018-10-04  Martin Liska  <mliska@suse.cz>
4137         PR ipa/82625
4138         * multiple_target.c (redirect_to_specific_clone): New function.
4139         (ipa_target_clone): Use it.
4140         * tree-inline.c: Fix comment.
4142 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4144         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4145         fields.
4146         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4147         (gcc::dump_manager::register_pass): New member function, adapted
4148         from loop body in gcc::pass_manager::register_pass, adding a
4149         call to update_dfi_for_opt_info.
4150         (gcc::dump_manager::opt_info_enable_passes): Store the
4151         -fopt-info options into the new fields.  Move the loop
4152         bodies into...
4153         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4154         function.
4155         * dumpfile.h (struct opt_pass): New forward decl.
4156         (gcc::dump_manager::register_pass): New decl.
4157         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4158         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4159         "m_optinfo_flags", and "m_optinfo_filename".
4160         * passes.c (gcc::pass_manager::register_pass): Move all of the
4161         dump-handling code to gcc::dump_manager::register_pass.
4163 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4165         PR rtl-optimization/87466
4166         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4167         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4168         * doc/tm.texi: Regenerate.
4169         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4170         * lra-lives.c (process_bb_lives): Likewise.
4171         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4172         Define.
4174 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
4176         * params.c (add_params): Fix initialization.
4178 2018-10-04  Martin Liska  <mliska@suse.cz>
4180         PR gcov-profile/84107
4181         * tree-profile.c (init_ic_make_global_vars):
4182         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4183         Come up with new ic_tuple* variables.  Emit
4184         __gcov_indirect_call{,_topn} variables.
4185         (gimple_gen_ic_profiler): Access the variable
4186         and emit gimple.
4187         (gimple_gen_ic_func_profiler): Access
4188         __gcov_indirect_call.callee field.
4189         (gimple_init_gcov_profiler): Use ptr_type_node.
4190         * value-prof.c (gimple_ic): Use ptr_type_node.
4192 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4194         PR tree-optimization/85787
4195         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4196         into this function and add support for detecting multiple phis.
4197         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4199 2018-10-04  Martin Liska  <mliska@suse.cz>
4201         PR ipa/87491
4202         * ipa-inline.c (inline_to_all_callers_1):
4203         Call ultimate_alias_target for node being inlined.
4205 2018-10-03  Jeff Law  <law@redhat.com>
4207         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4208         target's wchar_t.
4209         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4210         * tree.h (get_typenode_from_name): Prototype.
4212 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4214         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4215         Change operand 2 predicate to nonimmediate_operand.
4216         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4218 2018-10-03  Martin Sebor  <msebor@redhat.com>
4219             Jeff Law  <law@redhat.com>
4221         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4222         initialize it.
4223         (get_string_length): Detect unterminated arrays.
4224         (format_string): Same.
4225         (format_directive): Warn about unterminated arrays.
4226         (handle_gimple_call): Mark statements with no_warning as needed.
4228 2018-10-03  Jim Wilson  <jimw@sifive.com>
4230         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4231         also define __riscv_abi_rve.  Delete trailing white space.
4233 2018-10-03  Paul Koning  <ni1d@arrl.net>
4235         Enable LRA register allocator for PDP11.
4236         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4237         (R): Likewise.
4238         (D): Likewise.
4239         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4240         * config/pdp11/pdp11.opt (-mlra): New option.
4241         * doc/invoke.texi (PDP-11 Options): Document -mlra.
4243 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4245         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4246         (*<absneg:code>extend<mode>xf2): Ditto.
4248 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4250         PR tree-optimization/87415
4251         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4252         precision fields.
4254 2018-10-02  Jeff Law  <law@redhat.com>
4256         * gimple-fold.c (get_range_strlen): Only set *nonstr when
4257         an unterminated string is discovered.  Bubble up range
4258         even for unterminated strings.
4259         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4260         indicates the string was not terminated via NONSTR.
4262 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4264         * tree-vrp.c (extract_range_from_unary_expr): Special case all
4265         pointer conversions.
4266         Do not do anything special for anti-ranges.
4268 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
4270         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4271         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4273 2018-10-03  Martin Liska  <mliska@suse.cz>
4275         PR gcov-profile/86109
4276         * coverage.c (coverage_begin_function): Do not
4277         mark lambdas as artificial.
4278         * tree-core.h (struct GTY): Remove tm_clone_flag
4279         and introduce new lambda_function.
4280         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4282 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4284         PR target/87474
4285         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4286         P8_VECTOR and VSX are enabled.
4288 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4290         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4291         0x3907 as CPU model number.
4293 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4295         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4296         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4297         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4298         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4299         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4300         * config/s390/s390.md: Likewise. Rename also the cpu attribute
4301         value from arch12 to z14.
4303 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4305         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4306         (isinfxf2): Ditto.
4307         (isinf<mode>2): Ditto.
4309 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4311         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4312         before emitting fxam.  Perform calculations in XFmode.
4314 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4316         * match.pd (((X /[ex] A) +- B) * A): New transformation.
4318 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4320         PR middle-end/87319
4321         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
4322         * tree.c (signed_or_unsigned_type_for): Handle complex.
4324 2018-10-02  Jeff Law  <law@redhat.com>
4326         * gimple-fold.c (get_range_strlen): Remove dead code.
4328 2018-10-02  Martin Sebor  <msebor@redhat.com>
4329             Jeff Law  <law@redhat.com>
4331         * builtins.c (unterminated_array): Add new arguments.
4332         If argument is not terminated, bubble up size and exact
4333         state to callers.
4334         (expand_builtin_strnlen): Detect, avoid expanding
4335         and diagnose unterminated arrays.
4336         (c_strlen): Fill in offset of start of unterminated strings.
4337         * builtins.h (unterminated_array): Update prototype.
4339 2018-10-02  Richard Biener  <rguenther@suse.de>
4341         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
4342         of haddv4df, first reduce to SSE width and exploit the fact
4343         that we only need element zero with the reduction result.
4344         (reduc_plus_scal_v2df): Likewise.
4346 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4348         * dojump.h (do_jump): Delete.
4349         (do_jump_1): Likewise.
4350         (split_comparison): Move around.
4351         * dojump.c (do_jump): Make static.
4352         (do_jump_1): Likewise.
4353         (jumpifnot): Move around.
4354         (jumpifnot_1): Likewise.
4355         (jumpif): Likewise.
4356         (jumpif_1): Likewise.
4357         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
4359 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4361         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
4362         insns in the delay slot and add_insn_after for the jump insn.
4364 2018-10-02  Richard Biener  <rguenther@suse.de>
4366         * tree-inline.c (expand_call_inline): Use the location of
4367         the callee declaration for the inline-entry marker.
4368         * final.c (notice_source_line): Remove special-casing of
4369         NOTE_INSN_INLINE_ENTRY.
4371 2018-10-01  Carl Love  <cel@us.ibm.com>
4373         PR 69431
4374         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
4375         (__builtin_mtfsb0): New.
4376         (__builtin_mtfsb1): New.
4377         ( __builtin_set_fpscr_rn): New.
4378         (__builtin_set_fpscr_drn): New.
4379         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
4380         (rs6000_expand_set_fpscr_rn_builtin): Add.
4381         (rs6000_expand_set_fpscr_drn_builtin): Add.
4382         (rs6000_expand_builtin): Add case statement entries for
4383         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
4384         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
4385         RS6000_BUILTIN_MFFSL.
4386         (rs6000_init_builtins): Add ftype initialization and def_builtin
4387         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
4388         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
4389         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
4390         rs6000_mffscdrn): Add define_insn.
4391         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
4392         * doc/extend.texi: Add documentation for the builtins.
4394 2018-10-01  Richard Biener  <rguenther@suse.de>
4396         PR tree-optimization/87465
4397         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
4398         causing branch miscounts.
4400 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4402         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
4403         aarch64_option_default_param):  New.
4404         (params.h): Include.
4405         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
4406         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
4407         stack-clash protection validation code.
4409 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4411         * params.c (validate_param): New.
4412         (add_params): Use it.
4413         (set_param_value): Refactor param validation into validate_param.
4414         (diagnostic.h): Include.
4415         * diagnostic.h (diagnostic_ready_p): New.
4417 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4419         * params.c (set_param_value):
4420         Add index of parameter being validated.
4421         * common/common-target.def (option_validate_param): New.
4422         * common/common-targhooks.h (default_option_validate_param): New.
4423         * common/common-targhooks.c (default_option_validate_param): New.
4424         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
4425         * doc/tm.texi: Regenerate.
4427 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4429         PR target/86486
4430         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4431         Add validation for stack-clash parameters and set defaults.
4433 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4435         PR target/86486
4436         * configure.ac: Add stack-clash-protection-guard-size.
4437         * doc/install.texi: Document it.
4438         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
4439         * params.def: Update comment for guard-size.
4440         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
4441         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
4442         * configure: Regenerate.
4444 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4446         PR target/86486
4447         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
4448         STACK_DYNAMIC_OFFSET): New.
4449         * config/aarch64/aarch64.c (aarch64_layout_frame):
4450         Update outgoing args size.
4451         (aarch64_stack_clash_protection_alloca_probe_range,
4452         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4454 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4456         PR target/86486
4457         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
4458         probe ranges.
4459         * target.def (stack_clash_protection_alloca_probe_range): New.
4460         (stack_clash_protection_final_dynamic_probe): Remove.
4461         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
4462         (default_stack_clash_protection_final_dynamic_probe): Remove.
4463         * targhooks.c: Likewise.
4464         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4465         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
4466         * doc/tm.texi: Regenerate.
4468 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4470         PR target/86486
4471         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
4472         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
4473         aarch64_clamp_to_uimm12_shift): New.
4474         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
4475         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
4477 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4479         PR target/86486
4480         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
4482 2018-10-01  Jeff Law  <law@redhat.com>
4483             Richard Sandiford <richard.sandiford@linaro.org>
4484             Tamar Christina  <tamar.christina@arm.com>
4486         PR target/86486
4487         * config/aarch64/aarch64.md
4488         (probe_stack_range): Add k (SP) constraint.
4489         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
4490         STACK_CLASH_MAX_UNROLL_PAGES): New.
4491         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
4492         stack probes for stack clash.
4493         (aarch64_allocate_and_probe_stack_space): New.
4494         (aarch64_expand_prologue): Use it.
4495         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
4496         (aarch64_sub_sp): Add emit_move_imm optional param.
4498 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
4500         PR tree-optimization/87261
4501         * match.pd: Remove trailing whitespace.
4502         Add (x & y) | ~(x | y) -> ~(x ^ y),
4503         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
4505 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4507         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
4508         constraints.
4509         (ashlsi3_insn): Update instruction constraints.
4510         (ashrsi3_insn): Likewise.
4511         (rotrsi3): Likewise.
4512         (add_shift): Likewise.
4513         * config/arc/constraints.md (Csz): New 32 bit constraint. It
4514         avoids placing in the limm field small constants which, otherwise,
4515         could end into a small instruction.
4517 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4519         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
4520         destination register is not odd-even.
4521         (umaddsidi4_split): Likewise.
4523 2018-10-01  Richard Biener  <rguenther@suse.de>
4525         * tree-inline.c (expand_call_inline): Store origin of fn
4526         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
4527         * tree.c (block_ultimate_origin): Simplify and do some
4528         checking.
4530 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
4532         * config/i386/mmx.md (EMMS): New int iterator.
4533         (emms): New int attribute.
4534         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
4535         EMMS int iterator.  Explicitly declare clobbers.
4536         (mmx_emms): Remove expander.
4537         (mmx_femms): Ditto.
4538         * config/i386/predicates.md (emms_operation): Remove predicate.
4539         (vzeroall_pattern): New predicate.
4540         (vzeroupper_pattern): Rename from vzeroupper_operation.
4541         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
4542         vzeroupper_pattern and vzeroall_pattern predicates.
4544 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
4546         PR rtl-optimization/86939
4547         * ira-lives.c (make_hard_regno_born): Rename from this...
4548         (make_hard_regno_live): ... to this.  Remove update to conflict
4549         information.  Update function comment.
4550         (make_hard_regno_dead): Add conflict information update.  Update
4551         function comment.
4552         (make_object_born): Rename from this...
4553         (make_object_live): ... to this.  Remove update to conflict information.
4554         Update function comment.
4555         (make_object_dead):  Add conflict information update.  Update function
4556         comment.
4557         (mark_pseudo_regno_live): Call make_object_live.
4558         (mark_pseudo_regno_subword_live): Likewise.
4559         (mark_hard_reg_dead): Update function comment.
4560         (mark_hard_reg_live): Call make_hard_regno_live.
4561         (process_bb_node_lives): Likewise.
4562         * lra-lives.c (make_hard_regno_born): Rename from this...
4563         (make_hard_regno_live): ... to this.  Remove update to conflict
4564         information.  Remove now uneeded check_pic_pseudo_p argument.
4565         Update function comment.
4566         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
4567         to conflict information.  Update function comment.
4568         (mark_pseudo_live): Remove update to conflict information.  Update
4569         function comment.
4570         (mark_pseudo_dead): Add conflict information update.
4571         (mark_regno_live): Call make_hard_regno_live.
4572         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
4573         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
4575 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
4577         PR target/87370
4578         * config/i386/i386.c (construct_container): Use TImode for
4579         BLKmode values in 2 integer registers.
4581 2018-09-29  Jeff Law  <law@redhat.com>
4583         * builtins.c (unterminated_array): Pass in c_strlen_data * to
4584         c_strlen rather than just a tree *.
4585         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
4586         Update recursive calls appropriately.  If caller did not provide a
4587         suitable data pointer, create a local one.  When a non-terminated
4588         string is discovered, bubble up information about the string via the
4589         c_strlen_data object.
4590         * builtins.h (c_strlen): Update prototype.
4591         (c_strlen_data): New structure.
4592         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
4593         For a type 2 call, if c_strlen indicates a non-terminated string
4594         use the length of the non-terminated string.
4595         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
4597 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
4599         PR target/87467
4600         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
4601         __m512d type for __A argument rather than __m512.
4603 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
4605         * match.pd (simple_comparison): Don't optimize if either operand is
4606         a function pointer when target needs function pointer canonicalization.
4608 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4610         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
4611         power5 .. power9 to remove indirection.
4612         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
4613         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
4614         ASM_CPU_476_SPEC): Delete.
4615         (ASM_CPU_SPEC): Adjust.
4616         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
4617         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
4619 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4621         * config.in: Delete HAVE_AS_DCI.
4622         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
4623         * config/rs6000/rs6000.h: Ditto.
4624         * configure.ac: Delete HAVE_AS_DCI.
4625         * configure: Regenerate.
4627 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4629         * config.in (HAVE_AS_LWSYNC): Delete.
4630         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4631         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
4632         do it as a .long .
4633         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4634         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
4635         as a .long .
4636         * configure.ac: Delete HAVE_AS_LWSYNC.
4637         * configure: Regenerate.
4639 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4640             Pierre-Marie de Rodat  <derodat@adacore.com>
4642         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
4643         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
4644         (cgraph_node::create_thunk): Add indirect_offset parameter.
4645         (thunk_adjust): Likewise.
4646         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
4647         and initialize the corresponding field with it.
4648         (cgraph_node::dump): Dump indirect_offset field.
4649         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
4650         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
4651         (thunk_adjust): Add indirect_offset parameter and deal with it.
4652         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
4653         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
4654         or if the thunk is external or local.  Fix formatting.  Do not chain
4655         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
4656         if any, in the GIMPLE representation.
4657         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
4658         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
4659         (input_node): Read indirect_offset field.
4660         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
4661         call to thunk_adjust.
4662         * tree-nested.c (struct nesting_info): Add thunk_p field.
4663         (create_nesting_tree): Set it.
4664         (convert_all_function_calls): Copy static chain from targets to thunks.
4665         (finalize_nesting_tree_1): Return early for thunks.
4666         (unnest_nesting_tree_1): Do not finalize thunks.
4667         (gimplify_all_functions): Do not gimplify thunks.
4669 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4671         * opt-suggestions.c (option_proposer::build_option_suggestions):
4672         Release "option_values".
4674 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4676         * coverage.c (get_coverage_counts): Convert problem-reporting dump
4677         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
4678         * dumpfile.c (kind_as_string): New function.
4679         (dump_loc): Rather than a hardcoded prefix of "note: ", use
4680         kind_as_string to vary the prefix based on dump_kind.
4681         (selftest::test_capture_of_dump_calls): Update for above.
4683 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4685         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
4686         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
4688 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4690         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
4691         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
4692         INVALID_REGNUM instead of FPSR_REG.
4693         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
4694         * config/i386/i386.md: Update comment of FP compares.
4695         (fldenv): Do not clobber FPSR_REG.
4697 2018-09-28  Richard Biener  <rguenther@suse.de>
4699         * tree.h (BLOCK_ORIGIN): New.
4700         * omp-expand.c (grid_expand_target_grid_body): Assign
4701         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
4702         * tree-inline.c (remap_block): Likewise.
4703         * auto-profile.c (get_function_decl_from_block): Simplify
4704         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
4705         * langhooks.c (lhd_print_error_function): Likewise.
4706         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
4707         Likewise.
4708         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4709         * tree.c (block_nonartificial_location): Likewise.
4710         (block_ultimate_origin): Likewise.
4711         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
4712         no longer needed LTO case.
4714 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
4715             Jan Hubicka  <jh@suse.cz>
4716             Martin Jambor  <mjambor@suse.cz>
4718         * simplify-rtx.c (simplify_merge_mask): New function.
4719         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
4720         same masks are used in op1 or op2.
4721         (test_vec_merge): New function.
4722         (test_vector_ops): Call test_vec_merge.
4724 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4726         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
4727         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
4728         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
4729         (cypress_costs): Set it.
4730         (supersparc_costs): Likewise.
4731         (hypersparc_costs): Likewise.
4732         (leon_cost): Likewise.
4733         (leon3_costs): Likewise.
4734         (sparclet_costs): Likewise.
4735         (ultrasparc_costs): Likewise.
4736         (ultrasparc_costs): Likewise.
4737         (niagara_costs): Likewise.
4738         (niagara2_costs): Likewise.
4739         (niagara3_costs): Likewise.
4740         (niagara4_costs): Likewise.
4741         (niagara7_costs): Likewise.
4742         (m8_costs): Likewise.
4743         (TARGET_CAN_FOLLOW_JUMP): Define.
4744         (pass_work_around_errata::gate): Minor tweak.
4745         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
4746         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
4747         Automaitcally clear MASK_FSMULD mask for V7 processors.
4748         (sparc_can_follow_jump): New static function.
4749         (output_ubranch): Deal with CROSSING_JUMP_P.
4750         (sparc_use_sched_lookahead): Rewrite using switch statement.
4751         (sparc_issue_rate): Reorder.
4752         (sparc_branch_cost): New function.
4754 2018-09-27  Martin Sebor  <msebor@redhat.com>
4756         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
4757         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
4758         (int_fits_type_p): Same.
4760 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
4762         * config/i386/i386.md (FPCR_REG): Remove.
4763         (UNSPEC_FLDCW): Remove.
4764         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
4765         (x86_fldcw_1): Remove insn pattern.
4766         (fnstenv): Do not clobber FPCR_REG.
4767         (fldenv): Ditto.
4768         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
4769         (CALL_USED_REGISTERS): Ditto.
4770         (REG_ALLOC_ORDER): Ditto.
4771         (REG_CLASS_CONTENTS): Ditto.
4772         (HI_REGISTER_NAMES): Ditto.
4773         (ADDITIONAL_REGISTER_NAMES): Use defines instead
4774         of numerical constants.
4775         * config/i386/i386.c (regclass_map): Remove fpsr register.
4776         (dbx_register_map): Ditto.
4777         (dbx64_register_map): Ditto.
4778         (svr4_dbx_register_map): Ditto.
4779         (print_reg): Do not handle FPCR_REG.
4781 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
4783         PR target/87149
4784         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
4785         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
4786         Delete, always treat as true.
4787         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4788         Ditto.  Simplify remaining code.
4789         * config/powerpcspe/powerpcspe.h: Ditto.
4790         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
4791         Simplify remaining code.
4792         (rs6000_expand_builtin): Ditto.
4793         * config/rs6000/rs6000.h: Ditto.
4794         * configure.ac: Ditto.
4795         * configure: Regenerate.
4797 2018-09-27  Martin Liska  <mliska@suse.cz>
4799         * coverage.c (get_coverage_counts): Revert the formatting
4800         of missing profile opt info.
4802 2018-09-27  Richard Biener  <rguenther@suse.de>
4804         PR debug/37801
4805         PR debug/87440
4806         * dwarf2out.c (set_block_origin_self): Do not mark outermost
4807         block as we do not output that.
4808         (gen_inlined_subroutine_die): Elide the originally outermost
4809         block, matching what we do for concrete instances.
4810         (decls_for_scope): Add parameter specifying whether to recurse
4811         to subblocks.
4813 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
4814             Tom de Vries  <tom@codesourcery.com>
4816         PR 82089
4818         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
4819         STORE_FLAG_VALUE == 1.
4821 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
4823         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
4824         constant definitions.
4825         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
4826         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
4827         ("speculation_barrier"): New expander definition.
4829 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
4831         PR gcov-profile/86957
4832         * common.opt: New warning option -Wmissing-profile.
4833         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
4834         * doc/invoke.texi: Document -Wmissing-profile.
4836 2018-09-26  Jim Wilson  <jimw@sifive.com>
4838         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
4839         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
4841 2018-09-26  Martin Sebor  <msebor@redhat.com>
4843         * tree.c (zerop): Change return type to bool.
4844         (integer_zerop, integer_onep, integer_each_onep): Same.
4845         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4846         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4847         (real_onep, real_minus_onep, chain_index): Same.
4848         (print_type_hash_statistics, type_list_equal): Same.
4849         * tree.h (zerop): Same.
4850         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
4851         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4852         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4853         (real_onep, real_minus_onep, chain_index): Same.
4854         (print_type_hash_statistics, type_list_equal): Same.
4856 2018-09-26  Jim Wilson  <jimw@sifive.com>
4858         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
4860 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
4862         PR target/87414
4863         * config/i386/i386.c: Include debug.h and dwarf2out.h.
4864         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
4865         call.
4867 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
4869         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
4871 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
4873         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
4874         and FP_SECOND_SSE_REGS.
4875         (REG_CLASS_NAMES): Ditto.
4876         (REG_CLASS_CONTENTS): Ditto.
4877         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
4878         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
4879         (ix86_preferred_output_reload_class): Ditto.
4880         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
4881         clobber constraint to "=&f".
4882         (fix_truncdi_i387): Ditto.
4883         (lrintxfdi2): Ditto.
4884         (fistdi2_<rounding>): Ditto.
4885         (fpremxf4_i387): Change "=u" constraint to "=f".
4886         (fprem1xf4_i387): Ditto.
4887         (sincosxf3): Ditto.
4888         (fptanxf4_i387): Ditto.
4889         (fxtractxf3_i387): Ditto.
4890         (fscalexf4_i387): Ditto.
4891         (atan2xf3): Change "u" constraint to "f".
4892         (fyl2xxf3_i387): Ditto.
4893         (fyl2xp1xf3_i387): Ditto.
4895 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
4897         PR target/87439
4898         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
4899         for removed I387_MASK_PM entity.
4902 2018-09-26  Jeff Law  <law@redhat.com>
4903         Revert
4904         2018-09-26  Alexey Neyman  <stilor@att.net>
4906         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4907         headers are no longer pulled in by <isl/val.h>.
4909 2018-09-26  Richard Biener  <rguenther@suse.de>
4911         PR debug/87443
4912         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
4913         or concrete instance DIE to the tree.  Create abstract origin
4914         attributes also for concrete instances.
4916 2018-09-26  Alexey Neyman  <stilor@att.net>
4918         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4919         headers are no longer pulled in by <isl/val.h>.
4921 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
4923         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
4924         Use new helper functions.
4925         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
4926         Use new helper functions.
4927         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
4928         aarch_mm_needs_release): New declarations.
4929         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
4930         aarch_mm_needs_release): New.
4932 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
4934         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
4935         (arm32_output_mi_thunk): Deal with long calls.
4937 2018-09-26  Richard Biener  <rguenther@suse.de>
4939         PR debug/87428
4940         PR debug/87362
4941         * tree-inline.c (expand_call_inline): When the location
4942         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
4943         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
4944         the inserted BLOCK to make inlined_function_outer_scope_p
4945         recognize it.
4946         * dwarf2out.c (add_call_src_coords_attributes): Do not add
4947         coords for reserved locations.
4949 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4951         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
4952         (*call_indirect_nonlocal_sysv<mode>): Ditto.
4953         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
4954         (*sibcall_nonlocal_sysv<mode>): Ditto.
4955         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
4956         (<bd>_<mode>): Ditto.
4957         (<bd>tf_<mode>): Ditto.
4959 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4961         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
4962         control string as a list of templates instead of as C code.
4963         (*altivec_movti): Ditto.
4964         * config/rs6000/darwin.md (movdf_low_di): Ditto.
4966 2018-09-25  Jim Wilson  <jimw@sifive.com>
4968         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
4969         when target symbol is weak.
4971 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4973         PR c/87387
4974         * builtins.c (unterminated_array): Simplify.
4975         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
4976         where pointer arithmetic is safe.
4978 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4980         PR target/86987
4981         * config/rs6000/altivec.md (altivec_vspltb): Use
4982         const_0_to_15_operand instead of u5bit_cint_operand.
4983         (*altivec_vspltb_internal): Ditto.
4984         (altivec_vspltb_direct): Ditto.
4985         (altivec_vsplth): Use const_0_to_7_operand instead of
4986         u5bit_cint_operand.
4987         (*altivec_vsplth_internal): Ditto.
4988         (altivec_vsplth_direct): Ditto.
4989         (altivec_vspltw): Use const_0_to_3_operand instead of
4990         u5bit_cint_operand.
4991         (*altivec_vspltw_internal): Ditto.
4992         (altivec_vspltw_direct): Ditto.
4993         (altivec_vspltsf): Ditto.
4994         (*altivec_vspltsf_internal): Ditto.
4995         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
4996         various splats with the proper size immediate.  Reorder the various
4997         cases by ascending size of immediate, and put all such together.
4999 2018-09-25  Richard Biener  <rguenther@suse.de>
5001         PR debug/83941
5002         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5003         GC-ification.
5004         (maybe_create_die_with_external_ref): Do not create
5005         DW_TAG_imported_unit here.
5006         (add_abstract_origin_attribute): Handle external BLOCK refs.
5007         (dwarf2out_abstract_function): Simplify LTO case.
5008         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5009         rather than using maybe_create_die_with_external_ref.
5011 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5013         PR target/71278
5014         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5015         (frndintxf2_mask_pm_i387): Ditto.
5016         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5017         Enable for !flag_trapping_math.
5018         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5019         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5020         Change operand 1 predicate to nonimmediate_operand.
5021         (attr "i387_cw"): Remove mask_pm.
5022         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5023         (enum ix86_entity): Remove I387_MASK_PM.
5024         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5025         handle I387_MASK_PM.
5026         (ix86_mode_needed): Ditto.
5027         (ix86_mode_after): Ditto.
5028         (ix86_mode_entry): Ditto.
5029         (ix86_mode_exit): Ditto.
5030         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5032 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5034         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5035         to_update_switch_stmts to vNULL instead of calling create on them
5036         immediately.
5038 2018-09-25  Richard Biener  <rguenther@suse.de>
5040         PR tree-optimization/87402
5041         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5042         (visit_phi): Re-instantiate handling of supposed to be VARYING
5043         but non-VARYING backedge value.
5045 2018-09-25  Richard Biener  <rguenther@suse.de>
5047         PR debug/83941
5048         * dwarf2out.c (struct sym_off_pair): New.
5049         (external_die_map): New global.
5050         (lookup_decl_die): When in LTO create DIEs lazily from the
5051         external_die_map.
5052         (lookup_block_die): New function, create DIEs lazily in LTO.
5053         (equate_block_to_die): New function.
5054         (dwarf2out_die_ref_for_decl): During WPA get the association
5055         from the external DIE map.
5056         (dwarf2out_register_external_die): Record mapping into the
5057         external DIE map.
5058         (maybe_create_die_with_external_ref): New function split out from
5059         DIE generation part of old dwarf2out_register_external_die.
5060         (add_abstract_origin_attribute): Do not return the DIE.  When
5061         in LTO reference externals directly.
5062         (dwarf2out_abstract_function): When in LTO ignore calls for
5063         decls with external DIEs (already present abstract instances).
5064         (gen_call_site_die): Adjust.
5065         (add_high_low_attributes): Likewise.
5066         (gen_lexical_block_die): Likewise.
5067         (gen_inlined_subroutine_die): Likewie.
5068         (gen_block_die): Likewise.
5069         (dwarf2out_inline_entry): Likewise.
5070         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5071         DIEs.
5073 2018-09-25  Martin Liska  <mliska@suse.cz>
5075         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5076         integers and not by a float value.
5078 2018-09-25  Martin Liska  <mliska@suse.cz>
5080         PR fortran/87394
5081         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5082         instead of NULL.
5083         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5084         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5085         * godump.c (go_format_type): Remove extra parenthesis.
5087 2018-09-25  Martin Liska  <mliska@suse.cz>
5089         * alias.c (set_dest_equal_p): Remove unused function.
5090         * config/i386/i386.c (def_builtin_pure2): Likewise.
5091         * diagnostic-show-locus.c (class layout): Remove
5092         unused field.
5093         (layout::layout): Likewise here.
5094         * dump-context.h (class temp_dump_context): Likewise.
5095         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5096         (add_AT_loclistsptr): Likewise.
5097         (add_AT_offset): Likewise.
5098         (get_AT_hi_pc): Likewise.
5099         (is_comdat_die): Likewise.
5100         (type_is_enum): Likewise.
5101         (ceiling): Likewise.
5102         (add_AT_vms_delta): Likewise.
5103         (is_class_die): Likewise.
5104         * edit-context.c (class line_event): Remove unused field.
5105         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5106         unused function.
5107         * ipa-cp.c (ipa_get_vr_lat): Likewise.
5108         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5109         (ok_for_base_p_nonstrict): Likewise.
5110         * tree-chrec.c (is_not_constant_evolution): Likewise.
5111         (chrec_fold_poly_cst): Likewise.
5112         * tree-if-conv.c (has_pred_critical_p): Likewise.
5113         * tree-ssa-coalesce.c (print_exprs): Likewise.
5114         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5115         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5116         * tree-vrp.c (value_ranges_intersect_p): Likewise.
5117         (value_range_nonnegative_p): Likewise.
5119 2018-09-25  Martin Liska  <mliska@suse.cz>
5121         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5122         Do not handle "GNU Pascal".
5123         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5124         Likewise.
5125         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5126         from documentation. Likewise.
5127         * dbxout.c (dbxout_range_type): Likewise.
5128         * doc/cpp.texi: Likewise.
5129         * doc/extend.texi: Likewise.
5130         * doc/frontends.texi: Likewise.
5131         * doc/invoke.texi: Remove Pascal entry.
5132         * tree.def (CLEANUP_POINT_EXPR): Likewise.
5133         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5135 2018-09-25  Martin Liska  <mliska@suse.cz>
5137         PR middle-end/86078
5138         * doc/invoke.texi: Document all parameters and remove default
5139         of the parameters.
5141 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5143         PR bootstrap/87417
5144         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5145         contains HOST_WIDE_INTs when computing its size.
5147 2018-09-24  Jim Wilson  <jimw@sifive.com>
5149         PR target/87391
5150         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5151         not TARGET_RVE.
5152         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5154 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5156         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5157         access prev before checking it for NULLness in the
5158         AARCH64_FUSE_CMP_BRANCH case.
5160 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5162         PR target/82699
5163         * config/i386/i386.c (rest_of_insert_endbranch): Set
5164         endbr_queued_at_entrance to true and don't insert ENDBR if
5165         x86_function_profiler will be called.
5166         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5167         is true.
5168         * config/i386/i386.h (machine_function): Add
5169         endbr_queued_at_entrance.
5171 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5173         * genattrtab.c (mk_attr_alt): Use alternative_mask.
5174         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5175         types.
5176         (check_attr_test): Use alternative_mask.
5177         (get_attr_value): Likewise.
5178         (compute_alternative_mask): Use alternative_mask and XWINT.
5179         (make_alternative_compare): Use alternative_mask.
5180         (attr_alt_subset_p): Use XWINT.
5181         (attr_alt_subset_of_compl_p): Likewise.
5182         (attr_alt_intersection): Use alternative_mask and XWINT.
5183         (attr_alt_union): Likewise.
5184         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5185         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5186         (simplify_test_exp): Use alternative_mask and XWINT.
5187         (write_test_expr): Use alternative_mask and XWINT, adjust bit
5188         number calculation to support 64 bits.  Generate code that
5189         checks 64-bit masks.
5190         (main): Use alternative_mask.
5191         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5193 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5195         PR target/80080
5196         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5197         RETURN+USE when returning via %r14.
5199 2018-09-24  Martin Liska  <mliska@suse.cz>
5201         * gcov.c (output_lines): Print colorization legend
5202         for both flag_use_colors and flag_use_hotness_colors.
5203         Reword the help.
5205 2018-09-24  Martin Liska  <mliska@suse.cz>
5207         * coverage.c (get_coverage_counts): Use warning_at
5208         with current_function_decl location. Use %qD in warning
5209         message.
5211 2018-09-24  Martin Liska  <mliska@suse.cz>
5213         * memory-block.h (memory_block_pool::release): Annotate with
5214         valgrind that the memory is not accessible.
5216 2018-09-24  Martin Liska  <mliska@suse.cz>
5218         PR sanitizer/85774
5219         * asan.c: Make asan_handled_variables extern.
5220         * asan.h: Likewise.
5221         * cfgexpand.c (expand_stack_vars): Make sure
5222         a representative is unpoison if another
5223         variable in the partition is handled by
5224         use-after-scope sanitization.
5226 2018-09-24  Richard Biener  <rguenther@suse.de>
5228         PR tree-optimization/63155
5229         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5230         the worklist when the edge of the respective argument isn't
5231         executable.
5233 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5235         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5236         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5237         (MASK_CLASS_P): Update for rename.
5238         (MAYBE_MASK_CLASS_P): Ditto.
5239         (REG_CLASS_NAMES): Update.
5240         (REG_CLASS_CONTENT): Update.
5241         * config/i386/i386.c (regclass_map): Update for MASK_REG
5242         and ALL_MASK_REGS rename.
5243         * config/i386/constraints.md (Yk): Update for rename.
5244         (k): Ditto.
5246 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5248         * config/i386/i386.h (enum reg_class): Remove
5249         EVEX_SSE_REGS and MOD4_SSE_REGS.
5250         (REG_CLASS_NAMES): Update.
5251         (REG_CLASS_CONTENT): Update.
5252         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5253         registers as ALL_SSE_REGS.
5254         (ix86_additional_allocno_class_p): Remove.
5255         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5256         (ix86_register_priority): Lower priority of EVEX SSE registers.
5257         Use IN_RANGE macro where appropriate.
5258         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5259         AVX-5124VNNIW checks.
5260         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5261         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5262         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5263         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5264         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5265         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5266         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5267         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5268         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5269         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5270         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5271         * config/i386/constraints.md (Yh): Remove.
5273 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5275         * config/i386/i386.c (regclass_map): Declare integer REX registers
5276         as GENERAL_REGS.
5278 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
5280         * doc/service.texi (Service): Switch the fsf.org link to https.
5282 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5284         PR target/86798
5285         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5286         Define to speculation_safe_value_not_needed.
5288 2018-09-21  Florian Weimer  <fweimer@redhat.com>
5290         PR middle-end/81035
5291         * doc/extend.texi (Common Function Attributes): Mention that
5292         noreturn suppresses tail call optimization.
5294 2018-09-21  Jeff Law  <law@redhat.com>
5296         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5297         vr_values::cleanup_edges_and_switches.
5298         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5299         vr_values class.
5300         (identify_jump_threads): Remove EDGE_IGNORE handling.
5301         (execute_vrp): Move handling of to_remove_edges and
5302         to_update_switch_stmts into vr_values class member functions.
5303         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5304         (to_update_switch_stmts): Likewise.
5305         * vr-values.c: Include cfghooks.h.
5306         (vr_values::vr_values): Initialize to_remove_edges and
5307         to_update_switch_stmts.
5308         (vr_values::~vr_values): Verify to_remove_edges and
5309         to_update_switch_stmts are empty.
5310         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5311         (vr_values::cleanup_edges_and_switches): New member function.
5312         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5313         function.  Add new data members.
5315 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
5317         PR tree-optimization/87309
5318         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
5319         calls with pflags and alt_flags.
5320         (selftest::test_capture_of_dump_calls): Add test of interaction of
5321         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
5323 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5325         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
5327 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5329         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5331 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5333         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
5334         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
5335         (VXWORKS_LIBS_RTP): Minor reordering.
5337 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5339         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
5340         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5342 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5344         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
5345         (PTRDIFF_TYPE): Likewise.
5347 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5349         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
5350         triplet, similar to support for VxWorks7.
5351         * config/vxworks-dummy.h: Provide a default definition
5352         of TARGET_VXWORKS64 to 0.
5354 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5356         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
5357         * config/vxworks-dummy.h: here.
5359 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5361         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
5363 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
5364             Bo Zhou  <zbo.zhou@hisilicon.com>
5366         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5367         * config/aarch64/aarch64-tune.md: Regenerated.
5368         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
5369         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
5370         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
5372 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
5373             Julian Brown  <julian@codesourcery.com>
5375         * builtins.c (get_builtin_sync_mem): Handle address spaces.
5377 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5379         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
5380         if the call takes a static chain.
5382 2018-09-21  Martin Liska  <mliska@suse.cz>
5384         * auto-profile.c (autofdo_source_profile::read): Do not
5385         set sum_all.
5386         (read_profile): Do not add working sets.
5387         (read_autofdo_file): Remove sum_all.
5388         (afdo_callsite_hot_enough_for_early_inline): Remove const
5389         qualifier.
5390         * coverage.c (struct counts_entry): Remove gcov_summary.
5391         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
5392         do not support GCOV_TAG_PROGRAM_SUMMARY.
5393         (get_coverage_counts): Remove summary and expected
5394         arguments.
5395         * coverage.h (get_coverage_counts): Likewise.
5396         * doc/gcov-dump.texi: Remove -w option.
5397         * gcov-dump.c (dump_working_sets): Remove.
5398         (main): Do not support '-w' option.
5399         (print_usage): Likewise.
5400         (tag_summary): Likewise.
5401         * gcov-io.c (gcov_write_summary): Do not dump
5402         histogram.
5403         (gcov_read_summary): Likewise.
5404         (gcov_histo_index): Remove.
5405         (gcov_histogram_merge): Likewise.
5406         (compute_working_sets): Likewise.
5407         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
5408         it not obsolete.
5409         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
5410         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
5411         (GCOV_HISTOGRAM_SIZE): Remove.
5412         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
5413         (struct gcov_summary): Simplify rapidly just
5414         to runs and sum_max fields.
5415         (gcov_histo_index): Remove.
5416         (NUM_GCOV_WORKING_SETS): Likewise.
5417         (compute_working_sets): Likewise.
5418         * gcov-tool.c (print_overlap_usage_message): Remove
5419         trailing empty line.
5420         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
5421         (output_lines): Remove program related line.
5422         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
5423         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
5424         histogram.
5425         (input_profile_summary): Do not read it.
5426         (merge_profile_summaries): And do not merge it.
5427         (input_symtab): Do not call removed function.
5428         * modulo-sched.c (sms_schedule): Do not print sum_max.
5429         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
5430         removed when histogram method was invented.
5431         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
5432         mode.
5433         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
5434         GCOV coding style.
5435         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
5436         and dump selected value.
5437         * profile.c (add_working_set): Remove.
5438         (get_working_sets): Likewise.
5439         (find_working_set): Likewise.
5440         (get_exec_counts): Do not work with working sets.
5441         (read_profile_edge_counts): Do not inform as sum_max is removed.
5442         (compute_branch_probabilities): Likewise.
5443         (compute_value_histograms): Remove argument for call of
5444         get_coverage_counts.
5445         * profile.h: Do not make gcov_summary const.
5447 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
5449         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
5451 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5453         PR tree-optimization/86990
5454         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
5455         Check that the entire merged store group is made of constants only for
5456         overlapping stores.
5458 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
5460         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
5461         (VTABLE_VERIFICATION_SPEC): Likewise.
5462         (SANITIZER_EARLY_SPEC): Likewise.
5463         (SANITIZER_SPEC): Likewise.
5464         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
5465         * doc/invoke.texi (Link Options): Document -r.
5467 2018-09-20  Richard Biener <rguenther@suse.de>
5469         PR middle-end/87054
5470         * gimplify.c (gimplify_expr): Retain alignment of
5471         addressable lvalue in dereference.
5473 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
5475         PR bootstrap/87013
5476         * configure.ac: Check for .loc is_stmt support.
5477         * configure, config.in: Rebuilt.
5478         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
5479         if not supported.
5481 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5483         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
5484         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
5485         -misel=no.
5487 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5489         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5490         VECTOR_OTHER.
5491         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
5492         case VECTOR_OTHER.
5494 2018-09-20  Marek Polacek  <polacek@redhat.com>
5496         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
5498 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5500         PR tree-optimization/87288
5501         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
5502         into account when determining PEELING_FOR_NITERS.
5504 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5506         PR tree-optimization/86877
5507         * tree-vect-loop.c (vect_analyze_loop_2): Call
5508         vect_verify_datarefs_alignment.
5510 2018-09-19  Marek Polacek  <polacek@redhat.com>
5512         * doc/invoke.texi: Document -Wclass-conversion.
5514 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
5516         * config/pa/pa.c (pa_adjust_priority): Delete.
5517         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
5519         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
5520         (atomic_storehi): Likewise.
5521         (atomic_storesi): Likewise.
5522         (atomic_loaddi): Restore compare and swap exchange loop code.
5524 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
5526         PR rtl-optimization/86902
5527         * combine.c (try_combine): When changing the CC mode used, don't change
5528         an unrelated mode in other_insn to that new CC mode.
5530 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
5532         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
5533         with %T in place of calls to dump_generic_expr.
5534         (prune_runtime_alias_test_list): Likewise.
5535         (create_runtime_alias_checks): Likewise.
5536         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
5537         (vect_analyze_data_ref_dependence): Likewise.
5538         (vect_slp_analyze_data_ref_dependence): Likewise.
5539         (vect_record_base_alignment): Likewise.  Use %G in place of call
5540         to dump_gimple_stmt.
5541         (vect_compute_data_ref_alignment): Likewise.
5542         (verify_data_ref_alignment): Likewise.
5543         (vect_find_same_alignment_drs): Likewise.
5544         (vect_analyze_group_access_1): Likewise.
5545         (vect_analyze_data_ref_accesses): Likewise.
5546         (dependence_distance_ge_vf): Likewise.
5547         (dump_lower_bound): Likewise.
5548         (vect_prune_runtime_alias_test_list): Likewise.
5549         (vect_find_stmt_data_reference): Likewise.
5550         (vect_analyze_data_refs): Likewise.
5551         (vect_create_addr_base_for_vector_ref): Likewise.
5552         (vect_create_data_ref_ptr): Likewise.
5553         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
5554         (vect_can_advance_ivs_p): Likewise.
5555         (vect_update_ivs_after_vectorizer): Likewise.
5556         (vect_gen_prolog_loop_niters): Likewise.
5557         (vect_prepare_for_masked_peels): Likewise.
5558         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5559         (vect_determine_vectorization_factor): Likewise.
5560         (vect_is_simple_iv_evolution): Likewise.
5561         (vect_analyze_scalar_cycles_1): Likewise.
5562         (vect_analyze_loop_operations): Likewise.
5563         (report_vect_op): Likewise.
5564         (vect_is_slp_reduction): Likewise.
5565         (check_reduction_path): Likewise.
5566         (vect_is_simple_reduction): Likewise.
5567         (vect_create_epilog_for_reduction): Likewise.
5568         (vect_finalize_reduction:): Likewise.
5569         (vectorizable_induction): Likewise.
5570         (vect_transform_loop_stmt): Likewise.
5571         (vect_transform_loop): Likewise.
5572         (optimize_mask_stores): Likewise.
5573         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
5574         (vect_split_statement): Likewise.
5575         (vect_recog_over_widening_pattern): Likewise.
5576         (vect_recog_average_pattern): Likewise.
5577         (vect_determine_min_output_precision_1): Likewise.
5578         (vect_determine_precisions_from_range): Likewise.
5579         (vect_determine_precisions_from_users): Likewise.
5580         (vect_mark_pattern_stmts): Likewise.
5581         (vect_pattern_recog_1): Likewise.
5582         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5583         (vect_record_max_nunits): Likewise.
5584         (vect_build_slp_tree_1): Likewise.
5585         (vect_build_slp_tree_2): Likewise.
5586         (vect_print_slp_tree): Likewise.
5587         (vect_analyze_slp_instance): Likewise.
5588         (vect_detect_hybrid_slp_stmts): Likewise.
5589         (vect_detect_hybrid_slp_1): Likewise.
5590         (vect_slp_analyze_operations): Likewise.
5591         (vect_slp_analyze_bb_1): Likewise.
5592         (vect_transform_slp_perm_load): Likewise.
5593         (vect_schedule_slp_instance): Likewise.
5594         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5595         (vect_mark_stmts_to_be_vectorized): Likewise.
5596         (vect_init_vector_1): Likewise.
5597         (vect_get_vec_def_for_operand): Likewise.
5598         (vect_finish_stmt_generation_1): Likewise.
5599         (vect_check_load_store_mask): Likewise.
5600         (vectorizable_call): Likewise.
5601         (vectorizable_conversion): Likewise.
5602         (vectorizable_operation): Likewise.
5603         (vectorizable_load): Likewise.
5604         (vect_analyze_stmt): Likewise.
5605         (vect_is_simple_use): Likewise.
5606         (vect_get_vector_types_for_stmt): Likewise.
5607         (vect_get_mask_type_for_stmt): Likewise.
5608         * tree-vectorizer.c (increase_alignment): Likewise.
5610 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
5612         * doc/rtl.texi: Adjust vec_select description.
5613         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
5614         non-constant selectors.
5616 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
5618         * config/aarch64/aarch64-protos.h
5619         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
5620         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
5621         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
5622         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
5623         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
5624         (AARCH64_FL_PROFILE): Move index so flags are ordered.
5625         (AARCH64_ISA_RCPC8_4): New flag.
5626         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
5627         to aarch64_offset_9bit_signed_unscaled_p.
5628         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
5629         and use stlur.
5630         * config/aarch64/constraints.md (Ust): New constraint.
5631         * config/aarch64/predicates.md.
5632         (aarch64_9bit_offset_memory_operand): New predicate.
5633         (aarch64_rcpc_memory_operand): New predicate.
5635 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
5637         PR rtl-optimization/87361
5638         * rtlanal.c (nonzero_bits1): Revert accidental change.
5640 2018-09-19  Richard Biener  <rguenther@suse.de>
5642         PR tree-optimization/87349
5643         PR tree-optimization/87342
5644         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
5646 2018-09-18  Marek Polacek  <polacek@redhat.com>
5648         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
5649         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
5651 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5653         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
5655 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5657         PR rtl-optimization/86882
5658         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
5660 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
5662         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
5663         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
5665 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
5667         PR other/87353
5668         * doc/invoke.texi (Link Options): Fix formatting and grammar.
5670 2018-09-18  Richard Biener  <rguenther@suse.de>
5672         PR middle-end/63155
5673         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
5674         (compute_samebase_partition_bases): Likewise.
5675         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
5676         (gimple_can_coalesce_p): Simplify.
5678 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
5680         Handle a library implementation of ffs calling __builtin_ffs.
5681         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
5682         (mmix_init_libfuncs): New function: make __builtin_ffs expand
5683         to __ffsdi2.
5685 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5687         * diagnostic-show-locus.c (class layout_range): Add field
5688         "m_original_idx".
5689         (layout_range::layout_range): Add "original_idx" param and use it
5690         to initialize new field.
5691         (make_range): Use 0 for original_idx.
5692         (layout::layout): Pass in index to calls to
5693         maybe_add_location_range.
5694         (layout::maybe_add_location_range): Add param "original_idx" and
5695         pass it on to layout_range.
5696         (layout::print_any_labels): Pass on range->m_original_idx to
5697         get_text call.
5698         (gcc_rich_location::add_location_if_nearby): Use 0 for
5699         original_idx.
5700         * gcc-rich-location.h (text_range_label::get_text): Update for new
5701         param.
5702         (range_label_for_type_mismatch::get_text): Likewise.
5704 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5706         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
5708 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5710         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
5711         format_string_diagnostic_t.
5712         (fmtwarn_n): Likewise.
5713         * substring-locations.c
5714         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
5715         (format_warning_n_va): Convert to...
5716         (format_string_diagnostic_t::emit_warning_n_va): ...this.
5717         (format_warning_va): Convert to...
5718         (format_string_diagnostic_t::emit_warning_va): ...this.
5719         (format_warning_at_substring): Convert to...
5720         (format_string_diagnostic_t::emit_warning): ...this.
5721         (format_warning_at_substring_n): Convert to...
5722         (format_string_diagnostic_t::emit_warning_n): ...this.
5723         * substring-locations.h (class format_string_diagnostic_t): New
5724         class.
5725         (format_warning_va): Convert to
5726         format_string_diagnostic_t::emit_warning_va.
5727         (format_warning_n_va): Convert to
5728         format_string_diagnostic_t::emit_warning_n_va.
5729         (format_warning_at_substring): Convert to
5730         format_string_diagnostic_t::emit_warning.
5731         (format_warning_at_substring_n): Convert to
5732         format_string_diagnostic_t::emit_warning_n.
5734 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
5735             Bernd Schmidt <bernds_cb1@t-online.de>
5737         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
5738         SImode args.
5740 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5742         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
5743         operand 0 predicate to nonimmediate operand.
5744         (rint<mode>2_frndint): Remove insn pattern.
5745         (rint<mode>2): Change operand 1 predicate to general_operand.
5746         Extend operand 1 to XFmode and generate rintxf2 insn.
5747         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
5748         Do not use X87MODEF mode macro.
5749         (frndintxf2_<rounding>_i387): Rename from
5750         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
5751         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
5752         to XFmode and generate significandxf3 insn.
5754 2018-09-17  Richard Biener  <rguenther@suse.de>
5756         PR tree-optimization/87328
5757         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
5758         visiting unexecutable backedges when not iterating.
5759         (do_rpo_vn): Mark all edges not executable even when not
5760         iterating.
5762 2018-09-17  Martin Jambor  <mjambor@suse.cz>
5764         PR c/63886
5765         * doc/invoke.texi (Warning Options): Likewise.
5767 2018-09-17  Richard Biener  <rguenther@suse.de>
5769         PR tree-optimization/87301
5770         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
5771         clean EH info from leftover copy assignments.
5773 2018-09-17  Martin Liska  <mliska@suse.cz>
5775         PR gcov-profile/85871
5776         * gcov.c (output_intermediate_file): Fix out of bounds
5777         access.
5779 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
5781         * config/arc/arc.c: Object attributes for core4 not reflected
5782         correctly.
5783         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
5784         core3).
5786 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
5788         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
5790 2018-09-17  Martin Liska  <mliska@suse.cz>
5792         * doc/gcov.texi: Document new option --use-hotness-colors.
5793         * gcov.c (struct source_info): Declare new field.
5794         (source_info::source_info): Set default for maximum_count.
5795         (print_usage): Add new -q option.
5796         (process_args): Process it.
5797         (accumulate_line_info): Save src->maximum_count.
5798         (output_line_beginning): Make color line number if
5799         flag_use_hotness_colors is set.
5800         (output_line_details): Pass default argument value.
5801         (output_lines): Pass src->maximum_count.
5803 2018-09-17  Martin Liska  <mliska@suse.cz>
5805         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
5806         Use processor_names table.
5807         * config/i386/i386.c (ix86_default_align): Use
5808         processor_cost_table for alignment values.
5809         (ix86_option_override_internal): Use processor_names.
5810         (ix86_function_specific_print): Likewise.
5811         * config/i386/i386.h (struct processor_costs):
5812         Add alignment values.
5813         (struct ptt): Remove and replace with const char *.
5814         * config/i386/x86-tune-costs.h (struct processor_costs):
5815         Declare default alignments for all costs.
5817 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
5819         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
5820         symbolics or VR_VARYING ranges for ABS_EXPR.
5821         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
5822         when range will wrap.
5824 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
5826         PR middle-end/86864
5827         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
5828         before and after a JUMP_TABLE_DATA.
5830 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
5832         PR middle-end/87188
5833         * dojump.c (do_compare_and_jump): Canonicalize function pointers
5834         when one operand is a function pointer.  Use POINTER_TYPE_P and
5835         FUNC_OR_METHOD_TYPE_P.
5836         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
5837         * fold-const.c (build_range_check): Likewise.
5838         * match.pd (simple_comparison): Likewise.
5840 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
5842         PR c/82967
5843         * spellcheck.c (get_edit_distance_cutoff): New function.
5844         (selftest::test_edit_distance_unit_test_oneway): Rename to...
5845         (selftest::test_get_edit_distance_one_way): ...this.
5846         (selftest::test_get_edit_distance_unit): Rename to...
5847         (selftest::test_get_edit_distance_both_ways): ...this.
5848         (selftest::test_edit_distances): Move tests to this new function,
5849         and test some more pairs of strings.  Update for above renaming.
5850         (selftest::get_old_cutoff): New function.
5851         (selftest::test_get_edit_distance_cutoff): New function.
5852         (selftest::assert_suggested_for): New function.
5853         (ASSERT_SUGGESTED_FOR): New macro.
5854         (selftest::assert_not_suggested_for): New function.
5855         (ASSERT_NOT_SUGGESTED_FOR): New macro.
5856         (selftest::test_suggestions): New function.
5857         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
5858         tests to selftest::test_edit_distances and call it.  Add calls to
5859         selftest::test_get_edit_distance_cutoff and
5860         selftest::test_suggestions.
5861         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
5862         (best_match::consider): Replace hard-coded cutoff calculation with
5863         a call to...
5864         (best_match::get_cutoff): New declaration.
5865         (best_match::get_best_meaningful_candidate): Likewise.
5867 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5869         * builtins.c (fold_builtin_strlen): Remove TODO comment.
5871 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5873         revert:
5874         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5876         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5877         terminated string literal.
5879 2018-09-14  Martin Sebor  <msebor@redhat.com>
5881         * builtins.c (unterminated_array): Handle ARRAY_REF.
5882         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
5883         * builtins.h (unterminated_array): Declare extern.
5884         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
5885         arrays.
5886         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
5887         calls.
5889 2018-09-14  Martin Sebor  <msebor@redhat.com>
5890             Jeff Law  <law@redhat.com>
5892         * builtins.c (unterminated_array): New.
5893         (expand_builtin_strcpy): Adjust.
5894         (expand_builtin_strcpy_args): Detect unterminated arrays.
5895         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
5896         unterminated arrays.
5897         * gimple-fold.h (get_maxval_strlen): Add argument.
5898         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
5900         * gimple-fold.c (get_range_strlen): Add argument.
5901         (get_maxval_strlen): Adjust.
5902         * gimple-fold.h (get_range_strlen): Add argument.
5904 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
5906         * config/i386/movdirintrin.h: Fix copyright year.
5908 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
5910         * reg-stack.c: Include regs.h.
5911         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
5912         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
5913         FIRST_STACK_REG, not DFmode.
5914         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
5915         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
5916         (change stack): Default register mode to the reg_raw_mode of
5917         FIRST_STACK_REG, not DFmode.
5918         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
5919         (*swapxf): Rename from swapxf.
5921 2018-09-14  Carl Love  <cel@us.ibm.com>
5923         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
5924         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
5926 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
5928         PR target/87224
5929         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
5930         alternatives.
5932 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
5934         PR target/85628
5935         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
5937 2018-09-14  Jason Merrill  <jason@redhat.com>
5939         Fix --enable-gather-detailed-mem-stats.
5940         * hash-table.c (hash_table_usage): Change from variable to function.
5941         * hash-table.h: Adjust.
5942         * Makefile.in: Add missing dependencies on hash-table.h.
5944 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5946         PR tree-optimization/87259
5947         PR lto/87283
5948         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
5949         execute_cse_reciprocals_1 has tried transforming.
5951 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
5953         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
5954         VR_VARYING for PLUS/MINUS_EXPR.
5956 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5958         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
5959         formatting.
5961 2018-09-14  Richard Biener  <rguenther@suse.de>
5963         PR middle-end/63155
5964         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
5965         bits for the merged partition.
5967 2018-09-13  Martin Sebor  <msebor@redhat.com>
5968             Bernd Edlinger  <bernd.edlinger@hotmail.de>
5970         * builtins.h (c_srlen): Add argument.
5971         * builtins.c (warn_string_no_nul): New function.
5972         (c_strlen): Add argument and use it.  Update recursive calls.
5973         Pass DECL argument to string_constant to get info on non
5974         terminated strings.  Update *NONSTR as needed.
5975         (fold_builtin_strlen): Add argument to calls to c_strlen.
5976         Warn for unterminated arrays.
5977         (warn_string_no_null): Add prototype.
5978         * expr.c (string_constant): Update arguments.  Update recursive
5979         calls appropriately.  Detect missing NUL terminator and outermost
5980         declaration its missing in.
5981         Improve checks for arrays with nonzero lower bound or elements
5982         that are not a single byte.  Simplify offset computation.
5983         Simplify checks for non-NUL terminated strings.
5984         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
5985         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
5987 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5989         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
5990         correctly.
5991         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
5992         argument.  Fix range checks.
5993         * fold-const.h (c_getstr): Adjust protoype.
5994         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
5995         string is constant but contains no NUL byte.
5997         * expr.c (string_constant): Adjust function comment.
5998         Remove bogus check for zero termination.
6000         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6002         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6003         (get_constant_size): Don't make STRING_CSTs larger than they are.
6004         (check_string_literal): New check function for STRING_CSTs.
6005         (output_constant): Use it.
6007 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6009         PR target/86812
6010         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6012 2018-09-13  Richard Biener  <rguenther@suse.de>
6014         PR tree-optimization/87263
6015         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6016         (struct unwind_state): Add max_rpo field.
6017         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6018         Compute max_rpo, the max RPO number a block can be backwards reached
6019         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6020         separating it from the iterating mode.
6022 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6024         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6025         (rfs_decision): New scheduling decision.
6027 2018-09-13  Richard Biener  <rguenther@suse.de>
6029         PR bootstrap/87134
6030         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6031         (vn_nary_op_insert_pieces_predicated): Do not write useless
6032         valid_dominated_by_p entry outside of the allocated storage.
6034 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6035             Tom de Vries  <tdevries@suse.de>
6037         PR debug/86985
6038         * dwarf2out.c (is_c): New function.
6039         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6041 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6043         PR target/85628
6044         * config/aarch64/aarch64.md (*aarch64_bfxil):
6045         Define.
6046         * config/aarch64/constraints.md (Ulc): Define.
6047         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6048         Define.
6049         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6050         New function.
6052 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6054         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6055         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6056         aarch64_layout_frame call.
6057         (aarch64_expand_epilogue): Likewise.
6058         (aarch64_initial_elimination_offset): Likewise.
6059         (aarch64_get_separate_components): Likewise.
6060         (aarch64_use_return_insn_p): Likewise.
6061         (aarch64_layout_frame): Remove unneeded check.
6063 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6065         * configure.ac: Only append
6066         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6067         gcc_config_arguments if it was never reconfigured or last reconfigure
6068         was with different arguments.
6069         * configure: Regenerated.
6071 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6072             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6074         PR middle-end/87290
6075         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6076         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6078 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
6080         PR tree-optimization/87287
6081         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6082         X % C == 0 to X % (unsigned) C == 0 optimization to ...
6083         * match.pd (X % C == 0): ... here.  New optimization.
6085 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6087         PR middle-end/82853
6088         * expr.h (maybe_optimize_mod_cmp): Declare.
6089         * expr.c (mod_inv): New function.
6090         (maybe_optimize_mod_cmp): New function.
6091         (do_store_flag): Use it.
6092         * cfgexpand.c (expand_gimple_cond): Likewise.
6094 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
6095             Julian Brown  <julian@codesourcery.com>
6097         PR middle-end/86336
6098         * gimplify.c (gimplify_scan_omp_clauses): Set
6099         target_firstprivatize_array_bases in OpenACC parallel and kernels
6100         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6101         OpenACC data regions.
6103 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
6105         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6106         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6108 2018-09-12  Richard Biener  <rguenther@suse.de>
6110         PR tree-optimization/87280
6111         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6112         edge but unreachable target.
6113         (do_rpo_vn): For conservatively handling a PHI only mark
6114         the backedge executable but not the block reachable.
6116 2018-09-12  Richard Biener  <rguenther@suse.de>
6118         PR tree-optimization/87266
6119         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6120         visited blocks.
6122 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
6124         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6125         constants.
6126         ("trunc<BFP:mode><DFP_ALL:mode>2")
6127         ("trunc<DFP_ALL:mode><BFP:mode>2")
6128         ("extend<BFP:mode><DFP_ALL:mode>2")
6129         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6130         according to the target operand type.
6132 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6133             Andreas Krebbel  <krebbel@linux.ibm.com>
6135         PR tree-optimization/86844
6136         * gimple-ssa-store-merging.c
6137         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6138         there are any overlapping stores in between them, make sure they are
6139         also coalesced or we give up completely.
6141 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
6143         PR middle-end/87248
6144         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6145         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
6147 2018-09-12  Tom de Vries  <tdevries@suse.de>
6149         * common.opt (gdescribe-dies): Add option.
6150         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6151         attribute for artifical and nameless decls.
6152         (dwarf2out_register_external_die): Add description attribute to
6153         external reference die.
6154         (add_desc_attribute): New functions.
6155         (gen_subprogram_die): Add description attribute to
6156         DW_TAG_call_site_parameter.
6157         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6158         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6159         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6160         -gno-describe-dies.
6161         (@item -gdescribe-dies): Add.
6163 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6165         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6166         (extract_range_from_binary_expr_1: Call
6167         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6168         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6169         depend on sign.
6171 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6173         * gimple-ssa-warn-alloca.c
6174         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6175         field for ALLOCA_BOUND_*_LARGE.
6177 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6179         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6181 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6183         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6184         for clobbers.  Remove obsolete comment.
6186 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6188         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6189         mpxchk, mpxld and mpxst types.
6190         (define_attr length_immediate): Remove all processing of mpx types.
6191         (define_attr prefix_0f): Ditto.
6192         (define_attr memory): Ditto.
6194 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6196         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6197         (log<mode>2): Change operand 1 predicate to general_operand.
6198         Extend operand 1 to XFmode and generate logxf3 insn.
6199         (log10<mode>2): Change operand 1 predicate to general_operand.
6200         Extend operand 1 to XFmode and generate log10xf3 insn.
6201         (log2<mode>2): Change operand 1 predicate to general_operand.
6202         Extend operand 1 to XFmode and generate log2xf3 insn.
6203         (fyl2xp1_extend<mode>xf3_i387): Remove.
6204         (log1p<mode>2): Change operand 1 predicate to general_operand.
6205         Extend operand 1 to XFmode and generate log1pxf3 insn.
6206         (fxtract_extend<mode>xf3_i387): Remove.
6207         (logb<mode>2): Change operand 1 predicate to general_operand.
6208         Extend operand 1 to XFmode and generate logbxf3 insn.
6209         (ilogb<mode>2): Change operand 1 predicate to general_operand.
6210         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6211         (significand<mode>2): Change operand 1 predicate to general_operand.
6212         Extend operand 1 to XFmode and generate significandxf3 insn.
6214 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6216         * gcc.c (perror_with_name, pfatal_with_name): Delete.
6217         (load_specs): Use fatal_error.
6218         (DELETE_IF_ORDINARY, process_command): Use error.
6219         (execute, run_attempt): Use fatal_error.
6221 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
6223         * diagnostic-core.h (sorry_at): New prototype.
6224         * diagnostic.c (sorry_at): New function.
6226 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
6228         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6229         by zero as VR_UNDEFINED.
6231 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6233         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6234         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6235         (<sincos>mode2): New expander.
6236         (sincos_extend<mode>xf3_i387): Remove insn pattern.
6237         (sincos -> sin, cos splitters): Remove splitter patterns.
6238         (sincos<mode>3): Change operand 2 predicate to general_operand.
6239         Extend operand 2 to XFmode and generate sincosxf3 insn.
6240         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6241         Change operand 3 predicate to const1_operand.
6242         (fptan_extend<mode>xf4_i387): Remove insn pattern.
6243         (tanxf2): Update operands in the call to fptanxf4_i387.
6244         (tan<mode>2): Change operand 1 predicate to general_operand.
6245         Extend operand 1 to XFmode and generate tanxf3 insn.
6246         (atan2xf3): Rename from *fpatanxf3_i387.
6247         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6248         (atan2xf3): Remove expander.
6249         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
6250         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6251         (atan<mode>2): Change operand 1 predicate to general_operand.
6252         Extend operand 1 to XFmode and generate atanxf3 insn.
6254 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6256         * config/i386/i386.md (x87/SSE constant load splitter): Use
6257         memory_operand instead of nonimmediate_operand for input operand
6258         predicate.
6260 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
6262         * config/i386/i386.md (float partial SSE register stall splitter): Move
6263         splitter near its instruction pattern.
6264         (float_extend partial SSE register stall splitter): Ditto.
6265         (float_truncate partial SSE register stall splitter): Ditto.
6267 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
6269         PR target/86794
6270         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6271         to speculation_safe_value_not_needed.
6273         PR target/85666
6274         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6275         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6276         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6277         leaf_function_p, instead use has_hard_reg_initial_val.
6279 2018-09-09  Nathan Sidwell  <nathan@acm.org>
6281         * gcc.h (pfatal_with_name): Don't declare here.
6282         * gcc.c (pfatal_with_name): Make static.
6284 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
6286         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6287         earlyclobber.
6289 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
6291         PR rtl-optimization/85458
6292         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6293         priority hook to reduce the priority of EXPR.
6295 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
6297         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6298         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
6299         with X87_ENABLE_FLOAT.  Remove preparation code.
6300         (*float<SWI48:mode><MODEF:mode>2): Rename from
6301         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
6302         math using "enabled" attribute.
6303         (*floatdi<MODEF:mode>2_i387): Rename from
6304         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
6305         enable for 32bit targets only.
6306         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6307         splitter.
6308         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6309         as operand 1 predicate.  Rewrite as define_insn_and_split.
6310         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6312 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6314         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
6315         to fallthru to FLOAT case.
6317 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6319         PR target/86731
6320         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
6321         around folding of vec_sl to handle out of range shift values.
6323 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6325         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
6326         Update callers to gen_fix_trunc<mode>_i387_fisttp
6327         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
6328         nonimmediate_operand.
6329         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
6330         and corresponding splitters.
6331         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
6332         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
6333         (fix_truncdi_i387_with_temp): Remove insn pattern
6334         and corresponding splitters.
6335         (fix_trunc<mode>_i387): Change operand 0 predicate to
6336         nonimmediate_operand.
6337         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
6338         and corresponding splitters.
6339         (*fistdi2_1): Remove.
6340         (fistdi2): Ditto.
6341         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
6342         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
6343         (*fist<mode>2_1): Remove.
6344         (fist<mode>2): Ditto.
6345         (fist<mode>2_with_temp): Remove insn pattern and corresponding
6346         splitters.
6347         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
6348         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
6349         (fistdi2_<rounding>): Change operand 0 predicate to
6350         nonimmediate_operand.
6351         (fistdi2_<rounding>_with_temp): Remove insn pattern
6352         and corresponding splitters.
6353         (fist<mode>2_<rounding>): Change operand 0 predicate to
6354         nonimmediate_operand.
6355         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
6356         and corresponding splitters.
6358         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
6360 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6362         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6363         the init value.
6365 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6367         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
6368         early gimple folding of vec_splat().
6369         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
6370         * gimple-fold.h: Add an extern define for tree_vec_extract().
6372 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6374         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
6375         wrappers around TREE_TYPE comparisons.
6377 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6379         PR target/80080
6380         * config/s390/predicates.md: Add nonsym_memory_operand.
6381         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
6382         contains a SYMBOL_REF, load it into an intermediate pseudo.
6383         (s390_emit_compare_and_swap): Legitimize operand.
6384         * config/s390/s390.md: Use the new nonsym_memory_operand
6385         with UNSPECV_CAS patterns.
6387 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6389         PR target/80080
6390         * config/s390/s390-passes.def: New file.
6391         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
6392         declaration.
6393         (make_pass_s390_early_mach): Add declaration.
6394         * config/s390/s390.c (make_pass_s390_early_mach):
6395         (s390_option_override): Remove dynamic registration.
6396         * config/s390/t-s390: Add s390-passes.def.
6398 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6400         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6401         Remove UNSPEC_LTREL_BASE check.
6402         (annotate_constant_pool_refs): Likewise.
6403         (find_constant_pool_ref): Likewise.
6404         (find_ltrel_base): Removed.
6405         (replace_ltrel_base): Removed.
6406         (s390_mainpool_finish): Remove replace_ltrel_base call.
6407         (s390_chunkify_start): Remove pending LTREL_BASE logic.
6408         (s390_chunkify_finish): Remove replace_ltrel_base call.
6409         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
6411 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
6413         PR target/86779
6414         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6415         to speculation_safe_value_not_needed.
6417 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
6418             Bernd Schmidt  <bernds_cb1@t-online.de>
6420         * config/nvptx/nvptx-opts.h: New file.
6421         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
6422         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
6423         (ASM_SPEC): Define.
6424         (TARGET_SM35): New macro.
6425         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
6426         correct predicate.
6427         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
6428         values.
6429         (misa=): New option.
6430         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
6432 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
6434         * config/i386/i386.md (truncdfsf2): Remove expander.
6435         (truncdfsf2_with_temp): Ditto.
6436         (truncxf<mode>2): Ditto.
6437         (*truncdfsf_fast_mixed): Remove insn pattern.
6438         (*truncdfsf_fast_i387): Ditto.
6439         (*truncdfsf_mixed): Ditto.
6440         (*truncdfsf_i387): Ditto.
6441         (*truncdfsf2_i387_1): Ditto.
6442         (*truncxfsf2_mixed): Ditto.
6443         (*truncxfdf2_mixed): Ditto.
6444         (*truncxf<mode>2_i387_noop): Ditto. Update callers
6445         to call gen_truncxf<mode>2 instead.
6446         (*truncxf<mode>2_i387): Remove.
6447         (reg->reg splitters): Remove splitter pattern.
6448         (reg->mem splitters): Ditto.
6450         (truncdfsf2): New insn pattern.
6451         (truncxf<mode>2): Ditto.
6453 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6455         * tree-ssa-math-opts.c (is_mult_by): New function.
6456         (is_square_of): Use the above.
6457         (optimize_recip_sqrt): New function.
6458         (pass_cse_reciprocals::execute): Use the above.
6460 2018-09-05  Richard Biener  <rguenther@suse.de>
6462         PR bootstrap/87134
6463         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
6464         to zero-init the emplaced vec.
6466 2018-09-05  Martin Liska  <mliska@suse.cz>
6468         PR tree-optimization/87205
6469         * tree-switch-conversion.c (pass_lower_switch::execute):
6470         Group cases for switch statements.
6472 2018-09-05  Richard Biener  <rguenther@suse.de>
6474         PR tree-optimization/87217
6475         * tree-ssa-sccvn.c (vuse_valueize): New.
6476         (vn_reference_lookup_pieces): Use it.
6477         (vn_reference_lookup): Likewise.
6479 2018-09-05  Nathan Sidwell  <nathan@acm.org>
6481         PR c++/87137
6482         * stor-layout.c (place_field): Scan forwards to check last
6483         bitfield when ms_bitfield_placement is in effect.
6485 2018-09-05  Richard Biener  <rguenther@suse.de>
6487         PR bootstrap/87225
6488         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
6489         return.
6491 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6492             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
6494         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
6495         * config.gcc (extra_objs): Build it.
6496         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
6497         Likewise.
6498         * config/aarch64/aarch64-passes.def
6499         (pass_tag_collision_avoidance): New pass.
6500         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
6501         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
6502         (aarch64_classify_address): Remove static qualifier.
6503         (aarch64_address_info, aarch64_address_type): Move to...
6504         * config/aarch64/aarch64-protos.h: ... here.
6505         (make_pass_tag_collision_avoidance): New function.
6506         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
6507         New tuning flag.
6509 2018-09-05  Martin Liska  <mliska@suse.cz>
6511         * doc/gcov.texi: Update documentation of humar
6512         readable mode.
6513         * gcov.c (format_count): Print one decimal place, it provides
6514         more fine number of situations like '1G' vs. '1.4G'.
6516 2018-09-05  Martin Liska  <mliska@suse.cz>
6518         PR target/87164
6519         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
6520         * optc-gen.awk: Allow 'Var' for Deprecated options in order
6521         to generate a MASK value.
6523 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
6525         PR debug/86593
6526         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
6527         if frame pointer isn't used.
6528         (compute_frame_pointer_to_fb_displacement): Likewise.
6530 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
6532         PR target/87198
6533         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
6534         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
6535         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
6536         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
6537         and OPTION_MASK_ISA_XSAVEC_UNSET.
6539 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
6541         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
6542         XOR operations in NAND case.
6544 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6546         * wide-int-range.cc (wide_int_range_convert): New.
6547         * wide-int-range.h (wide_int_range_convert): New.
6548         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
6549         code into wide_int_range_convert.
6550         (extract_range_into_wide_ints): Do not munge anti range constants
6551         into the entire domain.  Just return the range back.
6553 2018-09-04  Martin Liska  <mliska@suse.cz>
6555         * genmatch.c (output_line_directive): Add new argument
6556         fnargs.
6557         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
6559 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6561         * doc/invoke.texi (Option Summary): Add whitespace.
6563         * doc/invoke.texi (Option Summary): Add -Waligned-new.
6565 2018-09-04  Richard Biener  <rguenther@suse.de>
6567         PR tree-optimization/87211
6568         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
6569         backedge value we're supposed to treat as VARYING also number
6570         the PHI to VARYING in case it got a different value-number already.
6572 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6574         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
6575         (extract_range_from_binary_expr_1): Do not call
6576         vrp_can_optimize_bit_op.
6577         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
6578         static.
6579         (wide_int_range_get_mask_and_bounds): New.
6580         (wide_int_range_optimize_bit_op): New.
6581         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
6582         (wide_int_range_bit_and): Same.
6583         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
6584         (wide_int_range_optimize_bit_op): New.
6585         (wide_int_range_get_mask_and_bounds): New.
6587 2018-09-04  Richard Biener  <rguenther@suse.de>
6589         PR tree-optimization/87176
6590         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
6591         variable.  When value-numbering a virtual PHI node make sure
6592         to not value-number to the backedge value.
6594 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6596         * doc/extend.texi (Long Long, Hex Floats): Document support for
6597         long long and hex floats in more recent versions of ISO C++.
6599 2018-09-03  Richard Biener  <rguenther@suse.de>
6601         PR tree-optimization/87177
6602         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
6603         cleanup.
6605 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6607         * bb-reorder.c (edge_order): Convert to C-qsort-style
6608         tri-state comparator.
6609         (reorder_basic_blocks_simple): Change std::stable_sort to
6610         gcc_stablesort.
6612 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6614         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
6615         tri-state comparator.
6616         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
6618 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6620         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
6621         (mergesort): ... here as maximum count for using netsort.
6622         (gcc_qsort): Set nlim to 3 if stable sort is requested.
6623         (gcc_stablesort): New.
6624         * system.h (gcc_stablesort): Declare.
6626 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6628         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
6629         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
6630         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
6632 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
6634         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
6635         lxsdx and stxsdx alternatives.
6636         (*mov<mode>_hardfloat64): Ditto.
6637         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
6639 2018-09-03  Richard Biener  <rguenther@suse.de>
6641         PR tree-optimization/87200
6642         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
6643         simplify result.
6645 2018-09-03  Martin Liska  <mliska@suse.cz>
6647         PR tree-optimization/87201
6648         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6649         Fix parenthesis in an expression.
6651 2018-09-03  Richard Biener  <rguenther@suse.de>
6653         PR tree-optimization/87197
6654         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
6655         visited.  CSE the VN_INFO hashtable lookup.
6657         PR tree-optimization/87169
6658         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
6659         iterating make sure there's no extra backedges from irreducible
6660         regions feeding the header.  Mark the destination block
6661         executable.
6663 2018-09-03  Martin Liska  <mliska@suse.cz>
6665         PR driver/83193
6666         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
6667         * common/common-targhooks.c (default_get_valid_option_values):
6668         New function.
6669         * common/common-targhooks.h (default_get_valid_option_values):
6670         Likewise.
6671         * common/config/i386/i386-common.c: Move processor_target_table
6672         from i386.c.
6673         (ix86_get_valid_option_values): New function.
6674         (TARGET_GET_VALID_OPTION_VALUES): New macro.
6675         * config/i386/i386.c (struct ptt): Move to i386-common.c.
6676         (PTA_*): Move all defined masks into i386-common.c.
6677         (ix86_function_specific_restore): Use new processor_cost_table.
6678         * config/i386/i386.h (struct ptt): Moved from i386.c.
6679         (struct pta): Likewise.
6680         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
6681         * doc/tm.texi.in: Likewise.
6682         * opt-suggestions.c (option_proposer::suggest_option):
6683         Pass prefix to build_option_suggestions.
6684         (option_proposer::get_completions): Likewise.
6685         (option_proposer::build_option_suggestions): Use the new target
6686         hook.
6687         * opts.c (struct option_help_tuple): New struct.
6688         (print_filtered_help): Use the new target hook.
6690 2018-09-03  Martin Liska  <mliska@suse.cz>
6692         PR middle-end/59521
6693         * predict.c (set_even_probabilities): Add likely_edges
6694         argument and handle cases where we have precisely one
6695         likely edge.
6696         (combine_predictions_for_bb): Catch also likely_edges.
6697         (tree_predict_by_opcode): Handle gswitch statements.
6698         * tree-cfg.h (find_case_label_for_value): New declaration.
6699         (find_taken_edge_switch_expr): Likewise.
6700         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6701         Find pivot in decision tree based on probabily, not by number of
6702         nodes.
6704 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
6706         * doc/standards.texi (Standards): Update Objective-C reference.
6708 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6710         * doc/install.texi (Prerequisites): Update link for MPC.
6712 2018-09-01  Michael Matz  <matz@suse.de>
6714         PR tree-optimization/87074
6715         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
6716         PHIs for outer-loop uses.
6718 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6720         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
6721         * doc/invoke.texi (C Dialect Options): Ditto.
6723 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6725         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
6727 2018-08-31  Richard Biener  <rguenther@suse.de>
6729         PR tree-optimization/87168
6730         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
6731         (rpo_elim::eliminate_avail): When OP was not visited it must
6732         be available.
6734 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
6736         * tree-vrp.c (copy_value_range): Convert param "from" from
6737         "value_range *" to "const value_range *".
6738         (range_is_null): Likewise for param "vr".
6739         (range_int_cst_p): Likewise.
6740         (range_int_cst_singleton_p): Likewise.
6741         (symbolic_range_p): Likewise.
6742         (value_ranges_intersect_p): Likewise for both params.
6743         (value_range_nonnegative_p): Likewise for param "vr".
6744         (value_range_constant_singleton): Likewise.
6745         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
6746         (extract_range_into_wide_ints): Likewise for param "vr".
6747         (extract_range_from_multiplicative_op): Likewise for params "vr0"
6748         and "vr1".
6749         (vrp_can_optimize_bit_op): Likewise.
6750         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
6751         "vr1_".
6752         (extract_range_from_unary_expr): Likewise.
6753         (debug_value_range): Likewise for param "vr".
6754         (value_range::dump): Add "const" qualifier.
6755         (vrp_prop::check_array_ref): Convert local "vr" from
6756         "value_range *" to "const value_range *".
6757         (vrp_prop::check_mem_ref): Likewise.
6758         (vrp_prop::visit_stmt): Likewise for local "old_vr".
6759         (vrp_intersect_ranges_1): Likewise for param "vr_1".
6760         (vrp_intersect_ranges): Likewise.
6761         (simplify_stmt_for_jump_threading): Likewise for local "vr".
6762         (vrp_prop::vrp_finalize): Likewise.
6763         * tree-vrp.h (value_range::dump): Add "const" qualifier.
6764         (vrp_intersect_ranges): Add "const" qualifier to params as above.
6765         (extract_range_from_unary_expr): Likewise.
6766         (value_range_constant_singleton): Likewise.
6767         (symbolic_range_p): Likewise.
6768         (copy_value_range): Likewise.
6769         (extract_range_from_binary_expr_1): Likewise.
6770         (range_int_cst_p): Likewise.
6771         (vrp_set_zero_nonzero_bits): Likewise.
6772         (range_int_cst_singleton_p): Likewise.
6774 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
6776         * config/aarch64/arm_neon.h (vabsd_s64): New.
6777         (vnegd_s64): Likewise.
6779 2018-08-31  Martin Jambor  <mjambor@suse.cz>
6781         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
6783 2018-08-31  Martin Liska  <mliska@suse.cz>
6785         * ipa-icf.c (sem_item::add_type): Use
6786         sem_item::m_type_hash_cache.
6787         * ipa-icf.h: Move the cache from sem_item_optimizer
6788         to sem_item.
6790 2018-08-31  Nathan Sidwell  <nathan@acm.org>
6792         * doc/extend.texi (Backwards Compatibility): Remove implicit
6793         extern C leeway of () being (...).
6795 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6797         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
6799 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
6801         PR target/86684
6802         PR target/87149
6803         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
6805 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
6807         PR middle-end/87138
6808         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
6809         gen_int_mode.  Formatting fixes.
6811 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
6813         * target.def (custom_function_descriptors): Improve documentation.
6814         * doc/tm.texi.in (Trampolines): Expand discussion of function
6815         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
6816         beginning of the section.
6817         * doc/tm.texi: Regenerated.
6819 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
6821         * cfg.h (class auto_edge_flag): Spell out the template-id of the
6822         base class in the initializer list.  This is a workaround for
6823         building with older GCC.
6824         (class auto_bb_flag): Likewise.
6826 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
6828         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
6829         (altivec_vcmpequ<VI_char>_p): Remove star.
6830         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
6831         vector load modes.
6832         (expand_strncmp_vec_sequence): New function.
6833         (emit_final_str_compare_vec): New function.
6834         (expand_strn_compare): Add support for vector strncmp.
6835         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
6836         length specification to bytes.
6837         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
6838         (vcmpnezb_p): New pattern.
6839         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
6840         for option -mstring-compare-inline-limit.
6842 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
6844         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
6845         (PTA_SKYLAKE): Add PTA_AES.
6846         (PTA_GOLDMONT): Likewise.
6848 2018-08-29  Jan Hubicka  <jh@suse.cz>
6850         PR lto/86517
6851         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
6852         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
6854 2018-08-29  Jan Hubicka  <jh@suse.cz>
6856         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
6857         TYPE_STUB_DECL.
6858         (hash_tree): Do not visit TYPE_STUB_DECL.
6859         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
6860         stream TYPE_STUB_DECL.
6861         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
6862         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
6863         after free_lang_data.
6864         (type_in_anonymous_namespace_p): Likewise.
6866 2018-08-29  Jan Hubicka  <jh@suse.cz>
6868         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6869         comment that it has to be even number.
6870         (class sreal): Change m_sig type to int32_t.
6871         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6872         int64_t for temporary calculations.
6873         (sreal_verify_basics): Drop one bit from minimum and maximum.
6875 2018-08-30  Richard Biener  <rguenther@suse.de>
6877         PR tree-optimization/87147
6878         * tree-ssa-sccvn.c (SSA_VISITED): New function.
6879         (visit_phi): When the degenerate result is from the backedge and
6880         we didn't visit its definition yet drop to VARYING.
6881         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
6883 2018-08-29  Jan Hubicka  <jh@suse.cz>
6885         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
6886         DECL_VINDEX.
6887         (hash_tree): Likewise.
6889 2018-08-29  Jan Hubicka  <jh@suse.cz>
6891         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
6892         and TYPE_NEXT_REF_TO.
6894 2018-08-29  Jan Hubicka  <jh@suse.cz>
6896         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6897         comment that it has to be even number.
6898         (class sreal): Change m_sig type to int32_t.
6899         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6900         int64_t for temporary calculations.
6901         (sreal_verify_basics): Drop one bit from minimum and maximum.
6903 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
6905         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
6907 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
6909         PR middle-end/86995
6910         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
6911         if to_add is negative.
6913 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6915         PR middle-end/87053
6916         * builtins.c (c_strlen): Improve range checks.
6918 2018-08-29  Martin Sebor  <msebor@redhat.com>
6919             Jeff Law  <law@redhat.com>
6921         PR tree-optimization/86714
6922         PR tree-optimization/86711
6923         * builtins.c (c_strlen): Add arguments to call to string_constant.
6924         * expr.c (string_constant): Add argument.  Detect missing nul
6925         terminator and outermost declaration it's missing in.
6926         * expr.h (string_constant): Add argument.
6927         * fold-const.c (read_from_constant_string): Add arguments to call to
6928         string_constant.
6929         (c_getstr): Likewise.
6930         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6931         to string_constant.
6932         * tree-ssa-strlen.c (get_stridx): Likewise.
6934 2018-08-29  Jan Hubicka  <jh@suse.cz>
6936         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
6937         Do not stream DECL_VINDEX.
6938         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
6939         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
6940         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
6942 2018-08-29  Richard Biener  <rguenther@suse.de>
6944         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
6945         virtual operands that are not default defs to honor region
6946         boundaries.
6947         (rpo_vn_valueize): Remove ineffective code here.
6949 2018-08-29  Richard Biener  <rguenther@suse.de>
6951         PR tree-optimization/87132
6952         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
6953         when skipping defs reachable over backedges.
6955 2018-08-29  Richard Biener  <rguenther@suse.de>
6957         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
6958         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
6959         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
6960         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
6961         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
6962         (vn_reference_lookup_3): Remove use of const_parms.
6963         (free_rpo_vn): Do not free const_parms.
6964         (do_rpo_vn): Do not call init_const_parms.
6965         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
6966         SSA_NAME_POINTS_TO_READONLY_MEMORY.
6967         (call_may_clobber_ref_p_1): Likewise.
6969 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
6971         PR other/86726
6972         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
6973         (-O): Ditto.
6974         (-ftree-scev-cprop): Document.
6976 2018-08-29  Jan Hubicka  <jh@suse.cz>
6978         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
6979         parameters.
6980         (sreal constructor): Update.
6981         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
6982         sreal:operator/): Update.
6984 2018-08-29  Martin Liska  <mliska@suse.cz>
6986         * tree-switch-conversion.c (switch_conversion::expand):
6987         Strenghten assumption about gswitch statements.
6989 2018-08-29  Richard Biener  <rguenther@suse.de>
6991         PR tree-optimization/87117
6992         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
6993         re-value-number released SSA VDEFs.
6995 2018-08-29  Richard Biener  <rguenther@suse.de>
6997         PR tree-optimization/87126
6998         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7000 2018-08-28  Jim Wilson  <jimw@sifive.com>
7002         * config/riscv/pic.md: Rewrite.
7003         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7004         invalid address.
7005         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7006         (SOFTF, default_load, softload, softstore): New.
7008 2018-08-28  Jeff Law  <law@redhat.com>
7010         * fold-const.c (fold_binary_loc): Remove recently added assert.
7012 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7014         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7015         to OP parmeter of generated function.
7017 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7019         PR tree-optimization/87009
7020         * match.pd: Add boolean optimizations.
7022 2018-08-28  Martin Sebor  <msebor@redhat.com>
7024         PR middle-end/86631
7025         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7026         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7027         (pass_walloca::gate): Use it.
7028         (alloca_call_type): Same.
7029         (pass_walloca::execute): Same.
7030         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7032 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7034         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7035         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7036         >= 3005 to >= 9000.
7038 2018-08-28  Richard Biener  <rguenther@suse.de>
7040         PR tree-optimization/87124
7041         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7042         constants before looking up avail.
7044 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7046         PR middle-end/87099
7047         * calls.c (maybe_warn_nonstring_arg): Punt early if
7048         warn_stringop_overflow is zero.  Don't call get_range_strlen
7049         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7050         Swap comparison operands to have constants on rhs.  Only use
7051         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
7052         increment lenrng[0].
7054 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
7056         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7057         use of tree_to_shwi.  Remove duplicated test for the size being
7058         a whole number of bytes.
7060 2018-08-28  Richard Biener  <rguenther@suse.de>
7062         PR tree-optimization/87117
7063         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7064         Handle removed stmt without LHS (GIMPLE_NOP).
7066 2018-08-28  Richard Biener  <rguenther@suse.de>
7068         PR tree-optimization/87117
7069         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7070         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7072 2018-08-28  Richard Biener  <rguenther@suse.de>
7074         PR tree-optimization/87117
7075         * tree-ssa-pre.c (compute_avail): Do not make expressions
7076         with predicated values available.
7077         (get_expr_value_id): Assert we do not run into predicated value
7078         expressions.
7080 2018-08-28  Richard Biener  <rguenther@suse.de>
7082         PR tree-optimization/87117
7083         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7084         get virtual operands.
7085         (get_expr_operands): Handle STRING_CST like other decls.
7087 2018-08-28  Martin Liska  <mliska@suse.cz>
7089         * tree.h: Update documentation of fndecl_built_in_p
7090         functions.
7093 2018-08-27  Jeff Law  <law@redhat.com>
7094         PR tree-optimization/87110
7095         * tree-ssa-dse.c (compute_trims): Handle non-constant
7096         TYPE_SIZE_UNIT.
7098 2018-08-27  Martin Sebor  <msebor@redhat.com>
7100         PR tree-optimization/86914
7101         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7103 2018-08-27  Martin Sebor  <msebor@redhat.com>
7105         PR tree-optimization/87112
7106         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7107         the type of the bound argument.
7109 2018-08-27  Jeff Law  <law@redhat.com>
7111         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7112         type does not have a TYPE_SIZE_UNIT.
7114 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
7116         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7117         with include of backend.h.
7119 2018-08-27  Richard Biener  <rguenther@suse.de>
7121         PR tree-optimization/86927
7122         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7123         use const cond reduction code.
7125 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
7127         PR tree-optimization/85758
7128         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7130 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7132         PR c++/87091
7133         * diagnostic-show-locus.c (class layout_range): Update for
7134         conversion of show_caret_p to a tri-state.
7135         (layout_range::layout_range): Likewise.
7136         (make_range): Likewise.
7137         (layout::maybe_add_location_range): Likewise.
7138         (layout::should_print_annotation_line_p): Don't show annotation
7139         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7140         (layout::get_state_at_point): Update for conversion of
7141         show_caret_p to a tri-state.  Bail out early for
7142         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7143         underlining or source colorization.
7144         (gcc_rich_location::add_location_if_nearby): Update for conversion
7145         of show_caret_p to a tri-state.
7146         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7147         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7148         Likewise.
7149         (selftest::test_one_liner_labels): Likewise.
7150         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7151         conversion of show_caret_p to a tri-state.
7152         * pretty-print.c (text_info::set_location): Likewise.
7153         * pretty-print.h (text_info::set_location): Likewise.
7154         * substring-locations.c (format_warning_n_va): Likewise.
7155         * tree-diagnostic.c (default_tree_printer): Likewise.
7156         * tree-pretty-print.c (newline_and_indent): Likewise.
7158 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7160         PR c++/87091
7161         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7162         line above for line-insertion fix-it hints.
7163         (selftest::test_fixit_insert_containing_newline): Update the
7164         expected results, and add a test with line-numbering enabled.
7166 2018-08-27  Martin Liska  <mliska@suse.cz>
7168         PR sanitizer/86962
7169         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7170         params with DECL_HAS_VALUE_EXPR_P.
7172 2018-08-27  Martin Liska  <mliska@suse.cz>
7174         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7175         selected expansion strategy.
7177 2018-08-27  Martin Liska  <mliska@suse.cz>
7179         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7180         * builtins.c (is_builtin_fn): Likewise.
7181         * attribs.c (diag_attr_exclusions): Use new function
7182         fndecl_built_in_p and remove check for FUNCTION_DECL if
7183         possible.
7184         (builtin_mathfn_code): Likewise.
7185         (fold_builtin_expect): Likewise.
7186         (fold_call_expr): Likewise.
7187         (fold_builtin_call_array): Likewise.
7188         (fold_call_stmt): Likewise.
7189         (set_builtin_user_assembler_name): Likewise.
7190         (is_simple_builtin): Likewise.
7191         * calls.c (gimple_alloca_call_p): Likewise.
7192         (maybe_warn_nonstring_arg): Likewise.
7193         * cfgexpand.c (expand_call_stmt): Likewise.
7194         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7195         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7196         (cgraph_node::verify_node): Likewise.
7197         * cgraphclones.c (build_function_decl_skip_args): Likewise.
7198         (cgraph_node::create_clone): Likewise.
7199         * config/arm/arm.c (arm_insert_attributes): Likewise.
7200         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7201         * dse.c (scan_insn): Likewise.
7202         * expr.c (expand_expr_real_1): Likewise.
7203         * fold-const.c (operand_equal_p): Likewise.
7204         (fold_binary_loc): Likewise.
7205         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7206         * gimple-low.c (lower_stmt): Likewise.
7207         * gimple-pretty-print.c (dump_gimple_call): Likewise.
7208         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7209         Likewise.
7210         * gimple.c (gimple_build_call_from_tree): Likewise.
7211         (gimple_call_builtin_p): Likewise.
7212         (gimple_call_combined_fn): Likewise.
7213         * gimplify.c (gimplify_call_expr): Likewise.
7214         (gimple_boolify): Likewise.
7215         (gimplify_modify_expr): Likewise.
7216         (gimplify_addr_expr): Likewise.
7217         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7218         * ipa-cp.c (determine_versionability): Likewise.
7219         * ipa-fnsummary.c (compute_fn_summary): Likewise.
7220         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7221         * ipa-split.c (visit_bb): Likewise.
7222         (split_function): Likewise.
7223         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7224         * lto-cgraph.c (input_node): Likewise.
7225         * lto-streamer-out.c (write_symbol): Likewise.
7226         * omp-low.c (setjmp_or_longjmp_p): Likewise.
7227         (lower_omp_1): Likewise.
7228         * predict.c (strip_predict_hints): Likewise.
7229         * print-tree.c (print_node): Likewise.
7230         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7231         * trans-mem.c (is_tm_irrevocable): Likewise.
7232         (is_tm_load): Likewise.
7233         (is_tm_simple_load): Likewise.
7234         (is_tm_store): Likewise.
7235         (is_tm_simple_store): Likewise.
7236         (is_tm_abort): Likewise.
7237         (tm_region_init_1): Likewise.
7238         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7239         * tree-cfg.c (verify_gimple_call): Likewise.
7240         (move_stmt_r): Likewise.
7241         (stmt_can_terminate_bb_p): Likewise.
7242         * tree-eh.c (lower_eh_constructs_2): Likewise.
7243         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7244         * tree-inline.c (remap_gimple_stmt): Likewise.
7245         (copy_bb): Likewise.
7246         (estimate_num_insns): Likewise.
7247         (fold_marked_statements): Likewise.
7248         * tree-sra.c (scan_function): Likewise.
7249         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7250         (optimize_stack_restore): Likewise.
7251         (pass_fold_builtins::execute): Likewise.
7252         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7253         (mark_all_reaching_defs_necessary_1): Likewise.
7254         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7255         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7256         (pass_forwprop::execute): Likewise.
7257         * tree-ssa-loop-im.c (stmt_cost): Likewise.
7258         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7259         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7260         * tree-ssa-strlen.c (get_string_length): Likewise.
7261         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7262         (find_func_aliases_for_call): Likewise.
7263         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7264         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7265         * tree-tailcall.c (find_tail_calls): Likewise.
7266         * tree.c (need_assembler_name_p): Likewise.
7267         (free_lang_data_in_decl): Likewise.
7268         (get_call_combined_fn): Likewise.
7269         * ubsan.c (is_ubsan_builtin_p): Likewise.
7270         * varasm.c (incorporeal_function_p): Likewise.
7271         * tree.h (DECL_BUILT_IN): Remove and replace with
7272         fndecl_built_in_p.
7273         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7274         (fndecl_built_in_p): New.
7276 2018-08-27  Martin Liska  <mliska@suse.cz>
7278         PR tree-optimization/86847
7279         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7280         Dump also subtree probability.
7281         (switch_decision_tree::do_jump_if_equal): New function.
7282         (switch_decision_tree::emit_case_nodes): Handle special
7283         situations in balanced tree that can be emitted much simpler.
7284         Fix calculation of probabilities that happen in tree expansion.
7285         * tree-switch-conversion.h (struct cluster): Add
7286         is_single_value_p.
7287         (struct simple_cluster): Likewise.
7288         (struct case_tree_node): Add new function has_child.
7289         (do_jump_if_equal): New.
7291 2018-08-27  Martin Liska  <mliska@suse.cz>
7293         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7294         Add new argument to bit_test_cluster constructor.
7295         (bit_test_cluster::emit): Set bits really number of values
7296         handlel by a test.
7297         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7298         probability argument.
7299         * tree-switch-conversion.h (struct bit_test_cluster):
7300         Add m_handles_entire_switch.
7302 2018-08-27  Martin Liska  <mliska@suse.cz>
7304         PR tree-optimization/86702
7305         * tree-switch-conversion.c (jump_table_cluster::emit):
7306         Make probabilities even for values in jump table
7307         according to number of cases handled.
7308         (switch_decision_tree::compute_cases_per_edge): Pass
7309         argument to reset_out_edges_aux function.
7310         (switch_decision_tree::analyze_switch_statement): Likewise.
7311         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7312         Make it static.
7314 2018-08-27  Martin Liska  <mliska@suse.cz>
7316         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
7317         cfun argument explicitly.
7318         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
7319         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
7320         function gimple_switch_default_bb.
7321         (convert_switch_statements):
7322         (expand_builtins):
7323         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
7324         * stmt.c (label_to_block_fn): Use label_to_block and pass
7325         cfun argument explicitly and use gimple_switch_label_bb.
7326         (expand_case): Likewise.
7327         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
7328         cfun argument explicitly. Likewise.
7329         (make_edges_bb): Likewise.
7330         (make_cond_expr_edges): Likewise.
7331         (get_cases_for_edge): Likewise.
7332         (make_gimple_switch_edges): Likewise.
7333         (label_to_block_fn): Likewise.
7334         (label_to_block): Likewise.
7335         (make_goto_expr_edges): Likewise.
7336         (make_gimple_asm_edges): Likewise.
7337         (main_block_label): Likewise.
7338         (group_case_labels_stmt): Likewise.
7339         (find_taken_edge_computed_goto): Likewise.
7340         (find_taken_edge_switch_expr): Likewise.
7341         (gimple_verify_flow_info): Likewise.
7342         (gimple_redirect_edge_and_branch): Likewise.
7343         (gimple_switch_label_bb): New function.
7344         (gimple_switch_default_bb): Likewise.
7345         (gimple_switch_edge): Likewise.
7346         (gimple_switch_default_edge): Likewise.
7347         * tree-cfg.h (label_to_block_fn): Remove and replace ...
7348         (label_to_block): ... with this.
7349         (gimple_switch_label_bb): New.
7350         (gimple_switch_default_bb): Likewise.
7351         (gimple_switch_edge): Likewise.
7352         (gimple_switch_default_edge): Likewise.
7353         * tree-cfgcleanup.c (convert_single_case_switch): Use
7354         new gimple functions and pass new argument to label_to_block.
7355         (cleanup_control_flow_bb):
7356         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
7357         cfun argument explicitly.
7358         (make_eh_edges): Likewise.
7359         (redirect_eh_dispatch_edge): Likewise.
7360         (lower_resx): Likewise.
7361         (lower_eh_dispatch): Likewise.
7362         (maybe_remove_unreachable_handlers): Likewise.
7363         (unsplit_eh): Likewise.
7364         (cleanup_empty_eh): Likewise.
7365         (verify_eh_edges): Likewise.
7366         (verify_eh_dispatch_edge): Likewise.
7367         * tree-ssa-dom.c (record_edge_info): Likewise.
7368         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
7369         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
7370         (thread_through_normal_block): Likewise.
7371         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
7372         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
7373         * tree-switch-conversion.c (switch_conversion::collect): Use new
7374         gimple functions.
7375         (switch_conversion::check_final_bb): Likewise.
7376         (switch_conversion::gather_default_values): Pass new argument
7377         to label_to_block.
7378         (switch_conversion::build_constructors): Likewise.
7379         (switch_decision_tree::compute_cases_per_edge): Use new
7380         gimple_switch_edge function.
7381         (switch_decision_tree::analyze_switch_statement): Pass new argument
7382         to label_to_block.
7383         (switch_decision_tree::try_switch_expansion): Use
7384         gimple_switch_default_edge.
7385         * tree-vrp.c (find_switch_asserts): Pass new argument
7386         to label_to_block.
7387         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
7388         (vr_values::simplify_switch_using_ranges): Likewise.
7390 2018-08-27  Richard Biener  <rguenther@suse.de>
7392         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
7393         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
7395         * tree-ssa-sccvn.h (struct vn_pval): New structure.
7396         (struct vn_nary_op_s): Add unwind_to member.  Add
7397         predicated_values flag and put result into a union together
7398         with a linked list of vn_pval.
7399         (struct vn_ssa_aux): Add name member to make maintaining
7400         a map of SSA name to vn_ssa_aux possible.  Remove no longer
7401         needed info, dfsnum, low, visited, on_sccstack, use_processed
7402         and range_info_anti_range_p members.
7403         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
7404         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
7405         New functions.
7406         (vn_valueize): New global.
7407         (vn_context_bb): Likewise.
7408         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
7409         VN_INFO_PTR_INFO): Remove.
7410         * tree-ssa-sccvn.c: ... (rewrite)
7411         (pass_fre::execute): For -O2+ initialize loops and run
7412         RPO VN in optimistic mode (iterating).  For -O1 and -Og
7413         run RPO VN in non-optimistic mode.
7414         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
7415         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
7416         * doc/invoke.texi (sccvn-max-scc-size): Remove.
7417         (rpo-vn-max-loop-depth): Document.
7418         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
7419         when valuezing the VUSE signals we walked out of the region.
7420         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
7421         (phi_translate): Set VN context block to use for availability
7422         lookup.
7423         (compute_avail): Likewise.
7424         (pre_valueize): New function.
7425         (pass_pre::execute): Adjust to the RPO VN API.
7427         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
7428         (propagate_constants_for_unrolling): Remove.
7429         (tree_unroll_loops_completely): Perform value-numbering
7430         on the unrolled bodies loop parent.
7432 2018-08-27  Richard Biener  <rguenther@suse.de>
7434         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
7435         for partial antic compute.
7437 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
7439         PR rtl-optimization/87065
7440         * combine.c (simplify_if_then_else): Formatting fix.
7441         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
7442         check.
7443         (known_cond): Don't return const_true_rtx for vector modes.  Use
7444         CONST0_RTX instead of const0_rtx.  Formatting fixes.
7446 2018-08-27  Martin Liska  <mliska@suse.cz>
7448         PR gcov-profile/87069
7449         * gcov.c (process_file): Record files already processed
7450         and warn about a file being processed multiple times.
7452 2018-08-27  Martin Liska  <mliska@suse.cz>
7454         PR driver/83193
7455         * config/aarch64/aarch64.c (aarch64_override_options_internal):
7456         Set default values for x_aarch64_*_string strings.
7457         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
7458         prefix.  For -mabi do not print '=ABI' in help and use
7459         <option_value> format for -msve-vector-bits and -moverride
7460         options.
7462 2018-08-26  Jeff Law  <law@redhat.com>
7464         * config/mips/frame-header-opt.c: Include "backend.h" rather than
7465         "cfg.h"
7467 2018-08-26  Marek Polacek  <polacek@redhat.com>
7469         PR c++/87029, Implement -Wredundant-move.
7470         * doc/invoke.texi: Document -Wredundant-move.
7472 2018-08-25  Martin Sebor  <msebor@redhat.com>
7474         PR tree-optimization/87059
7475         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
7476         to the same type as the other.
7477         * fold-const.c (fold_binary_loc): Assert expectation.
7479 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
7481         * config/darwin.c (machopic_legitimize_pic_address): Clean up
7482         extraneous parentheses, dead code section and formatting.
7484 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
7486         PR c++/87091
7487         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
7488         wide enough for jumps in the line-numbering to be visible.
7489         (layout::print_gap_in_line_numbering): New member function.
7490         (layout::calculate_line_spans): When using line numbering, merge
7491         line spans that are only 1 line apart.
7492         (diagnostic_show_locus): When printing line numbers, show gaps in
7493         line numbering directly, rather than printing headers.
7494         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
7495         line-numbering with multiple line spans.
7496         (selftest::test_fixit_insert_containing_newline_2): Add test of
7497         line-numbering, in which the spans are close enough to be merged.
7499 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7501         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
7502         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
7503         * tree-vrp.c (range_is_nonnull): Remove.
7504         (range_includes_zero_p): Accept value_range instead of min/max.
7505         (extract_range_from_binary_expr_1): Do not early bail on
7506         POINTER_PLUS_EXPR.
7507         Use range_includes_zero_p instead of range_is_nonnull.
7508         (extract_range_from_unary_expr): Use range_includes_zero_p instead
7509         of range_is_nonnull.
7510         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
7511         special case VR_ANTI_RANGE.
7512         (vrp_finalize): Same.
7513         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
7514         instead of min/max.
7515         (range_is_nonnull): Remove.
7516         * vr-values.c (vrp_stmt_computes_nonzero): Use
7517         range_includes_zero_p instead of range_is_nonnull.
7518         (extract_range_basic): Pass value_range to range_includes_zero_p
7519         instead of range_is_nonnull.
7521 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
7523         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
7524         * emit-rtl.h (rtl_data): Remove return_bnd.
7525         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
7526         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
7527         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
7528         (POINTER_BOUNDS_MODE): Remove definition.
7529         (make_pointer_bounds_mode): Remove.
7530         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
7531         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
7532         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
7533         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
7534         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
7535         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
7536         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
7538         * config/i386/i386-modes.def (BND32, BND64): Remove.
7539         * config/i386/i386.c (dbx_register_map): Remove bound registers.
7540         (dbx64_register_map): Ditto.
7541         (svr4_dbx_register_map): Ditto.
7542         (indirect_thunk_bnd_needed): Remove.
7543         (indirect_thunks_bnd_used): Ditto.
7544         (indirect_return_bnd_needed): Ditto.
7545         (indirect_return_via_cx_bnd): Ditto.
7546         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
7547         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
7548         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
7549         (output_indirect_thunk_function): Remove handling of
7550         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
7551         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
7552         (ix86_save_reg): Remove handling of crtl->return_bnd.
7553         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
7554         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
7555         and UNSPEC_BNDLX_ADDR.
7556         (ix86_output_indirect_branch_via_reg): Remove handling of
7557         indirect_thunk_prefix_bnd.
7558         (ix86_output_indirect_branch_via_push): Ditto.
7559         (ix86_output_function_return): Ditto.
7560         (ix86_output_indirect_function_return): Ditto.
7561         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
7562         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
7563         (CALL_USED_REGISTERS): Ditto.
7564         (REG_ALLOC_ORDER): Update for removal of bound registers.
7565         (HI_REGISTER_NAMES): Ditto.
7566         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
7567         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
7568         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
7569         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
7570         (FIRST_PSEUDO_REG): Update.
7571         (BND): Remove mode iterator.
7572         * config/i386/predicates.md (bnd_mem_operator): Remove.
7574 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7576         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
7577         vectors.
7579 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7581         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
7582         the case in which the permute needs only a single element and
7583         repeats for every vector of the result.  Extend that case to
7584         handle variable-length vectors.
7585         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7587 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
7589         PR debug/79342
7590         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
7591         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
7593 2018-08-24  Richard Biener  <rguenther@suse.de>
7595         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
7596         bb_flags_allocated members.
7597         (auto_flag): New RAII class for allocating flags.
7598         (auto_edge_flag): New RAII class for allocating edge flags.
7599         (auto_bb_flag): New RAII class for allocating bb flags.
7600         * cfgloop.c (verify_loop_structure): Allocate temporary edge
7601         flag dynamically.
7602         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
7603         in favor of temporarily allocated BB flag.
7604         * hsa-brig.c: Re-order includes.
7605         * hsa-dump.c: Likewise.
7606         * hsa-regalloc.c: Likewise.
7607         * print-rtl.c: Likewise.
7608         * profile-count.c: Likewise.
7610 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
7612         PR target/86989
7613         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
7614         the TOC register.
7616 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7618         PR 87073/bootstrap
7619         * wide-int-range.cc (wide_int_range_div): Do not ignore result
7620         from wide_int_range_multiplicative_op.
7622 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7624         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
7625         "permutaion".
7627 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
7629         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
7630         to 'expanded'.
7632 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
7634         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
7635         full GENERIC expression used for replacement.
7637 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
7639         * tree-vrp.c (abs_extent_range): Remove.
7640         (extract_range_into_wide_ints): Pass wide ints by reference.
7641         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
7642         Pass wide ints by reference in all calls to
7643         extract_range_into_wide_ints.
7644         * wide-int-range.cc (wide_int_range_div): New.
7645         * wide-int-range.h (wide_int_range_div): New.
7646         (wide_int_range_includes_zero_p): New.
7647         (wide_int_range_zero_p): New.
7649 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
7651         * config/aarch64/aarch64.md (arches): New enum.
7652         (arch): New enum attr.
7653         (arch_enabled): New attr.
7654         (enabled): Now uses arch_enabled only.
7655         (simd, sve, fp16): Removed attribute.
7656         (fp): Attr now defined in terms of 'arch'.
7657         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
7658         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
7659         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
7660         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
7661         attributes into 'arch'.
7662         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
7663         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
7664         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
7665         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
7666         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
7667         'simd' attribute into 'arch'.
7668         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
7669         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
7670         Convert use of 'fp' attribute to 'arch'.
7671         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
7672         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
7673         into 'arch'.
7674         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
7675         (different modes) Merge 'fp' and 'simd' into 'arch'.
7676         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
7677         'simd' into 'arch'.
7679 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
7681         PR rtl-optimization/87026
7682         * expmed.c (canonicalize_comparison): If we can no longer create
7683         pseudoregisters, don't.
7685 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
7687         PR target/86951
7688         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
7689         prototype.
7690         * config/arm/arm.c (speculation_barrier_libfunc): New static
7691         variable.
7692         (arm_init_libfuncs): Initialize it.
7693         (arm_emit_speculation_barrier): New function.
7694         * config/arm/arm.md (speculation_barrier): Call
7695         arm_emit_speculation_barrier for architectures that do not have
7696         DSB or ISB.
7697         (speculation_barrier_insn): Only match on Armv7 or later.
7699 2018-08-23  Richard Biener  <rguenther@suse.de>
7701         PR middle-end/87024
7702         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
7703         calls.
7705 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7707         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
7708         of single-vector TBLs.
7709         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
7710         one input is given.
7712 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7714         PR target/85910
7715         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
7716         aarch64_evpc_tbl guard.
7718 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7720         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
7721         behaviour.
7723 2018-08-22  Martin Sebor  <msebor@redhat.com>
7725         PR middle-end/87052
7726         * tree-pretty-print.c (pretty_print_string): Add argument.
7727         (dump_generic_node): Call to pretty_print_string with string size.
7729 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
7731         PR rtl-optimization/86771
7732         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
7733         of two SETs into those two SETs, one to be placed at i2, if that SETs
7734         destination is modified between i2 and i3.
7736 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7738         PR tree-optimization/86725
7739         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
7740         function.
7741         (vect_analyze_scalar_cycles_1): Check it.
7743 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7745         PR tree-optimization/86725
7746         * tree-vect-loop.c (vect_is_simple_reduction): When treating
7747         an outer loop phi as a double reduction, make sure that the
7748         single user of the phi result is an inner loop phi.
7750 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7752         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
7753         grouped stores with gaps to a strided group.
7755 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7757         * tree-vect-stmts.c (get_group_load_store_type)
7758         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
7759         first statement in a group.
7761 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7763         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
7764         the sequence used in gcc/gcc.c.
7766 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7768         PR other/704
7769         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
7770         building it.
7772 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7774         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
7775         Darwin10-specific unwinder-shim.
7776         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
7777         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
7778         New to cater for Darwin10 Rosetta.
7780 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7782         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
7783         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
7785 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7787         PR bootstrap/81033
7788         PR target/81733
7789         PR target/52795
7790         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
7791         (dwarf2out_switch_text_section): Generate a local label for the
7792         second function sub-section and apply it as the second FDE start
7793         label.
7794         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
7795         second sub-section start.
7797 2018-08-22  Richard Biener  <rguenther@suse.de>
7799         PR tree-optimization/86988
7800         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
7802 2018-08-22  Richard Biener  <rguenther@suse.de>
7804         PR tree-optimization/86945
7805         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
7807 2018-08-22  Alexandre Oliva <oliva@adacore.com>
7809         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
7810         a comment about how uses of r2 for .sdata2 come about.
7812 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
7814         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
7816 2018-08-21  Marek Polacek  <polacek@redhat.com>
7818         PR c++/86981, Implement -Wpessimizing-move.
7819         * doc/invoke.texi: Document -Wpessimizing-move.
7821 2018-08-21  Jan Hubicka  <jh@suse.cz>
7823         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
7824         * tree.h (is_redundant_typedef): Remove.
7825         * dwarf2out.c (is_redundant_typedef): Turn into static function.
7827 2018-08-21  Jan Hubicka  <jh@suse.cz>
7829         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
7830         when possible.
7832 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
7834         * expmed.c (extract_low_bits): Reject invalid subregs early.
7836 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7838         PR middle-end/86121
7839         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
7840         behaviour.
7842 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7844         * config/vxworks.h: Guard vxworks_asm_out_constructor and
7845         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
7846         * config/vxworks.c: Likewise.
7848 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7850         * config/vxworks.c: Set targetm.have_ctors_dtors
7851         if HAVE_INITFINI_ARRAY_SUPPORT.
7852         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
7853         if HAVE_INITFINI_ARRAY_SUPPORT.
7855 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7857         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
7858         default search path for VxWorks < 7.
7860 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7862         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
7863         (get_format_string): Refer to c_getstr.
7865 2018-08-21  Tom de Vries  <tdevries@suse.de>
7867         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7868         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7869         (debuginfo_early_stop): Declare.
7870         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7871         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7872         (debuginfo_early_stop): New function.
7873         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
7874         and debuginfo_early_stop.
7875         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
7876         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
7877         (general_init): Call debuginfo_early_init.
7878         (finalize): Call debuginfo_fini.
7879         (do_compile): Call debuginfo_init.
7880         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
7881         -fdump-early-debug.
7882         (@item -fdump-debug, @item -fdump-earlydebug): Add.
7884 2018-08-21  Tom de Vries  <tdevries@suse.de>
7886         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
7887         flag_dump_noaddr and flag_dump_unnumbered.
7889 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
7891         * wide-int-range.cc (wide_int_range_abs): New.
7892         (wide_int_range_order_set): Rename from wide_int_range_min_max.
7893         * wide-int-range.h (wide_int_range_abs): New.
7894         (wide_int_range_min_max): New.
7895         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
7896         case to call wide_int_range_abs.
7897         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
7898         (extract_range_from_abs_expr): Delete.
7900 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
7902         PR target/87033
7903         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
7904         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
7905         for indexed loads.
7907 2018-08-20  Nathan Sidwell  <nathan@acm.org>
7908             Jeff Law <law@redhat.com>
7910         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
7911         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
7913 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
7915         PR other/84889
7916         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
7917         (decl_attributes): Likewise.
7918         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
7919         instance.
7920         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7921         * diagnostic-core.h (class auto_diagnostic_group): New class.
7922         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
7923         (diagnostic_report_diagnostic): Handle the first diagnostics within
7924         a group.
7925         (emit_diagnostic): Add auto_diagnostic_group instance.
7926         (inform): Likewise.
7927         (inform_n): Likewise.
7928         (warning): Likewise.
7929         (warning_at): Likewise.
7930         (warning_n): Likewise.
7931         (pedwarn): Likewise.
7932         (permerror): Likewise.
7933         (error): Likewise.
7934         (error_n): Likewise.
7935         (error_at): Likewise.
7936         (sorry): Likewise.
7937         (fatal_error): Likewise.
7938         (internal_error): Likewise.
7939         (internal_error_no_backtrace): Likewise.
7940         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
7941         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
7942         * diagnostic.h (struct diagnostic_context): Add fields
7943         "diagnostic_group_nesting_depth",
7944         "diagnostic_group_emission_count", "begin_group_cb",
7945         "end_group_cb".
7946         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
7947         Add auto_diagnostic_group instance(s).
7948         (find_explicit_erroneous_behavior): Likewise.
7949         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
7950         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
7951         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
7952         (gimplify_va_arg_expr): Likewise.
7953         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
7954         (HSA_SORRY_AT): Likewise.
7955         * ipa-devirt.c (compare_virtual_tables): Likewise.
7956         (warn_odr): Likewise.
7957         * multiple_target.c (expand_target_clones): Likewise.
7958         * opts-common.c (cmdline_handle_error): Likewise.
7959         * reginfo.c (globalize_reg): Likewise.
7960         * substring-locations.c (format_warning_n_va): Likewise.
7961         * tree-inline.c (expand_call_inline): Likewise.
7962         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
7963         * tree-ssa-loop-niter.c
7964         (do_warn_aggressive_loop_optimizations): Likewise.
7965         * tree-ssa-uninit.c (warn_uninit): Likewise.
7966         * tree.c (warn_deprecated_use): Likewise.
7968 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
7970         PR target/87014
7971         * config/i386/i386.md (eh_return): Always update EH return
7972         address in word_mode.
7974 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
7976         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
7977         TARGET_SPLIT_COMPLEX_ARG is defined.
7979 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7981         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
7983 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7985         PR target/86984
7986         * expr.c (expand_assignment): Assert that bitpos is positive.
7987         (store_field): Likewise
7988         (expand_expr_real_1): Make sure that bitpos is positive.
7989         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
7990         integer overflow.
7992 2018-08-20  Nathan Sidwell  <nathan@acm.org>
7994         * Makefile.in (CPP_ID_DATA_H): Delete.
7995         (CPP_INTERNAL_H): Don't add it.
7996         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
7997         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
7999 2018-08-20  Richard Biener  <rguenther@suse.de>
8001         PR tree-optimization/78655
8002         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8003         pointer + offset nonnull if either operand is nonnull work.
8005 2018-08-20  Tom de Vries  <tdevries@suse.de>
8007         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8008         unless the referenced die describes the added property using
8009         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8010         Otherwise, add a DW_AT_location to the referenced die.
8012 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8014         PR target/86994
8015         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8016         register_operand when calling ix86_set_reg_reg_cost.
8017         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8018         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8019         predicate and to 1 otherwise.
8021 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8023         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8024         emit a diagnostic that it is not supported and reset the option.
8025         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8026         supported and consume the option.  (ASM_FINAL_SPEC): New.
8028 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8030         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8031         a sentence.
8033 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8035         C-SKY port: Documentation
8037         * doc/extend.texi (C-SKY Function Attributes): New section.
8038         * doc/invoke.texi (Option Summary): Add C-SKY options.
8039         (C-SKY Options): New section.
8040         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8042 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8043             Huibin Wang  <huibin_wang@c-sky.com>
8044             Sandra Loosemore  <sandra@codesourcery.com>
8045             Chung-Lin Tang  <cltang@codesourcery.com>
8047         C-SKY port: Backend implementation
8049         * config/csky/*: New.
8050         * common/config/csky/*: New.
8052 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8053             Huibin Wang  <huibin_wang@c-sky.com>
8054             Sandra Loosemore  <sandra@codesourcery.com>
8055             Chung-Lin Tang  <cltang@codesourcery.com>
8056             Andrew Jenner  <andrew@codesourcery.com>
8058         C-SKY port: Configury
8060         * config.gcc (csky-*-*): New.
8061         * configure.ac: Add csky to targets for dwarf2 debug_line support.
8062         * configure: Regenerated.
8064 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
8066         * dump-context.h: Include "dumpfile.h".
8067         (dump_context::dump_printf_va): Convert final param from va_list
8068         to va_list *.  Convert from ATTRIBUTE_PRINTF to
8069         ATTRIBUTE_GCC_DUMP_PRINTF.
8070         (dump_context::dump_printf_loc_va): Likewise.
8071         * dumpfile.c: Include "stringpool.h".
8072         (make_item_for_dump_printf_va): Delete.
8073         (make_item_for_dump_printf): Delete.
8074         (class dump_pretty_printer): New class.
8075         (dump_pretty_printer::dump_pretty_printer): New ctor.
8076         (dump_pretty_printer::emit_items): New member function.
8077         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8078         function.
8079         (dump_pretty_printer::emit_item): New member function.
8080         (dump_pretty_printer::stash_item): New member function.
8081         (dump_pretty_printer::format_decoder_cb): New member function.
8082         (dump_pretty_printer::decode_format): New member function.
8083         (dump_context::dump_printf_va): Reimplement in terms of
8084         dump_pretty_printer.
8085         (dump_context::dump_printf_loc_va): Convert final param from va_list
8086         to va_list *.
8087         (dump_context::begin_scope): Reimplement call to
8088         make_item_for_dump_printf.
8089         (dump_printf): Update for change to dump_printf_va.
8090         (dump_printf_loc): Likewise.
8091         (selftest::test_capture_of_dump_calls): Convert "stmt" from
8092         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
8093         with %T, %E, and %G.
8094         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8095         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8096         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8097         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8098         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8099         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8100         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8101         within a dump_printf_loc call to "%wu".
8102         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8103         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
8104         missing space after "=".
8105         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8106         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8107         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8108         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8109         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
8110         duplicate "vectorized" from message.
8112 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
8114         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8115         polyNxK_t element's TYPE_STRING_FLAG.
8117 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8119         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8120         (they were unnamed before).  Fix comments.
8122 2018-08-17  Nathan Sidwell  <nathan@acm.org>
8124         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8126 2018-08-17  Richard Biener  <rguenther@suse.de>
8128         PR tree-optimization/86841
8129         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8131 2018-08-17  Martin Liska  <mliska@suse.cz>
8133         * common.opt: Remove Warn, Init and Report for options with
8134         Ignore/Deprecated flag. Warning is done automatically for
8135         Deprecated flags.
8136         * config/i386/i386.opt: Likewise.
8137         * config/ia64/ia64.opt: Likewise.
8138         * config/rs6000/rs6000.opt: Likewise.
8139         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8140         Remove usage of flag_check_pointer_bounds.
8141         * lto-wrapper.c (merge_and_complain): Do not handle
8142         OPT_fcheck_pointer_bounds.
8143         (append_compiler_options): Likewise.
8144         * opt-functions.awk: Do not handle Deprecated.
8145         * optc-gen.awk: Check that Var, Report and Init are not
8146         used for an option with Ignore/Deprecated flag.
8147         * opts-common.c (decode_cmdline_option): Do not report
8148         CL_ERR_DEPRECATED.
8149         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8150         options.
8151         * opts.h (struct cl_option): Remove cl_deprecated flag.
8152         (CL_ERR_DEPRECATED): Remove error enum value.
8154 2018-08-17  Richard Biener  <rguenther@suse.de>
8156         PR middle-end/86505
8157         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8158         across a va-arg-pack using call adjust its return value accordingly.
8160 2018-08-16  Martin Sebor  <msebor@redhat.com>
8162         PR tree-optimization/86853
8163         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8164         (struct fmtresult): Add member and initialize it in ctors.
8165         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8166         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8167         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8168         (parse_directive): Handle %C same as %c.
8169         (sprintf_dom_walker::compute_format_length): Adjust.
8170         (is_call_safe): Adjust.
8172 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8174         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
8175         for determining how to count the elements.
8176         * builtins.h (c_strlen): Adjust prototype.
8177         * expr.c (string_constant): Add new parameter mem_size.
8178         Set *mem_size appropriately.
8179         * expr.h (string_constant): Adjust protoype.
8180         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8181         * gimple-fold.h (get_range_strlen): Adjust prototype.
8182         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8183         (format_string): Call get_string_length with eltsize.
8185 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8187         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8188         to emit the span, rather than setting it as the prefix.
8190 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8192         * diagnostic-show-locus.c (layout::start_annotation_line): Add
8193         "margin_char" parameter, defaulting to space.  Use it in place
8194         of pp_space for the initial part of the margin.
8195         (layout::print_leading_fixits): Use '+' when filling the margin
8196         of line-insertion fix-it hints.
8198 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8200         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8201         Delete.
8203 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8205         * config/rs6000/altivec.md: Don't set length attribute to the default
8206         value.
8207         * config/rs6000/darwin.md: Ditto.
8208         * config/rs6000/dfp.md: Ditto.
8209         * config/rs6000/htm.md: Ditto.
8210         * config/rs6000/rs6000.md: Ditto.
8211         * config/rs6000/sync.md: Ditto.
8212         * config/rs6000/vsx.md: Ditto.
8214 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8216         * config/rs6000/altivec.md: Don't set length attribute to the default
8217         value, for branch instructions.
8218         * config/rs6000/darwin.md: Ditto.
8219         * config/rs6000/rs6000.md: Ditto.
8221 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8223         * config/rs6000/rs6000.md (length): Always define as const_int 4.
8224         (unnamed conditional branch define_insn): Set length to 4 or 8
8225         depending on offset.
8226         (<bd>_<mode>): Similar, for alternative 0.
8227         (<bd>tf_<mode>): Ditto.
8229 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8231         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8233 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
8235         * doc/rtl.texi: Replace old RTX class names with new names.
8238 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
8240         * expmed.h (canonicalize_comparison): New declaration.
8241         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8242         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8243         * optabs.c (prepare_cmp_insn): Likewise.
8244         * rtl.h (unsigned_condition_p): New function which checks if a
8245         comparison operator is unsigned.
8247 2018-08-16  Nathan Sidwell  <nathan@acm.org>
8249         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8250         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8252 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8254         PR target/84711
8255         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8256         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8257         (mov<mov>): ..this and enable unconditionally.
8259 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8261         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8263 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
8265         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8266         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8267         "Common" with "Target".
8269 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8271         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8272         * doc/invoke.texi (mmitigate-rop): Remove.
8273         * config/i386/i386.c: Do not include "regrename.h".
8274         (ix86_rop_should_change_byte_p, reg_encoded_number)
8275         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8276         Remove.
8277         (ix86_reorg): Remove call to ix86_mitigate_rop.
8278         * config/i386/i386.md (attr "modrm_class"): Remove.
8279         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8280         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8281         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8283 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
8285         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8286         allow folding of mergeh() and mergel() for the float and double types.
8287         (fold_mergehl_helper): Rework to handle building a permute tree
8288         for float vectors.
8290 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8292         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8293         for TARGET_SSE.
8295 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
8297         * common.opt (fdiagnostics-show-labels): New option.
8298         * diagnostic-show-locus.c (class layout_range): Add field
8299         "m_label".
8300         (class layout): Add field "m_show_labels_p".
8301         (layout_range::layout_range): Add param "label" and use it to
8302         initialize m_label.
8303         (make_range): Pass in NULL for new "label" param of layout_range's
8304         ctor.
8305         (layout::layout): Initialize m_show_labels_p.
8306         (layout::maybe_add_location_range): Pass in loc_range->m_label
8307         when constructing layout_range instances.
8308         (struct line_label): New struct.
8309         (layout::print_any_labels): New member function.
8310         (layout::print_line): Call it if label-printing is enabled.
8311         (selftest::test_one_liner_labels): New test.
8312         (selftest::test_diagnostic_show_locus_one_liner): Call it.
8313         * diagnostic.c (diagnostic_initialize): Initialize
8314         context->show_labels_p.
8315         * diagnostic.h (struct diagnostic_context): Add field
8316         "show_labels_p".
8317         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8318         -fno-diagnostics-show-labels.
8319         * dwarf2out.c (gen_producer_string): Add
8320         OPT_fdiagnostics_show_labels to the ignored options.
8321         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
8322         param.
8323         (gcc_rich_location::maybe_add_expr): Likewise.
8324         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
8325         label" param, defaulting to NULL.
8326         (gcc_rich_location::add_expr): Add "label" param.
8327         (gcc_rich_location::maybe_add_expr): Likewise.
8328         (class text_range_label): New class.
8329         (class range_label_for_type_mismatch): New class.
8330         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
8331         of format_warning_va.
8332         (fmtwarn_n): Likewise for new params of format_warning_n_va.
8333         * lto-wrapper.c (merge_and_complain): Add
8334         OPT_fdiagnostics_show_labels to the "pick one setting" options.
8335         (append_compiler_options): Likewise to the dropped options.
8336         (append_diag_options): Likewise to the passed-on options.
8337         * opts.c (common_handle_option): Handle the new option.
8338         * selftest-diagnostic.c
8339         (test_diagnostic_context::test_diagnostic_context): Enable
8340         show_labels_p.
8341         * substring-locations.c: Include "gcc-rich-location.h".
8342         (format_warning_n_va): Add "fmt_label" and "param_label" params
8343         and use them as appropriate.
8344         (format_warning_va): Add "fmt_label" and "param_label" params,
8345         passing them on to format_warning_n_va.
8346         (format_warning_at_substring): Likewise.
8347         (format_warning_at_substring_n): Likewise.
8348         * substring-locations.h (format_warning_va): Add "fmt_label" and
8349         "param_label" params.
8350         (format_warning_n_va): Likewise.
8351         (format_warning_at_substring): Likewise.
8352         (format_warning_at_substring_n): Likewise.
8353         * toplev.c (general_init): Initialize global_dc->show_labels_p.
8355 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
8357         PR testsuite/86519
8358         * builtins.c (expand_builtin_memcmp): Do not expand the call
8359         when overflow is detected.
8361 2018-08-15  Martin Sebor  <msebor@redhat.com>
8363         PR tree-optimization/71625
8364         * config/aarch64/aarch64-builtins.c
8365         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
8367 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
8369         * config/s390/s390.c (s390_reorg): Remove loop.
8371 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8373         * config/darwin.c
8374          (darwin_function_switched_text_sections): Delete.
8375         * gcc/config/darwin.h
8376          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
8378 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8380         PR target/81685
8381         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
8382         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
8383         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
8385 2018-08-15  Martin Liska  <mliska@suse.cz>
8387         PR tree-optimization/86925
8388         * predict.c (expr_expected_value_1): When taking
8389         later predictor, assign also probability.
8390         Use fold_build2_initializer_loc in order to fold
8391         the expression in -frounding-math.
8393 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
8395         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
8396         patterns.
8397         (expand_vec_perm_1): Try the new method.
8399 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
8401         PR target/86547
8402         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
8403         Check whether lra_live_max_point is 0 before dividing.
8405 2018-08-14  Martin Sebor  <msebor@redhat.com>
8407         PR tree-optimization/86650
8408         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
8409         (vrp_prop::check_mem_ref): Same.
8411 2018-08-13  Liu Hao <lh_mouse@126.com>
8413         * pretty-print.c (eat_esc_sequence): Swap the foreground and
8414         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
8415         and clear it thereafter, as it only works for DBCS.
8417 2018-08-13  Liu Hao <lh_mouse@126.com>
8419         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
8420         handle returned by _get_osf_handle().
8422 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8424         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
8425         for folding vec_perm.
8427 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8429         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
8430         Add support for gimple-folding of vec_pack() and vec_unpack()
8431         intrinsics.
8433 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
8435         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
8436         vec_xst variants to the list.
8437         (rs6000_gimple_fold_builtin): Add support for folding unaligned
8438         vector loads and stores.
8440 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
8442         * config.gcc (rs6000-ibm-aix4.x): Delete.
8443         (rs6000-ibm-aix5.1): Delete.
8444         (rs6000-ibm-aix5.2): Delete.
8445         (rs6000-ibm-aix5.3): Delete.
8446         * config/rs6000/aix43.h: Delete.
8447         * config/rs6000/aix51.h: Delete.
8448         * config/rs6000/aix52.h: Delete.
8449         * config/rs6000/t-aix43: Delete.
8451 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
8453         * config/s390/s390.c (s390_decompose_constant_pool_ref):
8454         New function.
8455         (s390_decompose_address): Factor out constant pool ref
8456         decomposition.
8458 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8460         * config/nds32/nds32-predicates.c
8461         (nds32_can_use_bclr_p): Change return type as bool.
8462         (nds32_can_use_bset_p): Ditto.
8463         (nds32_can_use_btgl_p): Ditto.
8464         (nds32_can_use_bitci_p): Ditto.
8465         * config/nds32/nds32-protos.h
8466         (nds32_can_use_bclr_p): Change declaration.
8467         (nds32_can_use_bset_p): Ditto.
8468         (nds32_can_use_btgl_p): Ditto.
8469         (nds32_can_use_bitci_p): Ditto.
8471 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8473         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
8474         Support -msched-prolog-epilog option.
8475         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
8477 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8479         * common/config/nds32/nds32-common.c
8480         (nds32_option_optimization_table): Enalbe -malways-align.
8482 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8484         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
8485         extra_headers.
8486         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
8487         OPT_misr_secure_ case.
8488         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
8489         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
8490         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
8491         secure attribute.
8492         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
8493         (nds32_isr_info): New field security_level.
8494         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
8495         * config/nds32/nds32.md (return_internal): Consider critical attribute.
8496         * config/nds32/nds32.opt (misr-secure): New option.
8497         * config/nds32/nds32_init.inc: New file.
8498         * config/nds32/nds32_isr.h: New file.
8500 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
8502         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
8503         Update comment for atomic instructions.
8504         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
8505         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
8506         Remove.
8507         (atomic_loaddi): Revise fence expansion to only emit fence prior to
8508         load for __ATOMIC_SEQ_CST model.
8509         (atomic_loaddi_1): Remove float register target.
8510         (atomic_storedi): Handle CONST_INT values.
8511         (atomic_storedi_1): Remove float register source.  Add special case
8512         for zero value.
8513         (memory_barrier): New expander and insn.
8515 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
8517         PR tree-optimization/86835
8518         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
8519         new_stmt after def_gsi, make sure to insert new_square_stmt after
8520         that stmt, not 2 stmts before it.
8522 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
8524         PR target/82418
8525         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
8526         instead of SWI48.
8528 2018-08-10  Martin Liska  <mliska@suse.cz>
8530         PR target/83610
8531         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
8532         function type.
8533         * builtins.c (expand_builtin_expect_with_probability):
8534         New function.
8535         (expand_builtin_expect_with_probability): New function.
8536         (build_builtin_expect_predicate): Add new argumnet probability
8537         for BUILT_IN_EXPECT_WITH_PROBABILITY.
8538         (fold_builtin_expect):
8539         (fold_builtin_2):
8540         (fold_builtin_3):
8541         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
8542         * builtins.h (fold_builtin_expect): Set new argument.
8543         * doc/extend.texi: Document __builtin_expect_with_probability.
8544         * doc/invoke.texi: Likewise.
8545         * gimple-fold.c (gimple_fold_call): Pass new argument.
8546         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
8547         also BUILT_IN_EXPECT_WITH_PROBABILITY.
8548         * predict.c (get_predictor_value): New function.
8549         (expr_expected_value): Add new argument probability. Assume
8550         that predictor and probability are always non-null.
8551         (expr_expected_value_1): Likewise.  For __builtin_expect and
8552         __builtin_expect_with_probability set probability.  Handle
8553         combination in binary expressions.
8554         (tree_predict_by_opcode): Simplify code by simply calling
8555         get_predictor_value.
8556         (pass_strip_predict_hints::execute): Add handling of
8557         BUILT_IN_EXPECT_WITH_PROBABILITY.
8558         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
8559         new predictor.
8560         * tree.h (DECL_BUILT_IN_P): New function.
8562 2018-08-10  Martin Liska  <mliska@suse.cz>
8564         PR tree-optimization/85799
8565         * passes.def: Add argument for pass_strip_predict_hints.
8566         * predict.c (class pass_strip_predict_hints): Add new argument
8567         early_p.
8568         (strip_predictor_early): New function.
8569         (pass_strip_predict_hints::execute): Call the function to
8570         strip predictors.
8571         (strip_predict_hints): New function.
8572         * predict.def: Fix comment.
8574 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
8576         * Makefile.in: Clarify which tm.texi to copy over to assert the
8577         right to grant a GFDL license for all.
8579 2018-08-09  Jeff Law  <law@redhat.com>
8581         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
8582         unused variable.
8584 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
8586         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
8587         prototype.
8589 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8591         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
8592         reductions for variable-length vectors.
8594 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
8596         PR other/84889
8597         * common.opt (fdiagnostics-show-line-numbers): New option.
8598         * diagnostic-show-locus.c (class layout): Add fields
8599         "m_show_line_numbers_p" and "m_linenum_width";
8600         (num_digits): New function.
8601         (test_num_digits): New function.
8602         (layout::layout): Initialize new fields.  Update m_x_offset
8603         logic to handle any left margin.
8604         (layout::print_source_line): Print line number when requested.
8605         (layout::start_annotation_line): New member function.
8606         (layout::print_annotation_line): Call it.
8607         (layout::print_leading_fixits): Likewise.
8608         (layout::print_trailing_fixits): Likewise.  Update calls to
8609         move_to_column for new parameter.
8610         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
8611         it to potentially call start_annotation_line.
8612         (layout::show_ruler): Call start_annotation_line.
8613         (selftest::test_line_numbers_multiline_range): New selftest.
8614         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
8615         and selftest::test_line_numbers_multiline_range.
8616         * diagnostic.c (diagnostic_initialize): Initialize
8617         show_line_numbers_p.
8618         * diagnostic.h (struct diagnostic_context): Add field
8619         "show_line_numbers_p".
8620         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8621         -fno-diagnostics-show-line-numbers.
8622         * dwarf2out.c (gen_producer_string): Add
8623         OPT_fdiagnostics_show_line_numbers to the ignored options.
8624         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
8625         one setting" options.
8626         (append_compiler_options): Likewise to the dropped options.
8627         (append_diag_options): Likewise to the passed-on options.
8628         * opts.c (common_handle_option): Handle the new option.
8629         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
8631 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8633         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
8634         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
8635         third argument of type "const signed char" to descriptions of
8636         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
8637         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
8638         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
8639         __builtin_bcdsub_ov functions.
8641 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8643         PR tree-optimization/86858
8644         * tree-vect-loop.c (vect_is_simple_reduction): Restore
8645         flow_bb_inside_loop_p calls.
8647 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8649         PR tree-optimization/86871
8650         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
8651         instead of gimple_assign_lhs.
8653 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
8655         PR target/86887
8656         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
8657         register constraint to operand 0.
8658         (add<mode>3_carryinC): Likewise.
8659         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
8661 2018-08-09  Martin Liska  <mliska@suse.cz>
8663         PR c/86895
8664         * common.opt: Remove extra line.
8666 2018-08-09  Martin Liska  <mliska@suse.cz>
8668         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
8669         at the end of a line, make first letter capital and end up
8670         a sentence with a dot.
8671         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
8672         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
8673         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
8674         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
8675         (PARAM_MAX_ISL_OPERATIONS): Likewise.
8676         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
8677         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
8678         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
8679         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
8680         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
8681         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
8682         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
8683         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
8684         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
8685         (PARAM_TREE_REASSOC_WIDTH): Likewise.
8686         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
8687         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
8688         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
8690 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
8692         PR target/84332
8693         * config/s390/s390.c (s390_option_override_internal): Reduce the
8694         stack-clash-protection-probe-interval param if it would be too big
8695         for z900.
8697 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
8699         PR target/46179
8700         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
8701         * config/m68k/m68k.c (handle_move_double): Don't call
8702         m68k_final_prescan_insn.
8703         (m68k_adjust_decorated_operand): Renamed from
8704         m68k_final_prescan_insn, remove first and third operand and
8705         simplify.
8706         (print_operand): Call it.
8707         (print_operand_address): Call it.
8709 2018-08-08  Nathan Sidwell  <nathan@acm.org>
8711         * diagnostic.c (diagnostic_report_current_module): Use
8712         linemap_included_from & linemap_included_from_linemap.
8714 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
8716         * config/aarch64/aarch64-cores.def: Add phecda core.
8717         * config/aarch64/aarch64-tune.md: Regenerate.
8718         * doc/invoke.texi: Add phecda core.
8720 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8722         PR target/85295
8723         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
8724         definitions.
8725         * config/s390/s390.md ("movti"): Add more alternatives for
8726         constant to GPR copies.
8728 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8730         * config/s390/s390.c: Fix whitespace damage throughout the file.
8731         * config/s390/s390.h: Likewise.
8732         * config/s390/tpf.h: Likewise.
8734 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8736         * config/s390/s390.c (s390_loadrelative_operand_p):
8737         Remove TARGET_CPU_ZARCH usages.
8738         (s390_rtx_costs): Likewise.
8739         (s390_legitimate_constant_p): Likewise.
8740         (s390_cannot_force_const_mem): Likewise.
8741         (legitimate_reload_constant_p): Likewise.
8742         (s390_preferred_reload_class): Likewise.
8743         (legitimize_pic_address): Likewise.
8744         (legitimize_tls_address): Likewise.
8745         (s390_split_branches): Removed.
8746         (s390_add_execute): Removed.
8747         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
8748         (s390_mainpool_start): Likewise.
8749         (s390_mainpool_finish): Likewise.
8750         (s390_mainpool_cancel): Removed.
8751         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
8752         (s390_chunkify_cancel): Likewise.
8753         (s390_return_addr_rtx): Likewise.
8754         (s390_register_info): Remove split_branches_pending_p uages.
8755         (s390_optimize_register_info): Likewise.
8756         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
8757         split_branches_pending_p usages.
8758         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
8759         (s390_load_got): Likewise.
8760         (s390_expand_split_stack_prologue): Likewise.
8761         (output_asm_nops): Likewise.
8762         (s390_function_profiler): Likewise.
8763         (s390_emit_call): Likewise.
8764         (s390_conditional_register_usage): Likewise.
8765         (s390_optimize_prologue): Likewise.
8766         (s390_reorg): Remove TARGET_CPU_ZARCH and
8767         split_branches_pending_p usages.
8768         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
8769         usages.
8770         (s390_output_indirect_thunk_function): Likewise.
8771         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
8772         (TARGET_CPU_ZARCH_P): Removed.
8773         (struct machine_function): Remove split_branches_pending_p.
8774         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
8776 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8778         * common/config/s390/s390-common.c (processor_flags_table):
8779         Remove flags.
8780         * config.gcc: Remove with_arch/with_tune support.
8781         * config/s390/2064.md: Remove cpu attribute comparisons.
8782         * config/s390/driver-native.c (s390_host_detect_local_cpu):
8783         Remove MTN.
8784         * config/s390/linux.h (ASM_SPEC):
8785         Remove -march support.
8786         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
8787         Use a table to get an arch level.
8788         * config/s390/s390-opts.h (enum processor_type):
8789         Remove enum values.
8790         * config/s390/s390.c
8791         (processor_table): Remove entries, add arch_level values.
8792         (s390_issue_rate): Remove cases.
8793         (s390_option_override): Adjust
8794         s390_option_override_internal() call.
8795         (s390_option_override_internal): Remove deprecation warning.
8796         (s390_valid_target_attribute_tree): Adjust
8797         s390_option_override_internal() call.
8798         * config/s390/s390.h (struct s390_processor):
8799         Share with s390-c.c, add arch_level field.
8800         * config/s390/s390.md:
8801         Remove occurrences in cpu attribute.
8802         * config/s390/s390.opt: Remove -march/-mtune support.
8803         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
8804         * doc/invoke.texi: Remove deprecation warning.
8806 2018-08-08  Luis Machado  <luis.machado@linaro.org>
8808         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
8809         global.
8810         (qdf24xx_tunings): Set vector cost structure to
8811         qdf24xx_vector_cost.
8813         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
8814         <register_sextend>: Set to 3.
8816 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
8818         PR target/86838
8819         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
8820         * config/aarch64/aarch64-simd.md
8821         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
8822         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
8823         (aarch64_frecpx<mode>): ...this new pattern.
8824         * config/aarch64/aarch64-simd-builtins.def: Remove comment
8825         about aarch64_frecp<FRECP:frecp_suffix><mode>.
8827 2018-08-07  Martin Liska  <mliska@suse.cz>
8829         PR middle-end/83023
8830         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
8831         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
8832         * predict.def (PRED_MALLOC_NONNULL): New predictor.
8833         * doc/extend.texi: Document that malloc attribute adds
8834         hit to compiler.
8836 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
8838         PR target/86785
8839         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8840         Define to speculation_safe_value_not_needed.
8842 2018-08-06  Jeff Law  <law@redhat.com>
8844         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
8845         the vr_values instance to cprop_into_stmt.
8846         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
8847         (cprop_operand): Also query EVRP to determine if OP is a constant.
8849 2018-08-06  Nathan Sidwell  <nathan@acm.org>
8851         * diagnostic.c (diagnostic_report_current_module): Reroll
8852         included-at loop.  Translate text.
8854 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
8856         * function-tests.c (selftest::test_expansion_to_rtl): Call
8857         free_after_compilation.
8859 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8861         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
8863 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
8865         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
8866         loops with memory block operations from getting unrolled.
8868 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
8870         PR target/86807
8871         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8872         Define to speculation_safe_value_not_needed.
8874 2018-08-06  Jeff Law  <law@redhat.com>
8876         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
8877         assert.
8879 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8881         PR target/86662
8882         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
8883         with all enabled __intN types.
8885         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
8887 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8889         * alias.c (record_set): Check for clobber high.
8890         * cfgexpand.c (expand_gimple_stmt): Likewise.
8891         * combine-stack-adj.c (single_set_for_csa): Likewise.
8892         * combine.c (find_single_use_1): Likewise.
8893         (set_nonzero_bits_and_sign_copies): Likewise.
8894         (get_combine_src_dest): Likewise.
8895         (is_parallel_of_n_reg_sets): Likewise.
8896         (try_combine): Likewise.
8897         (record_dead_and_set_regs_1): Likewise.
8898         (reg_dead_at_p_1): Likewise.
8899         (reg_dead_at_p): Likewise.
8900         * dce.c (deletable_insn_p): Likewise.
8901         (mark_nonreg_stores_1): Likewise.
8902         (mark_nonreg_stores_2): Likewise.
8903         * df-scan.c (df_find_hard_reg_defs): Likewise.
8904         (df_uses_record): Likewise.
8905         (df_get_call_refs): Likewise.
8906         * dwarf2out.c (mem_loc_descriptor): Likewise.
8907         * haifa-sched.c (haifa_classify_rtx): Likewise.
8908         * ira-build.c (create_insn_allocnos): Likewise.
8909         * ira-costs.c (scan_one_insn): Likewise.
8910         * ira.c (equiv_init_movable_p): Likewise.
8911         (rtx_moveable_p): Likewise.
8912         (interesting_dest_for_shprep): Likewise.
8913         * jump.c (mark_jump_label_1): Likewise.
8914         * postreload-gcse.c (record_opr_changes): Likewise.
8915         * postreload.c (reload_cse_simplify): Likewise.
8916         (struct reg_use): Add source expr.
8917         (reload_combine): Check for clobber high.
8918         (reload_combine_note_use): Likewise.
8919         (reload_cse_move2add): Likewise.
8920         (move2add_note_store): Likewise.
8921         * print-rtl.c (print_pattern): Likewise.
8922         * recog.c (decode_asm_operands): Likewise.
8923         (store_data_bypass_p): Likewise.
8924         (if_test_bypass_p): Likewise.
8925         * regcprop.c (kill_clobbered_value): Likewise.
8926         (kill_set_value): Likewise.
8927         * reginfo.c (reg_scan_mark_refs): Likewise.
8928         * reload1.c (maybe_fix_stack_asms): Likewise.
8929         (eliminate_regs_1): Likewise.
8930         (elimination_effects): Likewise.
8931         (mark_not_eliminable): Likewise.
8932         (scan_paradoxical_subregs): Likewise.
8933         (forget_old_reloads_1): Likewise.
8934         * reorg.c (find_end_label): Likewise.
8935         (try_merge_delay_insns): Likewise.
8936         (redundant_insn): Likewise.
8937         (own_thread_p): Likewise.
8938         (fill_simple_delay_slots): Likewise.
8939         (fill_slots_from_thread): Likewise.
8940         (dbr_schedule): Likewise.
8941         * resource.c (update_live_status): Likewise.
8942         (mark_referenced_resources): Likewise.
8943         (mark_set_resources): Likewise.
8944         * rtl.c (copy_rtx): Likewise.
8945         * rtlanal.c (reg_referenced_p): Likewise.
8946         (single_set_2): Likewise.
8947         (noop_move_p): Likewise.
8948         (note_stores): Likewise.
8949         * sched-deps.c (sched_analyze_reg): Likewise.
8950         (sched_analyze_insn): Likewise.
8952 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8954         * cse.c (invalidate_reg): New function extracted from...
8955         (invalidate): ...here.
8956         (canonicalize_insn): Check for clobber high.
8957         (invalidate_from_clobbers): invalidate clobber highs.
8958         (invalidate_from_sets_and_clobbers): Likewise.
8959         (count_reg_usage): Check for clobber high.
8960         (insn_live_p): Likewise.
8961         * cselib.c (cselib_expand_value_rtx_1):Likewise.
8962         (cselib_invalidate_regno): Check for clobber in setter.
8963         (cselib_invalidate_rtx): Pass through setter.
8964         (cselib_invalidate_rtx_note_stores):
8965         (cselib_process_insn): Check for clobber high.
8966         * cselib.h (cselib_invalidate_rtx): Add operand.
8968 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8970         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
8971         (mark_not_eliminable): Likewise.
8972         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
8973         * lra-lives.c (process_bb_lives): Check for clobber high.
8974         * lra.c (new_insn_reg): Remember clobber highs.
8975         (collect_non_operand_hard_regs): Check for clobber high.
8976         (lra_set_insn_recog_data): Likewise.
8977         (add_regs_to_insn_regno_info): Likewise.
8978         (lra_update_insn_regno_info): Likewise.
8980 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8982         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
8983         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
8985 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8987         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
8988         (copy_insn_1): Likewise.
8989         (gen_hard_reg_clobber_high): New gen function.
8990         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
8991         * genemit.c (gen_exp): Likewise.
8992         (gen_emit_seq): Pass through info.
8993         (gen_insn): Check for CLOBBER_HIGH.
8994         (gen_expand): Pass through info.
8995         (gen_split): Likewise.
8996         (output_add_clobbers): Likewise.
8997         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
8998         (remove_clobbers): Likewise.
8999         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9001 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9003         * doc/rtl.texi (clobber_high): Add.
9004         (parallel): Add in clobber high
9005         * rtl.c (rtl_check_failed_code3): Add function.
9006         * rtl.def (CLOBBER_HIGH): Add expression.
9007         * rtl.h (RTL_CHECKC3): Add macro.
9008         (rtl_check_failed_code3): Add declaration.
9009         (XC3EXP): Add macro.
9011 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9013         PR target/86386
9014         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9015         cfun->machine->max_used_stack_alignment if needed.
9017 2018-08-04  Martin Sebor  <msebor@redhat.com>
9019         PR tree-optimization/86571
9020         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9021         NaN output to 4.
9023 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9025         PR target/86799
9026         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9027         Define.
9029 2018-08-03  Jeff Law  <law@redhat.com>
9031         PR target/86795
9032         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9033         Define to speculation_safe_value_not_needed.
9035 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9037         * doc/gcov.texi (-x): Remove duplicate "to".
9038         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9039         (-Wif-not-aligned): Remove duplicate "is".
9040         (-flto): Remove duplicate "the".
9041         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9042         duplicate "v5.00.b".
9043         (MSP430 Options): Remove duplicate "and" from the description
9044         of "-mgprel-sec=regexp".
9045         (x86 Options): Remove duplicate copies of "vmldLog102" and
9046         vmlsLog104 from description of "-mveclibabi=type".
9048 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9050         * internal-fn.h (first_commutative_argument): Declare.
9051         * internal-fn.c (first_commutative_argument): New function.
9052         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9053         restrictions for pattern statements.  Use first_commutative_argument
9054         to look for commutative operands in calls to internal functions.
9056 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
9058         * Makefile.in (wide-int-range.o): New.
9059         * tree-vrp.c: Move all the wide_int_* functions to...
9060         * wide-int-range.cc: ...here.
9061         * tree-vrp.h: Move all the wide_int_* prototypes to...
9062         * wide-int-range.h: ...here.
9064 2018-08-03  Tom de Vries  <tdevries@suse.de>
9066         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9067         UI_NONE.
9068         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9069         * except.c (output_function_exception_table): Do early exit if
9070         targetm_common.except_unwind_info (&global_options) == UI_NONE.
9072 2018-08-03  Martin Liska  <mliska@suse.cz>
9074         * predict.c (dump_prediction): Change to 2 digits
9075         in fraction part.
9077 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9079         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9080         neon_dup_q to...
9081         (falkor_am_1_gtov_gtov): ... a new insn reservation.
9083 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
9085         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9086         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9087         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9088         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9089         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9091 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9093         * diagnostic-show-locus.c (diagnostic_show_locus): Use
9094         pp_take_prefix when saving the existing prefix.
9095         * diagnostic.c (diagnostic_append_note): Likewise.
9096         * langhooks.c (lhd_print_error_function): Likewise.
9097         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9098         param's type.  Free the existing prefix.
9099         (pp_take_prefix): New function.
9100         (pretty_printer::pretty_printer): Drop the prefix parameter.
9101         Rename the length parameter to match the comment.
9102         (pretty_printer::~pretty_printer): Free the prefix.
9103         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9104         parameter.
9105         (struct pretty_printer): Drop the "const" from "prefix" field's
9106         type and clarify memory management.
9107         (pp_set_prefix): Drop the "const" from the 2nd param.
9108         (pp_take_prefix): New decl.
9110 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
9112         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9113         for word_mode_ok here instead of passing as argument.
9114         (expand_block_compare): Change select_block_compare_mode() call.
9115         (expand_strncmp_gpr_sequence): New function.
9116         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9118 2018-08-02  Jeff Law  <law@redhat.com>
9120         PR target/86790
9121         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9122         Define to speculation_safe_value_not_needed.
9124         PR target/86784
9125         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9126         Define to speculation_safe_value_not_needed.
9128 2018-08-02  Tom de Vries  <tdevries@suse.de>
9130         PR target/86660
9131         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9132         function.  Return UI_TARGET unconditionally.
9133         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9134         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9136 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9138         * genemit.c (print_overload_test): Fix typo.
9140 2018-08-02  Richard Biener  <rguenther@suse.de>
9142         PR tree-optimization/86816
9143         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9144         which checks for value availability before querying it.
9145         (gvn_uses_equal): Use it.
9146         (same_succ_hash): Likewise.
9147         (gimple_equal_p): Likewise.
9149 2018-08-02  Nick Clifton  <nickc@redhat.com>
9151         PR target/86813
9152         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9153         Define to speculation_safe_value_not_needed.
9155         PR target/86810
9156         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9157         Define to speculation_safe_value_not_needed.
9159         PR target/86810
9160         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9161         Define to speculation_safe_value_not_needed.
9163         PR target/86803
9164         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9165         Define to speculation_safe_value_not_needed.
9167         PR target/86797
9168         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9169         Define to speculation_safe_value_not_needed.
9171         PR target/86791
9172         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9173         Define to speculation_safe_value_not_needed.
9175         PR target/86789
9176         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9177         Define to speculation_safe_value_not_needed.
9179         PR target/86787
9180         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9181         Define to speculation_safe_value_not_needed.
9183         PR target/86782
9184         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9185         speculation_safe_value_not_needed.
9187         PR target/86781
9188         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9189         to speculation_safe_value_not_needed.
9191 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9193         * doc/md.texi: Expand the documentation of instruction names
9194         to mention port-local uses.  Document '@' in pattern names.
9195         * read-md.h (overloaded_instance, overloaded_name): New structs.
9196         (mapping): Declare.
9197         (md_reader::handle_overloaded_name): New member function.
9198         (md_reader::get_overloads): Likewise.
9199         (md_reader::m_first_overload): New member variable.
9200         (md_reader::m_next_overload_ptr): Likewise.
9201         (md_reader::m_overloads_htab): Likewise.
9202         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9203         m_next_overload_ptr and m_overloads_htab.
9204         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9205         (get_mode_token, get_code_token, get_int_token): New functions.
9206         (map_attr_string): Add an optional argument that passes back
9207         the associated iterator.
9208         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9209         (md_reader::handle_overloaded_name, add_overload_instance): New
9210         functions.
9211         (apply_iterators): Handle '@' names.  Report an error if '@'
9212         is used without iterators.
9213         (initialize_iterators): Initialize the new iterator_group fields.
9214         * genopinit.c (handle_overloaded_code_for)
9215         (handle_overloaded_gen): New functions.
9216         (main): Use them to print declarations of maybe_code_for_* and
9217         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9218         * genemit.c (print_overload_arguments, print_overload_test)
9219         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9220         (main): Use it to print definitions of maybe_code_for_* and
9221         maybe_gen_* functions.
9222         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9223         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9224         instead of explicit mode checks.
9225         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9226         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9227         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9228         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9229         (aarch64_expand_compare_and_swap): Likewise
9230         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9231         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9232         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9233         (aarch64_constant_pool_reload_icode): Delete.
9234         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9235         instead of aarch64_constant_pool_reload_icode.  Use
9236         code_for_aarch64_reload_mov instead of explicit mode checks.
9237         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9238         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9239         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9240         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9241         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9242         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9243         (aarch64_atomic_load_op_code): Delete.
9244         (aarch64_emit_atomic_load_op): Likewise.
9245         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9246         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
9247         instead of aarch64_emit_atomic_load_op.
9248         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9249         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9250         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9251         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9252         character before the pattern name.
9253         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9254         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9255         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9256         (aarch64_frecps<mode>): Likewise.
9257         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9258         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9259         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9260         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9261         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9263 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9265         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9266         Allow HFmode constants if TARGET_FP_F16INST.
9268 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
9270         PR target/86014
9271         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9272         No longer check last store for clobber of address register.
9274 2018-08-02  Martin Liska  <mliska@suse.cz>
9276         PR gcov-profile/86817
9277         * gcov.c (process_all_functions): New function.
9278         (main): Call it.
9279         (process_file): Move functions processing to
9280         process_all_functions.
9282 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9284         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9285         "const" to the "gimple *" and "rtx_insn *" parameters.
9286         * dumpfile.h (dump_user_location_t::dump_user_location_t):
9287         Likewise.
9288         (dump_location_t::dump_location_t): Likewise.
9290 2018-08-01  Martin Sebor  <msebor@redhat.com>
9292         PR tree-optimization/86650
9293         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9294         rather than a "gcall *".  Directly pass the data of interest
9295         to percent_K_format, rather than building a temporary CALL_EXPR
9296         to hold it.
9297         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9298         (gimple_fold_builtin_strncat): Adjust.
9299         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9300         gcall* argument with gimple*.
9301         * gimple-ssa-warn-restrict.c (check_call): Same.
9302         (wrestrict_dom_walker::before_dom_children): Same.
9303         (builtin_access::builtin_access): Same.
9304         (check_bounds_or_overlap): Same
9305         (maybe_diag_overlap): Same.
9306         (maybe_diag_offset_bounds): Same.
9307         * tree-diagnostic.c (default_tree_printer): Move usage of
9308         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9309         to this callsite.
9310         * tree-pretty-print.c (percent_K_format): Add argument.
9311         * tree-pretty-print.h: Add argument.
9312         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9313         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
9314         (maybe_diag_stxncpy_trunc): Same.
9315         (handle_builtin_stxncpy): Same.
9316         (handle_builtin_strcat): Same.
9318 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9320         * match.pd: Optimise pointer range checks.
9322 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9324         PR tree-optimization/86758
9325         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
9326         to remove pattern statements.
9328 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9330         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
9331         result of dfs_enumerate_from when constructing stmt_vec_infos,
9332         instead of additionally calling get_loop_body.
9334 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9336         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
9337         parameter.
9338         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
9339         When creating an iv, assert that the step is not known to be zero.
9340         (vect_setup_realignment): Update call accordingly.
9341         * tree-vect-stmts.c (vectorizable_store): Likewise.
9342         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
9344 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9346         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
9347         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
9348         (vectorizable_reduction): Likewise.
9349         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9350         (vect_detect_hybrid_slp_stmts): Likewise.
9351         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9353 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
9355         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
9356         (wide_int_set_zero_nonzero_bits): ...this.
9357         (zero_nonzero_bits_from_vr): Rename to...
9358         (vrp_set_zero_nonzero_bits): ...this.
9359         (extract_range_from_multiplicative_op_1): Abstract wide int
9360         code...
9361         (wide_int_range_multiplicative_op): ...here.
9362         (extract_range_from_binary_expr_1): Extract wide int binary
9363         operations into their own functions.
9364         (wide_int_range_lshift): New.
9365         (wide_int_range_can_optimize_bit_op): New.
9366         (wide_int_range_shift_undefined_p): New.
9367         (wide_int_range_bit_xor): New.
9368         (wide_int_range_bit_ior): New.
9369         (wide_int_range_bit_and): New.
9370         (wide_int_range_trunc_mod): New.
9371         (extract_range_into_wide_ints): New.
9372         (vrp_shift_undefined_p): New.
9373         (extract_range_from_multiplicative_op): New.
9374         (vrp_can_optimize_bit_op): New.
9375         * tree-vrp.h (value_range::dump): New.
9376         (wide_int_range_multiplicative_op): New.
9377         (wide_int_range_lshift):New.
9378         (wide_int_range_shift_undefined_p): New.
9379         (wide_int_range_bit_xor): New.
9380         (wide_int_range_bit_ior): New.
9381         (wide_int_range_bit_and): New.
9382         (wide_int_range_trunc_mod): New.
9383         (zero_nonzero_bits_from_bounds): Rename to...
9384         (wide_int_set_zero_nonzero_bits): ...this.
9385         (zero_nonzero_bits_from_vr): Rename to...
9386         (vrp_set_zero_nonzero_bits): ...this.
9387         (range_easy_mask_min_max): Rename to...
9388         (wide_int_range_can_optimize_bit_op): this.
9389         * vr-values.c (simplify_bit_ops_using_ranges): Rename
9390         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
9392 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9394         * tree-vectorizer.h (vect_orig_stmt): New function.
9395         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
9396         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
9397         (vect_create_epilog_for_reduction): Likewise.
9398         (vectorizable_live_operation): Likewise.
9399         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9400         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
9401         * tree-vect-stmts.c (vectorizable_call): Likewise.
9402         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
9404 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9406         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
9407         argument.
9408         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
9409         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
9410         (vect_transform_loop): Likewise.
9411         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9413 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9415         * tree-vectorizer.h (vect_schedule_slp): Return void.
9416         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9417         (vect_schedule_slp): Likewise.
9419 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9421         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
9422         argument.
9423         (vect_transform_loop): Update calls accordingly.  Schedule SLP
9424         instances before the main loop, if any exist.
9426 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9428         PR tree-optimization/86749
9429         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
9430         If the lhs is used in a COND_EXPR, check that it is being used
9431         as the "then" or "else" value.
9433 2018-08-01  Tom de Vries  <tdevries@suse.de>
9435         PR target/86800
9436         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9437         speculation_safe_value_not_needed.
9439 2018-08-01  Richard Biener  <rguenther@suse.de>
9441         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
9442         as base and offset.
9444 2018-08-01  Martin Liska  <mliska@suse.cz>
9446         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
9447         format how successful transformation is dumped.
9448         (gimple_mod_pow2_value_transform): Likewise.
9449         (gimple_mod_subtract_transform): Likewise.
9450         (gimple_stringops_transform): Likewise.
9452 2018-08-01  Martin Liska  <mliska@suse.cz>
9454         PR value-prof/35543
9455         * value-prof.c (interesting_stringop_to_profile_p):
9456         Simplify the code and add BUILT_IN_MEMMOVE.
9457         (gimple_stringops_transform): Likewise.
9459 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
9461         * config/aarch64/aarch64-simd.md
9462         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
9463         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
9464         use GPI iterator instead of SI mode.
9466 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9468         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
9469         rs6000_speculation_barrier.
9470         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
9471         new barrier pattern name.
9473 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9475         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
9476         (speculation_barrier): New insn.
9478 2018-08-01  Richard Biener  <rguenther@suse.de>
9480         PR bootstrap/86724
9481         * graphite.h: Include isl/id.h and isl/space.h to allow build
9482         with ISL 0.20.
9484 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
9486         PR target/86651
9487         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9488         mode for COFF targets.
9489         * defaults.h (TARGET_COFF): Define.
9490         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
9491         TARGET_COFF): Define.
9492         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
9493         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
9494         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
9496 2018-07-31  Alexandre Oliva <oliva@adacore.com>
9498         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
9499         * gimple-streamer-out.c (output_bb): Save it.
9500         * lto-streamer-in.c (input_struct_function_base): Restore
9501         instance discriminator if available.  Create map on demand.
9502         * lto-streamer-out.c (output_struct_function_base): Save it if
9503         available.
9504         * final.c (decl_to_instance_map): Document LTO strategy.
9506 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
9507             Olivier Hainque  <hainque@adacore.com>
9509         * debug.h (decl_to_instance_map_t): New type.
9510         (decl_to_instance_map): Declare.
9511         (maybe_create_decl_to_instance_map): New inline function.
9512         * final.c (bb_discriminator, last_bb_discriminator): New statics,
9513         to track basic block discriminators.
9514         (final_start_function_1): Initialize them.
9515         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
9516         bb_discriminator.
9517         (decl_to_instance_map): New variable.
9518         (map_decl_to_instance, maybe_set_discriminator): New functions.
9519         (notice_source_line): Set discriminator.
9521 2018-07-31  Ian Lance Taylor  <iant@golang.org>
9523         * targhooks.c (default_have_speculation_safe_value): Add
9524         ATTRIBUTE_UNUSED.
9526 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9528         * dump-context.h: Include "pretty-print.h".
9529         (dump_context::refresh_dumps_are_enabled): New decl.
9530         (dump_context::emit_item): New decl.
9531         (class dump_context): Add fields "m_test_pp" and
9532         "m_test_pp_flags".
9533         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9534         (temp_dump_context::get_dumped_text): New decl.
9535         (class temp_dump_context): Add field "m_pp".
9536         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
9537         (dump_context::refresh_dumps_are_enabled): ...and add a test for
9538         m_test_pp.
9539         (set_dump_file): Update for above change.
9540         (set_alt_dump_file): Likewise.
9541         (dump_loc): New overload, taking a pretty_printer *.
9542         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
9543         to any test pretty-printer.
9544         (make_item_for_dump_gimple_stmt): New function, adapted from
9545         optinfo::add_gimple_stmt.
9546         (dump_context::dump_gimple_stmt): Call it, and use the result,
9547         eliminating the direct usage of dump_file and alt_dump_file in
9548         favor of indirectly using them via emit_item.
9549         (make_item_for_dump_gimple_expr): New function, adapted from
9550         optinfo::add_gimple_expr.
9551         (dump_context::dump_gimple_expr): Call it, and use the result,
9552         eliminating the direct usage of dump_file and alt_dump_file in
9553         favor of indirectly using them via emit_item.
9554         (make_item_for_dump_generic_expr): New function, adapted from
9555         optinfo::add_tree.
9556         (dump_context::dump_generic_expr): Call it, and use the result,
9557         eliminating the direct usage of dump_file and alt_dump_file in
9558         favor of indirectly using them via emit_item.
9559         (make_item_for_dump_printf_va): New function, adapted from
9560         optinfo::add_printf_va.
9561         (make_item_for_dump_printf): New function.
9562         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
9563         and use the result, eliminating the direct usage of dump_file and
9564         alt_dump_file in favor of indirectly using them via emit_item.
9565         (make_item_for_dump_dec): New function.
9566         (dump_context::dump_dec): Call it, and use the result,
9567         eliminating the direct usage of dump_file and alt_dump_file in
9568         favor of indirectly using them via emit_item.
9569         (make_item_for_dump_symtab_node): New function, adapted from
9570         optinfo::add_symtab_node.
9571         (dump_context::dump_symtab_node): Call it, and use the result,
9572         eliminating the direct usage of dump_file and alt_dump_file in
9573         favor of indirectly using them via emit_item.
9574         (dump_context::begin_scope): Reimplement, avoiding direct usage
9575         of dump_file and alt_dump_file in favor of indirectly using them
9576         via emit_item.
9577         (dump_context::emit_item): New member function.
9578         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9579         Set up test pretty-printer on the underlying context.  Call
9580         refresh_dumps_are_enabled.
9581         (temp_dump_context::~temp_dump_context): Call
9582         refresh_dumps_are_enabled.
9583         (temp_dump_context::get_dumped_text): New member function.
9584         (selftest::verify_dumped_text): New function.
9585         (ASSERT_DUMPED_TEXT_EQ): New macro.
9586         (selftest::test_capture_of_dump_calls): Run all tests twice, with
9587         and then without optinfo enabled.  Add uses of
9588         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
9589         * dumpfile.h: Update comment for the dump_* API.
9590         * optinfo-emit-json.cc
9591         (selftest::test_building_json_from_dump_calls): Update for new
9592         param for temp_dump_context ctor.
9593         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
9594         and "m_owned" field.
9595         (optinfo_item::~optinfo_item): Likewise.
9596         (optinfo::add_item): New member function.
9597         (optinfo::emit): Update comment.
9598         (optinfo::add_string): Delete.
9599         (optinfo::add_printf): Delete.
9600         (optinfo::add_printf_va): Delete.
9601         (optinfo::add_gimple_stmt): Delete.
9602         (optinfo::add_gimple_expr): Delete.
9603         (optinfo::add_tree): Delete.
9604         (optinfo::add_symtab_node): Delete.
9605         (optinfo::add_dec): Delete.
9606         * optinfo.h (class dump_context): New forward decl.
9607         (optinfo::add_item): New decl.
9608         (optinfo::add_string): Delete.
9609         (optinfo::add_printf): Delete.
9610         (optinfo::add_printf_va): Delete.
9611         (optinfo::add_gimple_stmt): Delete.
9612         (optinfo::add_gimple_expr): Delete.
9613         (optinfo::add_tree): Delete.
9614         (optinfo::add_symtab_node): Delete.
9615         (optinfo::add_dec): Delete.
9616         (optinfo::add_poly_int): Delete.
9617         (optinfo_item::optinfo_item): Remove "owned" param.
9618         (class optinfo_item): Remove field "m_owned".
9620 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9622         PR middle-end/86705
9623         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
9624         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
9625         requested variable alignment.
9626         (expand_one_ssa_partition): Likewise.
9627         (expand_one_var): Likewise.
9629 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9631         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9632         to speculation_safe_value_not_needed.
9634 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9636         * targhooks.h (speculation_safe_value_not_needed): New prototype.
9637         * targhooks.c (speculation_safe_value_not_needed): New function.
9638         * target.def (have_speculation_safe_value): Update documentation.
9639         * doc/tm.texi: Regenerated.
9641 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9643         * config/aarch64/iterators.md (ALLI_TI): New iterator.
9644         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
9645         expand.
9646         (despeculate_copy<ALLI:mode>_insn): New insn.
9647         (despeculate_copyti_insn): New insn.
9648         (despeculate_simple<ALLI:mode>): New insn
9649         (despeculate_simpleti): New insn.
9650         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
9651         function.
9652         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
9653         aarch64_speculation_safe_value.
9654         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
9656 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9658         * config/aarch64/aarch64-speculation.cc: New file.
9659         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
9660         before pass_reorder_blocks.
9661         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
9662         prototype.
9663         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
9664         X14 and X15 when tracking speculation.
9665         * config/aarch64/aarch64.md (register name constants): Add
9666         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
9667         (unspec): Add UNSPEC_SPECULATION_TRACKER.
9668         (speculation_barrier): New insn attribute.
9669         (cmp<mode>): Allow SP in comparisons.
9670         (speculation_tracker): New insn.
9671         (speculation_barrier): Add speculation_barrier attribute.
9672         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
9673         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
9674         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
9676 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9678         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
9679         aarch64_track_speculation is true.
9680         (tb<optab><mode>1): Likewise.
9681         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
9682         generate CB[N]Z when tracking speculation.
9683         (aarch64_split_compare_and_swap): Likewise.
9684         (aarch64_split_atomic_op): Likewise.
9686 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9688         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
9690 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9692         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
9693         (speculation_barrier): New insn.
9695 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9697         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
9698         * config/arm/arm.md (speculation_barrier): New expand.
9699         (speculation_barrier_insn): New pattern.
9701 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9703         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
9704         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
9705         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
9706         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
9707         list.
9708         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
9709         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
9710         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
9711         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
9712         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
9713         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
9714         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
9715         * builtins.c (expand_speculation_safe_value): New function.
9716         (expand_builtin): Call it.
9717         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
9718         * doc/extend.texi: Document __builtin_speculation_safe_value.
9719         * doc/md.texi: Document "speculation_barrier" pattern.
9720         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
9721         TARGET_HAVE_SPECULATION_SAFE_VALUE.
9722         * doc/tm.texi: Regenerated.
9723         * target.def (have_speculation_safe_value, speculation_safe_value): New
9724         hooks.
9725         * targhooks.c (default_have_speculation_safe_value): New function.
9726         (default_speculation_safe_value): New function.
9727         * targhooks.h (default_have_speculation_safe_value): Add prototype.
9728         (default_speculation_safe_value): Add prototype.
9730 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9732         * dump-context.h (dump_context::dump_loc): New decl.
9733         * dumpfile.c (dump_context::dump_loc): New member function.
9734         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
9735         and dump_gimple_stmt.
9736         (dump_context::dump_gimple_expr_loc): Likewise, using
9737         dump_gimple_expr.
9738         (dump_context::dump_generic_expr_loc): Likewise, using
9739         dump_generic_expr.
9740         (dump_context::dump_printf_loc_va): Likewise, using
9741         dump_printf_va.
9742         (dump_context::begin_scope): Explicitly using the global function
9743         "dump_loc", rather than the member function.
9745 2018-07-31  Martin Sebor  <msebor@redhat.com>
9747         PR tree-optimization/86741
9748         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
9750 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
9752         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
9753         depend on whether prefetch instructions will be emitted or not.
9754         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
9755         will be emitted or not.
9756         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
9757         (TARGET_SETMEM_PFD): New macros.
9759 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9761         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
9762         (NULL_STMT_VEC_INFO): Delete.
9763         (stmt_vec_info::operator*): Likewise.
9764         (stmt_vec_info::operator gimple *): Likewise.
9765         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
9766         of NULL_STMT_VEC_INFO.
9767         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9768         (vect_reassociating_reduction_p): Likewise.
9769         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9770         (vectorizable_store): Likewise.
9771         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
9772         (vec_info::free_stmt_vec_infos): Likewise.
9774 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9776         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
9777         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
9779 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9781         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
9782         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9783         (vec_info::free_stmt_vec_info): New private member functions.
9784         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
9785         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
9786         * tree-parloops.c (gather_scalar_reductions): Remove calls to
9787         set_stmt_vec_info_vec and free_stmt_vec_infos.
9788         * tree-vect-loop.c (_loop_vec_info): Remove call to
9789         set_stmt_vec_info_vec.
9790         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
9791         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
9792         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
9793         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9794         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
9795         assignments in {vec_info::,}new_stmt_vec_info that are redundant
9796         with the clearing in the xcalloc.
9797         (stmt_vec_info_vec): Delete.
9798         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
9799         (vectorize_loops): Likewise.
9800         (vec_info::~vec_info): Remove argument from call to
9801         free_stmt_vec_infos.
9802         (vec_info::add_stmt): Remove vinfo argument from call to
9803         new_stmt_vec_info.
9805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9807         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
9808         rather than a gimple stmt.
9809         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
9810         information for pattern statements when passed the original
9811         statement; instead wait to be passed the pattern statement itself.
9812         Don't call set_vinfo_for_stmt here.
9813         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
9814         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
9815         stmt_vec_infos here.
9816         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
9817         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
9818         stmt_vec_infos entry.
9820 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9822         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
9823         * tree-vectorizer.c (vec_info::replace_stmt): New function.
9824         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
9825         * tree-vect-stmts.c (vectorizable_call): Likewise.
9826         (vectorizable_simd_clone_call): Likewise.
9828 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9830         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
9831         * tree-vectorizer.c (vec_info::remove_stmt): New function.
9832         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
9833         * tree-vect-loop.c (vect_transform_loop): Likewise.
9834         * tree-vect-slp.c (vect_schedule_slp): Likewise.
9835         * tree-vect-stmts.c (vect_remove_stores): Likewise.
9837 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9839         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
9840         (vect_dr_stmt): Delete.
9841         * tree-vectorizer.c (vec_info::lookup_dr): New function.
9842         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
9843         of DR_VECT_AUX.
9844         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9845         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
9846         (vect_verify_datarefs_alignment, vect_peeling_supportable)
9847         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9848         (vect_analyze_data_refs): Likewise.
9849         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
9850         argument.
9851         (vect_find_same_alignment_drs): Likewise.
9852         (vect_slp_analyze_node_dependences): Update calls accordingly.
9853         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
9854         instead of DR_VECT_AUX.
9855         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
9856         of a vector data references.  Use vec_info::lookup_dr instead of
9857         DR_VECT_AUX.
9858         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
9859         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
9860         instead of DR_VECT_AUX.
9862 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9864         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
9865         dr_vec_info.
9866         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9867         accordingly.
9868         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9869         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9870         (vect_gen_prolog_loop_niters): Likewise.
9872 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9874         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
9875         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
9876         (vect_known_alignment_in_bytes, vect_dr_behavior)
9877         (vect_get_scalar_dr_size): Take references as dr_vec_infos
9878         instead of data_references.  Update calls to other routines for
9879         which the same change has been made.
9880         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
9881         dr_vec_infos instead of stmt_vec_infos.
9882         (vect_analyze_data_ref_dependence): Update call accordingly.
9883         (vect_slp_analyze_data_ref_dependence)
9884         (vect_record_base_alignments): Use DR_VECT_AUX.
9885         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
9886         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
9887         (vector_alignment_reachable_p, vect_get_data_access_cost)
9888         (vect_peeling_supportable, vect_analyze_group_access_1)
9889         (vect_analyze_group_access, vect_analyze_data_ref_access)
9890         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
9891         (vect_compile_time_alias, vect_small_gap_p)
9892         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
9893         (vect_supportable_dr_alignment): Take references as dr_vec_infos
9894         instead of data_references.  Update calls to other routines for
9895         which the same change has been made.
9896         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
9897         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9898         (vect_slp_analyze_and_verify_node_alignment)
9899         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9900         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
9901         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
9902         above changes.
9903         (_vect_peel_info::dr): Replace with...
9904         (_vect_peel_info::dr_info): ...this new field.
9905         (vect_peeling_hash_get_most_frequent)
9906         (vect_peeling_hash_choose_best_peeling): Update accordingly.
9907         (vect_peeling_hash_get_lowest_cost):
9908         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
9909         routines for which the same change has been made.
9910         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
9911         data_reference.
9912         * tree-vect-loop-manip.c (get_misalign_in_elems)
9913         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
9914         above changes.
9915         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9916         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
9917         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
9918         (get_group_load_store_type, get_negative_load_store_type)
9919         (vect_get_data_ptr_increment, vectorizable_store)
9920         (vectorizable_load): Likewise.
9921         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
9922         Update calls to other routines for which the same change has been made.
9924 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9926         * tree-vectorizer.h (vec_info::move_dr): New member function.
9927         (dataref_aux): Rename to...
9928         (dr_vec_info): ...this and add "dr" and "stmt" fields.
9929         (_stmt_vec_info::dr_aux): Update accordingly.
9930         (_stmt_vec_info::data_ref_info): Delete.
9931         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
9932         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
9933         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
9934         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
9935         of data_ref.
9936         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
9937         (STMT_VINFO_DR_INFO): New macro.
9938         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
9939         (set_dr_misalignment): Update after rename of dataref_aux.
9940         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
9941         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
9942         initialization of STMT_VINFO_DATA_REF.
9943         * tree-vectorizer.c (vec_info::move_dr): New function.
9944         * tree-vect-patterns.c (vect_recog_bool_pattern)
9945         (vect_recog_mask_conversion_pattern)
9946         (vect_recog_gather_scatter_pattern): Use it.
9947         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
9948         the "dr" and "stmt" fields of dr_vec_info instead of
9949         STMT_VINFO_DATA_REF.
9951 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9953         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
9954         (is_pattern_stmt_p): Use it.
9955         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
9956         on pattern statements.
9958 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9960         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
9961         original stmt as a stmt_vec_info rather than a gimple stmt.
9962         (vect_pattern_recog_1): Take the statement directly as a
9963         stmt_vec_info, rather than via a gimple_stmt_iterator.
9964         Update call to vect_mark_pattern_stmts.
9965         (vect_pattern_recog): Update calls accordingly.
9967 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9969         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
9970         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
9971         a vect_def_type for the first argument.
9972         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
9973         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
9974         operand if it isn't defined by a vectorized statement.
9975         (vect_build_gather_load_calls): Remove the mask_dt argument and
9976         update calls to vect_get_vec_def_for_stmt_copy.
9977         (vectorizable_bswap): Likewise the dt argument.
9978         (vectorizable_call): Update calls to vectorizable_bswap and
9979         vect_get_vec_def_for_stmt_copy.
9980         (vectorizable_simd_clone_call, vectorizable_assignment)
9981         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
9982         (vectorizable_comparison): Update calls to
9983         vect_get_vec_def_for_stmt_copy.
9984         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
9985         vect_is_simple_use.
9986         (vect_get_loop_based_defs): Remove dt argument and update call
9987         to vect_get_vec_def_for_stmt_copy.
9988         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
9989         and vect_get_vec_def_for_stmt_copy.
9990         (vectorizable_load): Update calls to vect_build_gather_load_calls
9991         and vect_get_vec_def_for_stmt_copy.
9992         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9993         (vectorizable_reduction, vectorizable_live_operation): Update calls
9994         to vect_get_vec_def_for_stmt_copy.
9996 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9998         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
9999         and gimple stmt arguments with a stmt_vec_info.
10000         (vect_record_base_alignments): Update calls accordingly.
10001         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10002         and gimple stmt arguments with a stmt_vec_info.
10003         (vect_build_slp_tree_1): Remove vinfo argument and update call
10004         to vect_record_max_nunits.
10005         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10006         and vect_record_max_nunits.
10008 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10010         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10011         file and take a stmt_vec_info instead of a gimple stmt.
10012         (supportable_widening_operation, vect_finish_replace_stmt)
10013         (vect_finish_stmt_generation, vect_get_store_rhs)
10014         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10015         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10016         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10017         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10018         (vect_create_data_ref_ptr, bump_vector_ptr)
10019         (vect_permute_store_chain, vect_setup_realignment)
10020         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10021         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10022         (vectorizable_reduction, vectorizable_induction)
10023         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10024         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10025         than gimple stmts as arguments.
10026         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10027         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10028         (can_group_stmts_p, vect_check_gather_scatter)
10029         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10030         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10031         (vect_permute_load_chain, vect_shift_permute_load_chain)
10032         (vect_transform_grouped_load)
10033         (vect_record_grouped_load_vectors): Likewise.
10034         * tree-vect-loop.c (vect_fixup_reduc_chain)
10035         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10036         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10037         (vectorizable_reduction, vectorizable_induction)
10038         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10039         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10040         (vect_get_load_store_mask): Likewise.
10041         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10042         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10043         * tree-vect-stmts.c (vect_mark_relevant)
10044         (is_simple_and_all_uses_invariant)
10045         (exist_non_indexing_operands_for_use_p, process_use)
10046         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10047         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10048         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10049         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10050         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10051         (get_negative_load_store_type, get_load_store_type)
10052         (vect_check_load_store_mask, vect_check_store_rhs)
10053         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10054         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10055         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10056         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10057         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10058         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10059         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10060         (supportable_widening_operation): Likewise.
10062 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10064         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10065         a stmt_vec_info instead of a gcall.
10066         (vect_check_gather_scatter): Update call accordingly.
10067         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10068         of a gphi.
10069         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10070         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10071         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10072         instead of a gimple stmt.
10073         (vect_transform_loop): Update calls accordingly.
10074         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10075         stmt_vec_infos instead of gimple stmts.
10076         (vect_analyze_slp_instance): Update use accordingly.
10077         * tree-vect-stmts.c (read_vector_array, write_vector_array)
10078         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10079         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10080         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10081         (vect_gen_widened_results_half, vect_get_loop_based_defs)
10082         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10083         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10084         down to subroutines.
10086 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10088         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10089         of the worklist from a vector of gimple stmts to a vector of
10090         stmt_vec_infos.
10091         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10092         (vect_mark_stmts_to_be_vectorized): Likewise
10094 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10096         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10097         statement before passing it to vect_analyze_stmt.
10098         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10099         the chain of phi vector definitions.  Track the exit phi via its
10100         stmt_vec_info.
10101         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10102         STMT_VINFO_REDUC_DEF.
10103         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10104         stmt_vec_infos to handle the statement chains.
10105         (vect_get_slp_defs): Record the first statement in the node
10106         using a stmt_vec_info.
10107         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10108         statements here and pass their stmt_vec_info down to subroutines.
10109         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10110         down to vect_finish_stmt_generation.
10111         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10112         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10113         stmt_vec_infos to subroutines.
10114         (vect_remove_stores): Use stmt_vec_infos to handle the statement
10115         chains.
10117 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10119         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10120         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10121         (vect_permute_store_chain, vect_setup_realignment)
10122         (vect_permute_load_chain, vect_shift_permute_load_chain)
10123         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10124         stmts internally, and when passing values to other vectorizer routines.
10125         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10126         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10127         (vect_analyze_loop_operations, get_initial_def_for_reduction)
10128         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10129         (vectorizable_reduction, vectorizable_induction)
10130         (vectorizable_live_operation, vect_transform_loop_stmt)
10131         (vect_transform_loop): Likewise.
10132         * tree-vect-patterns.c (vect_reassociating_reduction_p)
10133         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10134         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10135         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10136         (vect_slp_analyze_node_operations_1): Likewise.
10137         * tree-vect-stmts.c (vect_mark_relevant, process_use)
10138         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10139         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10140         (vect_finish_stmt_generation_1, get_group_load_store_type)
10141         (get_load_store_type, vect_build_gather_load_calls)
10142         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10143         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10144         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10145         (vectorizable_store, vectorizable_load, vectorizable_condition)
10146         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10147         (supportable_widening_operation): Likewise.
10148         (vect_get_vector_types_for_stmt): Likewise.
10149         * tree-vectorizer.h (vect_dr_behavior): Likewise.
10151 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10153         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10154         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10155         (vect_permute_store_chain, vect_permute_load_chain)
10156         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10157         repeated stmt_vec_info lookups.
10158         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10159         (vect_update_ivs_after_vectorizer): Likewise.
10160         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10161         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10162         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10163         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10164         (vect_bb_slp_scalar_cost): Likewise.
10165         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10167 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10169         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10170         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10171         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10172         to get gassigns and gcalls, rather than operating on generc gimple
10173         stmts.
10174         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10175         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10176         (vectorizable_load, vect_analyze_stmt): Likewise.
10177         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10179 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10181         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10182         return stmt_vec_infos rather than gimple stmts.  Do not accept
10183         null arguments.
10184         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10185         of a gimple stmt.
10186         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10187         Update use of get_later_stmt.
10188         (vect_get_constant_vectors): Update call accordingly.
10189         (vect_schedule_slp_instance): Likewise
10190         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10191         (vect_slp_analyze_instance_dependence): Likewise.
10192         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10194 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10196         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10197         (stmt_info_for_cost::stmt_info): ...this new field.
10198         (add_stmt_costs): Update accordingly.
10199         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10200         (vect_get_known_peeling_cost): Likewise.
10201         (vect_estimate_min_profitable_iters): Likewise.
10202         * tree-vect-stmts.c (record_stmt_cost): Likewise.
10204 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10206         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10207         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10208         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10209         accordingly.
10210         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10212 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10214         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10215         a gimple stmt to a stmt_vec_info.
10216         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10218 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10220         * tree-vectorizer.h (vec_info::grouped_stores): Change from
10221         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10222         (_loop_vec_info::reduction_chains): Likewise.
10223         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10224         accordingly.
10225         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10227 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10229         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10230         a gimple stmt to a stmt_vec_info.
10231         (_stmt_vec_info::next_element): Likewise.
10232         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10233         (vect_slp_analyze_and_verify_node_alignment)
10234         (vect_analyze_group_access_1, vect_analyze_group_access)
10235         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10236         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10237         (vect_supportable_dr_alignment): Update accordingly.
10238         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10239         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10240         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10241         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10242         * tree-vect-slp.c (vect_build_slp_tree_1)
10243         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10244         (vect_split_slp_store_group, vect_analyze_slp_instance)
10245         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10246         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10247         (get_group_load_store_type, get_load_store_type)
10248         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10249         (vect_transform_stmt, vect_remove_stores): Likewise.
10251 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10253         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10254         than a gimple stmt.
10255         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10256         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10257         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10258         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10259         (vector_alignment_reachable_p, vect_get_data_access_cost)
10260         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10261         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10262         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10263         (vect_analyze_group_access_1, vect_analyze_group_access)
10264         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10265         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10266         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10267         result of vect_dr_stmt and use the stmt_vec_info instead of
10268         the associated gimple stmt.
10269         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10270         (vect_gen_prolog_loop_niters): Likewise.
10271         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10273 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10275         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10276         to a vec<stmt_vec_info>.
10277         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10278         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10279         vec<stmt_vec_info>.
10280         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10281         to a vec<stmt_vec_info>.
10282         (bst_traits::value_type, bst_traits::value_type): Likewise.
10283         (bst_traits::hash): Update accordingly.
10284         (vect_get_and_check_slp_defs): Change the stmts parameter from
10285         a vec<gimple *> to a vec<stmt_vec_info>.
10286         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10287         (vect_build_slp_tree): Likewise.
10288         (vect_build_slp_tree_2): Likewise.  Update uses of
10289         SLP_TREE_SCALAR_STMTS.
10290         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10291         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10292         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10293         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10294         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10295         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10296         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10297         (vect_get_constant_vectors, vect_get_slp_defs)
10298         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10299         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10300         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10301         instead of gimple stmts.
10302         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10303         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10304         (vect_slp_analyze_instance_dependence): Update uses of
10305         SLP_TREE_SCALAR_STMTS.
10306         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10307         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10308         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10309         (get_initial_defs_for_reduction): Likewise.
10310         (vect_create_epilog_for_reduction): Likewise.
10311         (vectorize_fold_left_reduction): Likewise.
10312         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10313         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
10314         (can_vectorize_live_stmts): Likewise.
10316 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10318         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
10319         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10320         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
10321         than gimple stmts.
10322         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
10323         of a gimple stmt.
10324         (gather_scalar_reductions): Update after above interface changes.
10325         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
10326         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
10327         than gimple stmts.
10328         (vect_force_simple_reduction): Likewise.
10329         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
10330         LOOP_VINFO_REDUCTIONS.
10331         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10333 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10335         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
10336         a gimple stmt to a stmt_vec_info.
10337         * tree-vect-loop.c (vect_active_double_reduction_p)
10338         (vect_force_simple_reduction, vectorizable_reduction): Update
10339         accordingly.
10341 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10343         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
10344         vec<gimple *> to a vec<stmt_vec_info>.
10345         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
10346         the reduction_phis argument from a vec<gimple *> to a
10347         vec<stmt_vec_info>.
10348         (vectorizable_reduction): Likewise the phis local variable that
10349         is passed to vect_create_epilog_for_reduction.  Update for new type
10350         of SLP_TREE_VEC_STMTS.
10351         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
10352         (vectorizable_live_operation): Likewise.
10353         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
10354         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
10356 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10358         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
10359         a gimple stmt to a stmt_vec_info.
10360         (vectorizable_condition, vectorizable_live_operation)
10361         (vectorizable_reduction, vectorizable_induction): Pass back the
10362         vectorized statement as a stmt_vec_info.
10363         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
10364         use of STMT_VINFO_VEC_STMT.
10365         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
10366         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
10367         as stmt_vec_infos rather than gimple stmts.
10368         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
10369         to a stmt_vec_info.
10370         (vectorizable_live_operation): Likewise.
10371         (vectorizable_reduction, vectorizable_induction): Likewise,
10372         updating use of STMT_VINFO_VEC_STMT.
10373         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
10374         of STMT_VINFO_VEC_STMT.
10375         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
10376         (vectorizable_simd_clone_call, vectorizable_conversion)
10377         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10378         (vectorizable_store, vectorizable_load, vectorizable_condition)
10379         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
10380         from a gimple stmt to a stmt_vec_info.
10381         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
10382         pointer to a stmt_vec_info to the vectorizable_* routines.
10384 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10386         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
10387         a gimple stmt to a stmt_vec_info.
10388         (is_pattern_stmt_p): Update accordingly.
10389         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
10390         (vect_record_grouped_load_vectors): Likewise.
10391         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10392         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
10393         (vect_model_reduction_cost): Likewise.
10394         (vect_create_epilog_for_reduction): Likewise.
10395         (vectorizable_reduction, vectorizable_induction): Likewise.
10396         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10397         Return the stmt_vec_info for the pattern statement.
10398         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
10399         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
10400         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10401         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
10402         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10403         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
10404         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
10405         (free_stmt_vec_info, vect_is_simple_use): Likewise.
10407 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10409         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
10410         (vect_finish_stmt_generation): Likewise.
10411         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10412         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
10413         (vect_build_gather_load_calls): Use the return value of the above
10414         functions instead of a separate call to vinfo_for_stmt.  Use narrow
10415         scopes for the input gimple stmt and wider scopes for the associated
10416         stmt_vec_info.  Use vec_info::lookup_def when setting these
10417         stmt_vec_infos from an SSA_NAME definition.
10418         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10419         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10420         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10421         (vectorizable_store, vectorizable_load, vectorizable_condition)
10422         (vectorizable_comparison): Likewise.
10423         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
10424         (vectorizable_reduction): Likewise.
10426 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10428         * tree-vectorizer.h (vect_is_simple_use): Add an optional
10429         stmt_vec_info * parameter before the optional gimple **.
10430         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10431         (process_use, vect_get_vec_def_for_operand_1): Update callers.
10432         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
10433         * tree-vect-loop.c (vectorizable_reduction): Likewise.
10434         (vectorizable_live_operation): Likewise.
10435         * tree-vect-patterns.c (type_conversion_p): Likewise.
10436         (vect_look_through_possible_promotion): Likewise.
10437         (vect_recog_rotate_pattern): Likewise.
10438         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10440 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10442         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
10443         a typedef to a wrapper class.
10444         (NULL_STMT_VEC_INFO): New macro.
10445         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
10446         (stmt_vec_info::operator*): New function.
10447         (stmt_vec_info::operator gimple *): Likewise.
10448         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
10449         (add_stmt_costs): Likewise.
10450         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
10451         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10452         (vect_get_known_peeling_cost): Likewise.
10453         (vect_estimate_min_profitable_iters): Likewise.
10454         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10455         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
10456         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10457         (vectorizable_store, free_stmt_vec_infos): Likewise.
10458         (new_stmt_vec_info): Change return type of xcalloc to
10459         _stmt_vec_info *.
10461 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10463         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
10464         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
10465         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
10466         a single_imm_use-based sequence.
10467         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
10469 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10471         * tree-vectorizer.h (vec_info::lookup_def): Declare.
10472         * tree-vectorizer.c (vec_info::lookup_def): New function.
10473         * tree-vect-patterns.c (vect_get_internal_def): Use it.
10474         (vect_widened_op_tree): Likewise.
10475         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10476         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
10477         (vectorizable_reduction): Likewise.
10478         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
10479         of a gimple *.
10480         (vect_is_slp_reduction): Update calls accordingly.  Use
10481         vec_info::lookup_def.
10482         (vect_is_simple_reduction): Likewise
10483         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
10485 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10487         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
10488         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
10489         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
10490         of vinfo_for_stmt.
10491         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
10492         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
10493         (vect_update_vf_for_slp, vect_analyze_loop_operations)
10494         (vect_is_slp_reduction, vectorizable_induction)
10495         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
10496         * tree-vect-patterns.c (vect_init_pattern_stmt):
10497         (vect_determine_min_output_precision_1, vect_determine_precisions)
10498         (vect_pattern_recog): Likewise.
10499         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
10500         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
10501         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
10502         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10503         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
10504         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
10505         info field from a loop to a loop_vec_info.
10507 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10509         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
10510         (vec_info::add_stmt): Declare.
10511         * tree-vectorizer.c (vec_info::add_stmt): New function.
10512         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
10513         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
10514         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10515         (vectorizable_induction): Likewise.
10516         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
10517         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10518         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
10519         (vectorizable_load): Likewise.
10520         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10521         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
10522         (vect_recog_gather_scatter_pattern): Likewise.
10523         (append_pattern_def_seq): Likewise.  Remove a check that is
10524         performed by add_stmt itself.
10526 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10528         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
10529         which make_ssa_name was called with new_stmt before new_stmt
10530         had been created.
10532 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10534         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
10535         split out from...
10536         (vect_is_slp_reduction): ...here...
10537         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
10538         that are already known to be false.
10540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10542         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
10543         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
10544         STMT_VINFO_NUM_SLP_USES when it's true.
10545         (vect_free_slp_instance): Add a final_p parameter and pass it to
10546         vect_free_slp_tree.
10547         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
10548         (vect_analyze_slp_instance): Likewise.
10549         (vect_slp_analyze_operations): Likewise.
10550         (vect_slp_analyze_bb_1): Likewise.
10551         * tree-vectorizer.c (vec_info): Likewise.
10552         * tree-vect-loop.c (vect_transform_loop): Likewise.
10554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10556         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10557         function is not called for second and subsequent members of
10558         a reduction group.
10560 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10562         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
10563         cases for nested loops from here to ...
10564         (vect_create_epilog_for_reduction): ...here.  Only call
10565         vect_is_simple_use for inner-loop reductions.
10567 2018-07-31  Martin Liska  <mliska@suse.cz>
10569         PR gcov-profile/85338
10570         PR gcov-profile/85350
10571         PR gcov-profile/85372
10572         * profile.c (struct location_triplet): New.
10573         (struct location_triplet_hash): Likewise.
10574         (output_location): Do not output a BB that
10575         is already recorded for a line.
10576         (branch_prob): Use streamed_locations.
10578 2018-07-31  Martin Liska  <mliska@suse.cz>
10580         PR gcov-profile/85370
10581         * coverage.c (coverage_begin_function): Do not mark target
10582         clones as artificial functions.
10584 2018-07-31  Martin Liska  <mliska@suse.cz>
10586         PR gcov-profile/83813
10587         PR gcov-profile/84758
10588         PR gcov-profile/85217
10589         PR gcov-profile/85332
10590         * profile.c (branch_prob): Do not record GOTO expressions
10591         for GIMPLE statements which locations are already streamed.
10593 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10595         * gcc.c (handle_spec_function): Accept a soft_matched_part
10596         argument, as do_spec_1.  Pass it down to ...
10597         (eval_spec_function): Accept a soft_matched_part argument,
10598         and pass it down to ...
10599         (do_spec_2): Accept a soft_matched_part argument, and pass
10600         it down to do_spec_1.
10601         (do_spec_1): Pass soft_matched_part to handle_spec_function.
10602         (handle_braces): Update call to handle_spec_function.
10603         (driver::set_up_specs): Update calls to do_spec_2.
10604         (compare_debug_dump_opt_spec_function): Likewise.
10605         (compare_debug_self_opt_spec_function): Likewise.
10607 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10609         * common.opt (nolibc): New option.
10610         * doc/invoke.texi (Link Options): Document it.
10611         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
10612         * config/alpha/linux.h: Likewise.
10613         * config/arc/elf.h: Likewise.
10614         * config/arm/uclinux-elf.h: Likewise.
10615         * config/arm/unknown-elf.h: Likewise.
10616         * config/avr/avrlibc.h: Likewise.
10617         * config/bfin/bfin.h: Likewise.
10618         * config/bfin/linux.h: Likewise.
10619         * config/bfin/uclinux.h: Likewise.
10620         * config/darwin.h: Likewise.
10621         * config/darwin10.h: Likewise.
10622         * config/darwin12.h: Likewise.
10623         * config/gnu-user.h: Likewise.
10624         * config/lm32/uclinux-elf.h: Likewise.
10625         * config/pa/pa-hpux11.h: Likewise.
10626         * config/pa/pa64-hpux.h: Likewise.
10627         * config/sparc/sparc.h: Likewise.
10629 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10631         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
10632         undefined variables.
10634 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10636         PR target/86640
10637         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
10638         instead of GEN_INT.
10640 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10642         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10643         terminated string literal.
10645 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10647         PR rtl-optimization/85160
10648         * combine.c (is_just_move): New function.
10649         (try_combine): Allow combining two instructions into two if neither of
10650         the original instructions was a move.
10652 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
10654         PR target/86673
10655         * doc/extend.texi (Global Register Variables): Discourage use of type
10656         qualifiers.
10657         (Local Register Variables): Likewise.
10659 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
10661         PR tree-optimization/86506
10662         * hwint.h (ceil_log2): Resync with hwint.c implementation.
10664 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
10666         PR target/86547
10667         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
10668         hard_regno, make sure no insn between `from` and `to` clobbers it.
10670 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
10671             Tom de Vries  <tdevries@suse.de>
10673         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
10674         (PTX_DEFAULT_RUNTIME_DIM): ... this.
10675         (nvptx_goacc_validate_dims): Set default worker and gang dims to
10676         PTX_DEFAULT_RUNTIME_DIM.
10677         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
10679 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
10681         * config/pa/pa.c (pa_output_addr_vec): Align address table.
10682         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
10683         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
10685 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
10687         * config/rs6000/constraints.md (wG constraint): Delete, no longer
10688         used.
10689         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
10690         predicate to reflect toc fusion has been deleted.
10691         (toc_fusion_mem_raw): Delete, no longer used.
10692         (toc_fusion_mem_wrapped): Likewise.
10693         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
10694         fusion mask bit.
10695         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
10696         Delete, no longer used.
10697         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
10698         meant to be used for toc fusion.
10699         (rs6000_debug_print_mode): Delete toc fusion debugging.
10700         (rs6000_debug_reg_global): Likewise.
10701         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
10702         fusion and secondary reload support that were never used.
10703         (rs6000_option_override_internal): Delete TOC fusion, that was only
10704         partially defined, and it did not work unless you also used the
10705         -mcmodel= switch.
10706         (rs6000_legitimate_address_p): Delete TOC fusion support.
10707         (rs6000_opt_masks): Likewise.
10708         (fusion_wrap_memory_address): Delete function, no longer used.
10709         (fusion_split_address); Delete TOC fusion support.
10710         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
10711         longer used with toc fusion being deleted.
10712         (TARGET_TOC_FUSION_FP): Likewise.
10713         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
10714         UNSPEC.
10715         (toc fusion spliter): Delete TOC fusion support.
10716         (toc_fusionload_<mode>): Likewise.
10717         (toc_fusionload_di): Likewise.
10718         (fusion_gpr_load_<mode>): Delete generator function, this insn no
10719         longer needs to be named.  Rename predicate to delete TOC fusion.
10720         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10721         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10722         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10723         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10724         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
10726 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10728         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
10729         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
10730         __int128 in built-in function prototypes.
10731         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
10732         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
10734 2018-07-27  Martin Sebor  <msebor@redhat.com>
10736         PR tree-optimization/86696
10737         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
10738         types, including enums.
10739         (handle_char_store): Be prepared for the above function to fail.
10741 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
10743         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
10744         when optimization level is lower than 2 or optimize for size.
10746 2018-07-26  Martin Sebor  <msebor@redhat.com>
10748         PR tree-optimization/86043
10749         PR tree-optimization/86042
10750         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
10751         (get_string_cst_length): Rename...
10752         (get_min_string_length): ...to this.  Add argument.
10753         (handle_char_store): Extend to handle multi-character stores by
10754         MEM_REF.
10755         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
10756         * tree.h (initializer_zerop): Add argument.
10758 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
10760         PR middle-end/86660
10761         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
10762         declare target to variables if they have always,{to,from,tofrom} map
10763         kinds.
10765 2018-07-26  Martin Liska  <mliska@suse.cz>
10767         PR lto/86548
10768         * lto-wrapper.c: Add linker_output as prefix
10769         for ltrans_output_file.
10771 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
10773         PR rtl-optimization/85805
10774         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
10775         value for hard registers if that was written in the same mode.
10777 2018-07-26  Martin Liska  <mliska@suse.cz>
10779         PR gcov-profile/86536
10780         * gcov.c (format_gcov): Use printf format %.*f directly
10781         and do not handle special values.
10783 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10785         * common/config/arc/arc-common.c (arc_option_optimization_table):
10786         Update default optimizations for size.
10788 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10790         * config/arc/arc.md (movsf_insn): Add short instruction selection.
10791         * config/arc/constraints.md (CfZ): New constraint.
10792         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
10793         (subsf3_fpu): Likewise.
10794         (cmpsf_fpu): Likewise.
10795         (cmpsf_fpu_uneq): Likewise.
10797 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10799         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
10800         accesses as well.
10801         (arc_is_uncached_mem_p): uncached applies to both the variable and
10802         the pointer.
10804 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10806         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
10807         register names.
10809 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
10811         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
10812         field "m_scopes" from vec to auto_vec.
10814 2018-07-25  Martin Liska  <mliska@suse.cz>
10816         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
10817         return type.
10819 2018-07-25  Richard Biener  <rguenther@suse.de>
10821         PR debug/86654
10822         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
10823         special wrt context_die late.
10824         (gen_subprogram_die): Re-use DIEs in local scope.
10826 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
10828         PR tree-optimization/86644
10829         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
10831 2018-07-25  Martin Liska  <mliska@suse.cz>
10833         PR middle-end/86645
10834         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
10835         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
10837 2018-07-25  Martin Liska  <mliska@suse.cz>
10839         PR sanitizer/79635
10840         * params.def: Explain ASan abbreviation and provide
10841         a documentation link.
10843 2018-07-24  Martin Sebor  <msebor@redhat.com>
10845         PR tree-optimization/86622
10846         PR tree-optimization/86532
10847         * builtins.h (string_length): Declare.
10848         * builtins.c (c_strlen): Correct handling of non-constant offsets.
10849         (check_access): Be prepared for non-constant length ranges.
10850         (string_length): Make extern.
10851         * expr.c (string_constant): Only handle the minor non-constant
10852         array index.  Use string_constant to compute the length of
10853         a generic string constant.
10855 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
10857         PR tree-optimization/86618
10858         * tree-vect-stmts.c (vectorizable_call): Don't take the address
10859         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
10861 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
10863         PR tree-optimization/86636
10864         * json.cc (json::object::set): Fix comment.  Add assertions.
10865         (json::array::append): Move here from json.h.  Add comment and an
10866         assertion.
10867         (json::string::string): Likewise.
10868         * json.h (json::array::append): Move to json.cc.
10869         (json::string::string): Likewise.
10870         * optinfo-emit-json.cc
10871         (optrecord_json_writer::impl_location_to_json): Assert that we
10872         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
10873         wrapper around it.  Expand the location once, rather than three
10874         times.
10875         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
10876         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
10877         wrappers.
10878         (optrecord_json_writer::optinfo_to_json): Likewise, in four
10879         places.  Fix some overlong lines.
10881 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
10883         * config/aarch64/aarch64-simd.md
10884         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
10885         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
10886         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
10887         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
10888         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
10889         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
10890         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
10891         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
10892         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
10894 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
10896         PR middle-end/86627
10897         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
10898         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
10899         and abs_d == d, do the power of two handling if profitable.
10901 2018-07-24  Richard Biener  <rguenther@suse.de>
10903         * match.pd: Add BIT_FIELD_REF canonicalizations.
10905 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10907         PR c/86617
10908         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
10910 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10912         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
10913         terminated STRING_CST object.
10915 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10917         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
10919 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
10921         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
10922         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
10923         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
10924         the elements into a register.
10925         (rs6000_split_v4si_init_di_reg): Delete.
10926         (rs6000_split_v4si_init): Delete.
10927         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
10928         (vsx_init_v4si): Rewrite as a define_expand.
10930 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
10932         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
10933         zero_extend argument from memory): New.
10935 2018-07-22  Martin Sebor  <msebor@redhat.com>
10937         PR bootstrap/86621
10938         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
10939         diagnosing calls with unknown arguments unless -Walloca-larger-than
10940         is restricted to less than PTRDIFF_MAX bytes.
10942 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
10944         * doc/gcov.texi (Invoking Gcov): Editorial changes.
10946 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
10948         * pretty-print.c (text_info::set_location): Remove redundant
10949         "line_table" parameter from call to rich_location::set_range.
10951 2018-07-20  Martin Sebor  <msebor@redhat.com>
10953         PR middle-end/82063
10954         * builtins.c (expand_builtin_alloca): Adjust.
10955         * calls.c (alloc_max_size): Simplify.
10956         * cgraphunit.c (cgraph_node::expand): Adjust.
10957         * common.opt (larger_than_size, warn_frame_larger_than): Remove
10958         variables.
10959         (frame_larger_than_size): Same.
10960         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
10961         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
10962         Initialize.
10963         * doc/invoke.texi (GCC Command Options): Document option arguments.
10964         Explain byte-size arguments and suffixes.
10965         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
10966         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
10967         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
10968         * doc/options.texi (UInteger): Expand.
10969         (Host_Wide_Int, ByteSize): Document new properties.
10970         * final.c (final_start_function_1): Include sizes in an error message.
10971         * function.c (frame_offset_overflow): Same.
10972         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
10973         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
10974         Diagnose unbounded alloca calls only for limits of less than
10975         PTRDIFF_MAX.
10976         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
10977         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
10978         for alloca(0).
10979         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
10980         only for limits of less than PTRDIFF_MAX.
10981         * langhooks-def.h (lhd_handle_option): Change function argument
10982         to HOST_WIDE_INT.
10983         * langhooks.c (lhd_handle_option): Same.
10984         * langhooks.h (handle_option): Same.
10985         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
10986         ByteSize flags.
10987         (var_type, var_type_struct): Same.
10988         (var_set): Handle ByteSize flag.
10989         * optc-gen.awk: Add comments to output to ease debugging.  Make
10990         use of HOST_WIDE_INT where appropriate.
10991         * opts-gen-save.awk:  Use %lx to format unsigned long.
10992         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
10993         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
10994         arguments.  Parse bytes-size suffixes.
10995         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
10996         (enum_value_to_arg): Same.
10997         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
10998         (handle_option): Adjust.
10999         (generate_option): Change function argument to HOST_WIDE_INT.
11000         (cmdline_handle_error): Adjust.
11001         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11002         (set_option): Change function argument to HOST_WIDE_INT.
11003         (option_enabled): Handle cl_host_wide_int.
11004         (get_option_state): Handle CLVC_SIZE.
11005         (control_warning_option): Same.
11006         * opts.c (common_handle_option): Change function argument to
11007         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11008         OPT_Wvla_larger_than_.
11009         * opts.h (enum cl_var_type): Add an enumerator.
11010         * stor-layout.c (layout_decl): Print a more meaningful warning.
11011         * toplev.c (output_stack_usage): Adjust.
11013 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11015         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11016         call to inline_expand_builtin_string_cmp.
11017         (expand_builtin_strcmp): Likewise.
11018         (expand_builtin_strncmp): Likewise.
11019         (inline_string_cmp): Delete the last parameter, change char_type_node
11020         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11021         two operands.
11022         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11023         the inlining expansion on target where the type of the call has same or
11024         narrower precision than unsigned char.
11026 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11028         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11029         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11030         * common.opt (fsave-optimization-record): New option.
11031         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11032         * doc/invoke.texi (-fsave-optimization-record): New option.
11033         * dumpfile.c: Include "optinfo-emit-json.h".
11034         (struct kv_pair): Move to coretypes.h.
11035         (optgroup_options): Make non-static.
11036         (dump_context::end_scope): Call
11037         optimization_records_maybe_pop_dump_scope.
11038         * dumpfile.h (optgroup_options): New decl.
11039         * json.cc: New file.
11040         * json.h: New file.
11041         * optinfo-emit-json.cc: New file.
11042         * optinfo-emit-json.h: New file.
11043         * optinfo.cc: Include "optinfo-emit-json.h".
11044         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11045         (optinfo_enabled_p): Check optimization_records_enabled_p.
11046         (optinfo_wants_inlining_info_p): Likewise.
11047         * optinfo.h: Update comment.
11048         * profile-count.c (profile_quality_as_string): New function.
11049         * profile-count.h (profile_quality_as_string): New decl.
11050         (profile_count::quality): New accessor.
11051         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11052         and optinfo_emit_json_cc_tests.
11053         * selftest.h (selftest::json_cc_tests): New decl.
11054         (selftest::optinfo_emit_json_cc_tests): New decl.
11055         * toplev.c: Include "optinfo-emit-json.h".
11056         (compile_file): Call optimization_records_finish.
11057         (do_compile): Call optimization_records_start.
11058         * tree-ssa-live.c: Include optinfo.h.
11059         (remove_unused_scope_block_p): Retain inlining information if
11060         optinfo_wants_inlining_info_p returns true.
11062 2018-07-20  Richard Biener  <rguenther@suse.de>
11064         PR debug/86585
11065         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11066         to cover -flto-partition=none.
11068 2018-07-20  Martin Liska  <mliska@suse.cz>
11070         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11071         (get_decl_source_range): Remove unused function.
11073 2018-07-20  Richard Biener  <rguenther@suse.de>
11075         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11076         (struct vn_phi_s): Likewise.
11077         (struct vn_reference_s): Likewise.
11078         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11079         for searching the slot of an entry known to be in the hash itself.
11080         (vn_phi_hasher::equal): Likewise.
11081         (vn_reference_hasher::equal): Likewise.
11082         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11083         globals.
11084         (optimistic_info, current_info): Remove, keeping only valid_info.
11085         (vn_reference_lookup_1): Remove fallback lookup.
11086         (vn_reference_lookup_2): Likewise.
11087         (vn_nary_op_lookup_1): Likewise.
11088         (vn_phi_lookup): Likewise.
11089         (vn_nary_build_or_lookup_1): Make sure to not chain the built
11090         hash element.
11091         (vn_reference_insert): Adjust, chain the inserted hash element
11092         at last_inserted_ref.
11093         (vn_reference_insert_pieces): Likewise.
11094         (visit_reference_op_call): Likewise.
11095         (vn_nary_op_insert_into): Chain the inserted hash element at
11096         last_inserted_nary.
11097         (vn_nary_op_insert_pieces): Adjust.
11098         (vn_nary_op_insert): Likewise.
11099         (vn_nary_op_insert_stmt): Likewise.
11100         (vn_phi_insert): Adjust, chain the inserted hash element at
11101         last_inserted_phi.
11102         (process_scc): Remove clearing and copying the optimistic
11103         table.  Instead remove elements inserted during an optimistic
11104         iteration from the single table we maintain.
11105         (init_scc_vn): Adjust.
11106         (free_scc_vn): Likewise.
11107         (sccvn_dom_walker::record_cond): Likewise.
11108         (sccvn_dom_walker::after_dom_children): Likewise.
11110 2018-07-19  Martin Sebor  <msebor@redhat.com>
11112         PR tree-optimization/84047
11113         PR tree-optimization/83776
11114         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11115         (check_array_bounds): Call it.
11117 2018-07-19  Martin Sebor  <msebor@redhat.com>
11119         * align.h (align_flags): Use member initialization.
11121 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
11123         * Makefile.in (OBJS): Add optinfo.o.
11124         * coretypes.h (class symtab_node): New forward decl.
11125         (struct cgraph_node): New forward decl.
11126         (class varpool_node): New forward decl.
11127         * dump-context.h: New file.
11128         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11129         "tree-pass.h".
11130         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11131         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11132         (set_alt_dump_file): Likewise.
11133         (dump_context::~dump_context): New dtor.
11134         (dump_gimple_stmt): Move implementation to...
11135         (dump_context::dump_gimple_stmt): ...this new member function.
11136         Add the stmt to any pending optinfo, creating one if need be.
11137         (dump_gimple_stmt_loc): Move implementation to...
11138         (dump_context::dump_gimple_stmt_loc): ...this new member function.
11139         Start a new optinfo and add the stmt to it.
11140         (dump_gimple_expr): Move implementation to...
11141         (dump_context::dump_gimple_expr): ...this new member function.
11142         Add the stmt to any pending optinfo, creating one if need be.
11143         (dump_gimple_expr_loc): Move implementation to...
11144         (dump_context::dump_gimple_expr_loc): ...this new member function.
11145         Start a new optinfo and add the stmt to it.
11146         (dump_generic_expr): Move implementation to...
11147         (dump_context::dump_generic_expr): ...this new member function.
11148         Add the tree to any pending optinfo, creating one if need be.
11149         (dump_generic_expr_loc): Move implementation to...
11150         (dump_context::dump_generic_expr_loc): ...this new member
11151         function.  Add the tree to any pending optinfo, creating one if
11152         need be.
11153         (dump_printf): Move implementation to...
11154         (dump_context::dump_printf_va): ...this new member function.  Add
11155         the text to any pending optinfo, creating one if need be.
11156         (dump_printf_loc): Move implementation to...
11157         (dump_context::dump_printf_loc_va): ...this new member function.
11158         Start a new optinfo and add the stmt to it.
11159         (dump_dec): Move implementation to...
11160         (dump_context::dump_dec): ...this new member function.  Add the
11161         value to any pending optinfo, creating one if need be.
11162         (dump_context::dump_symtab_node): New member function.
11163         (dump_context::get_scope_depth): New member function.
11164         (dump_context::begin_scope): New member function.
11165         (dump_context::end_scope): New member function.
11166         (dump_context::ensure_pending_optinfo): New member function.
11167         (dump_context::begin_next_optinfo): New member function.
11168         (dump_context::end_any_optinfo): New member function.
11169         (dump_context::s_current): New global.
11170         (dump_context::s_default): New global.
11171         (dump_scope_depth): Delete global.
11172         (dumpfile_ensure_any_optinfo_are_flushed): New function.
11173         (dump_symtab_node): New function.
11174         (get_dump_scope_depth): Reimplement in terms of dump_context.
11175         (dump_begin_scope): Likewise.
11176         (dump_end_scope): Likewise.
11177         (selftest::temp_dump_context::temp_dump_context): New ctor.
11178         (selftest::temp_dump_context::~temp_dump_context): New dtor.
11179         (selftest::verify_item): New function.
11180         (ASSERT_IS_TEXT): New macro.
11181         (ASSERT_IS_TREE): New macro.
11182         (ASSERT_IS_GIMPLE): New macro.
11183         (selftest::test_capture_of_dump_calls): New test.
11184         (selftest::dumpfile_c_tests): Call it.
11185         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11186         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11187         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11188         descriptive comment.
11189         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11190         (dump_node, dump_bb): Move these unrelated decls.
11191         (class dump_manager): Add leading comment.
11192         * optinfo.cc: New file.
11193         * optinfo.h: New file.
11195 2018-07-19  Michael Collison  <michael.collison@arm.com>
11196             Richard Henderson <rth@redhat.com>
11198         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11199         (subti): Handle op1 zero.
11200         (subvti4, usub4ti4): New.
11201         (*sub<GPI>3_compare1_imm): New.
11202         (sub<GPI>3_carryinCV): New.
11203         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11204         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11206 2018-07-19  Michael Collison  <michael.collison@arm.com>
11207             Richard Henderson <rth@redhat.com>
11209         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11210         (addti3): Create simpler code if low part is already known to be 0.
11211         (addvti4, uaddvti4): New.
11212         (*add<GPI>3_compareC_cconly_imm): New.
11213         (*add<GPI>3_compareC_cconly): New.
11214         (*add<GPI>3_compareC_imm): New.
11215         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11216         handle constants within this pattern..
11217         (*add<GPI>3_compareV_cconly_imm): New.
11218         (*add<GPI>3_compareV_cconly): New.
11219         (*add<GPI>3_compareV_imm): New.
11220         (add<GPI>3_compareV): New.
11221         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11222         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11223         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11224         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11225         with 'comparison' operator.
11226         (*add<GPI>3_compareV_cconly_imm): Ditto.
11227         (*add<GPI>3_compareV_cconly): Ditto.
11228         (*add<GPI>3_compareV_imm): Ditto.
11229         (add<GPI>3_compareV): Ditto.
11230         (add<mode>3_carryinC): Ditto.
11231         (*add<mode>3_carryinC_zero): Ditto.
11232         (*add<mode>3_carryinC): Ditto.
11233         (add<mode>3_carryinV): Ditto.
11234         (*add<mode>3_carryinV_zero): Ditto.
11235         (*add<mode>3_carryinV): Ditto.
11237 2018-07-19  Michael Collison  <michael.collison@arm.com>
11238             Richard Henderson <rth@redhat.com>
11240         * config/aarch64/aarch64-modes.def (CC_V): New.
11241         * config/aarch64/aarch64-protos.h
11242         (aarch64_addti_scratch_regs): Declare
11243         (aarch64_subvti_scratch_regs): Declare.
11244         (aarch64_expand_subvti): Declare.
11245         (aarch64_gen_unlikely_cbranch): Declare
11246         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11247         for signed overflow using CC_Vmode.
11248         (aarch64_get_condition_code_1): Handle CC_Vmode.
11249         (aarch64_gen_unlikely_cbranch): New function.
11250         (aarch64_addti_scratch_regs): New function.
11251         (aarch64_subvti_scratch_regs): New function.
11252         (aarch64_expand_subvti): New function.
11254 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11256         * config/aarch64/aarch64-option-extensions.def: New entry for profile
11257         extension.
11258         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11259         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11260         extension.
11262 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11264         PR target/83009
11265         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11266         address check not strict.
11268 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11270         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11271         Umq with Umn.
11272         (store_pair_lanes<mode>): Likewise.
11273         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11274         enum value 'ADDR_QUERY_LDP_STP_N'.
11275         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11276         (aarch64_print_address_internal): Add declaration.
11277         (aarch64_print_ldpstp_address): Remove.
11278         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11279         (aarch64_print_operand): Change printing of 'y'.
11280         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11281         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11282         'true' rather than '1'.
11283         * config/aarch64/constraints.md (Uml): Likewise.
11284         (Uml): Rename to Umn.
11285         (Umq): Remove.
11287 2018-07-19  Richard Biener  <rguenther@suse.de>
11289         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11290         a trailing array.
11291         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11292         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11293         (vn_reference_hasher): Likewise.
11294         (struct vn_tables_s): Remove obstack and alloc-pool members.
11295         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11296         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11297         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11298         (vn_reference_insert_pieces): Likewise.
11299         (alloc_vn_nary_op_noinit): Adjust.
11300         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11301         (vn_phi_eq): Adjust.
11302         (shared_lookup_phiargs): Remove.
11303         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11304         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11305         (visit_reference_op_call): Likewise.
11306         (copy_nary, copy_phi, copy_reference): Remove.
11307         (process_scc): Rewind the obstack when iterating.  Do not
11308         copy the elements to valid_info but just move them from one
11309         hashtable to the other.
11310         (allocate_vn_table): Adjust.
11311         (free_vn_table): Likewise.
11312         (init_scc_vn): Likewise.
11313         (free_scc_vn): Likewise.
11315 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
11317         PR target/86560
11318         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
11319         indirect_return as function type attribute.
11320         (ix86_attribute_table): Change indirect_return to function
11321         type attribute.
11322         * doc/extend.texi: Update indirect_return attribute.
11324 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
11326         * wide-int.h (widest2_int): New.
11327         * gimple-fold.c (arith_overflowed_p): Use it.
11328         * tree.h (widest2_int_cst): New.
11329         * tree-vrp.c (wide_int_binop_overflow): Rename from
11330         vrp_int_const_binop.
11331         Rewrite to work on trees.
11332         (extract_range_from_multiplicative_op_1): Abstract code to...
11333         (wide_int_range_min_max): ...here.
11334         (wide_int_range_cross_product): ...and here.
11335         (extract_range_from_binary_expr_1): Abstract overflow code to...
11336         (wide_int_range_mult_wrapping): ...here.
11337         * tree-vrp.h (wide_int_range_cross_product): New.
11338         (wide_int_range_mult_wrapping): New.
11340 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
11341             Julia Koval  <julia.koval@intel.com>
11343         * config/i386/x86-tune-costs.h (skylake_memcpy,
11344         skylake_memset): Replace rep_prefix with unrolling for size 512.
11346 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
11348         PR middle-end/86544
11349         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
11350         comparision with EQ_EXPR in last stmt.
11352 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11354         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
11355         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
11356         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
11357         previously known as "PowerPC AltiVec Built-in Functions".  Move
11358         some material to new subsubsections "PowerPC AltiVec Built-in
11359         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
11360         ISA 2.07".
11361         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
11362         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
11363         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
11364         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
11366 2018-07-18  Richard Biener  <rguenther@suse.de>
11368         PR tree-optimization/86557
11369         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
11370         EXACT_DIV_EXPR.
11372 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
11374         * config/s390/s390.c (s390_function_profiler): Generate CFI.
11376 2018-07-17  Jeff Law  <law@redhat.com>
11378         * config/arm/arm.c (get_label_padding): Update for recent
11379         changes to label_to_alignment.
11381         PR tree-optimization/86010
11382         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
11384         * config/mips/mips.c (vr4130_align_insns): Update for recent
11385         changes to label_to_alignment.
11387         * config/frv/frv.c (frv_label_align): Update for recent changes
11388         to label_to_alignment.
11390         * config/nios2/nios2.c (nios2_label_align): Update for recent
11391         changes which dropped ALIGN_LABELS_LOG.
11393 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
11395         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
11396         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
11398 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
11400         * config/arc/arc.c (arc_label_align): Use align_labels instead of
11401         deprecated align_labels_log.
11403 2018-07-17  Richard Biener  <rguenther@suse.de>
11405         PR lto/86456
11406         * dwarf2out.c (init_sections_and_labels): Always generate
11407         a debug_line_str_section for early LTO debug.
11408         (dwarf2out_finish): Reset debug_line_str_hash output early.
11409         Bump counter for extra dwarf5 .debug_loc labels to not conflict
11410         with fat LTO part.
11411         (dwarf2out_early_finish): Output debug_line_str.
11413 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11415         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
11416         index register on z196 or later.
11418 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11420         * config/s390/s390.c (s390_default_align): Set default function
11421         alignment to 16.
11422         (s390_override_options_after_change): Call s390_default align.
11423         (s390_option_override_internal): Call s390_default_align.
11424         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
11426 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11428         PR middle-end/86542
11429         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
11430         to _looptemp_ clauses, other than the first two.
11432 2018-07-17  Martin Liska  <mliska@suse.cz>
11434         * opts.c: Do not enable OPT_falign_* for -Os.
11436 2018-07-17  Martin Liska  <mliska@suse.cz>
11438         * align.h (MAX_CODE_ALIGN): New.
11439         (MAX_CODE_ALIGN_VALUE): New.
11440         * common/config/i386/i386-common.c (ix86_handle_option):
11441         (MAX_CODE_ALIGN): Moved to align.h.
11442         * final.c (MAX_CODE_ALIGN): Likewise.
11443         * opts.c (parse_and_check_align_values):
11444         (MAX_CODE_ALIGN): Likewise.
11445         (MAX_CODE_ALIGN_VALUE): Likewise.
11447 2018-07-17  Martin Liska  <mliska@suse.cz>
11449         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
11450         in order to fulfil coding style.
11451         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
11452         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11453         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11454         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
11455         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
11456         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
11457         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
11458         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
11459         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11461 2018-07-17  Martin Liska  <mliska@suse.cz>
11463         * align.h: New file.
11464         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
11465         directly.
11466         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
11467         align_flags of label_to_alignment.
11468         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
11469         align_flags class.
11470         * config/m68k/m68k.c: Do not use removed align_labels_value and
11471         align_loops_value.
11472         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
11473         (LOOP_ALIGN): Likewise.
11474         (LABEL_ALIGN): Likewise.
11475         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11476         Remove not used macro.
11477         (rs6000_loop_align): Change return type to align_flags.
11478         (rs6000_loop_align_max_skip): Remove.
11479         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
11480         Change return type to align_flags.
11481         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11482         Remove not used macro.
11483         (rs6000_loop_align):  Change return type to align_flags.
11484         (rs6000_loop_align_max_skip): Remove.
11485         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
11486         * config/rx/rx-protos.h (rx_align_for_label): Make it
11487         static function.
11488         * config/rx/rx.c (rx_align_for_label): Change return type
11489         to align_flags.
11490         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
11491         macro definitions.
11492         into align_flags class.
11493         (LABEL_ALIGN): Likewise.
11494         (LOOP_ALIGN): Likewise.
11495         * config/s390/s390.c (s390_label_align): Use align_flags
11496         class member.
11497         (s390_asm_output_function_label): Likewise.
11498         * config/sh/sh.c (sh_override_options_after_change):
11499         Use align_flags class directly without macros.
11500         (find_barrier): Likewise.
11501         (barrier_align): Likewise.
11502         (sh_loop_align): Likewise.
11503         * config/spu/spu.c (spu_option_override):
11504         Use align_flags_tuple::get_value instead of removed macros.
11505         (spu_sched_init): Likewise.
11506         * config/spu/spu.h (GTY): Likewise.
11507         * config/visium/visium.c (visium_option_override):
11508         Set "8" as default secondary alignment.
11509         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
11510         in order to guarantee secondary alignment of 8.
11511         * coretypes.h: Include align.h header file.
11512         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
11513         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
11514         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
11515         * doc/tm.texi.in: Likewise.
11516         * final.c (struct label_alignment): Remove not used structure.
11517         (LABEL_ALIGN): Change type to align_flags.
11518         (LOOP_ALIGN): Likewise.
11519         (JUMP_ALIGN): Likewise.
11520         (default_loop_align_max_skip): Remove.
11521         (default_label_align_max_skip): Likewise.
11522         (default_jump_align_max_skip): Likewise.
11523         (default_label_align_after_barrier_max_skip):
11524         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
11525         (LABEL_TO_MAX_SKIP): Remove.
11526         (label_to_alignment): Return align_flags type instead of integer.
11527         (label_to_max_skip): Remove.
11528         (align_fuzz): Use align_flags type.
11529         (compute_alignments): Use align_flags type and use align_flags::max
11530         to combine multiple alignments.
11531         (grow_label_align): Grow vec instead of C array.
11532         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
11533         (shorten_branches):  Use align_flags type and use align_flags::max
11534         to combine multiple alignments.
11535         (final_scan_insn_1): Remove usage of secondary alignment that comes
11536         from label alignment, but instead use proper secondary alignment
11537         which is computed in grow_label_align.
11538         * flags.h (struct align_flags_tuple): Move to align.h.
11539         (struct align_flags): Likewise.
11540         (state_align_loops): Rename to align_loops.
11541         (state_align_jumps): Rename to align_jumps.
11542         (state_align_labels): Rename to align_labels.
11543         (state_align_functions): Rename to align_functions.
11544         (align_loops_log): Remove.
11545         (align_jumps_log): Remove.
11546         (align_labels_log): Remove.
11547         (align_functions_log): Remove.
11548         (align_loops_max_skip): Remove.
11549         (align_jumps_max_skip): Remove.
11550         (align_labels_max_skip): Remove.
11551         (align_functions_max_skip): Remove.
11552         (align_loops_value): Remove.
11553         (align_jumps_value): Remove.
11554         (align_labels_value): Remove.
11555         (align_functions_value): Remove.
11556         * output.h (label_to_alignment): Change return type to align_flags.
11557         (label_to_max_skip): Remove.
11558         * target.def: Remove loop_align_max_skip, label_align_max_skip,
11559         jump_align_max_skip macros.
11560         * targhooks.h (default_loop_align_max_skip): Remove.
11561         (default_label_align_max_skip): Likewise.
11562         (default_jump_align_max_skip): Likewise.
11563         (default_label_align_after_barrier_max_skip): Remove.
11564         * toplev.c (read_log_maxskip): Use ::normalize function.
11565         (parse_N_M): Remove not used argument and also call ::normalize.
11566         (parse_alignment_opts): Do not pass unused arguments.
11567         * varasm.c (assemble_start_function): Use directly align_functions
11568         instead of removed macros.
11569         * system.h: Do not poison removed macros.
11571 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11573         PR middle-end/86539
11574         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
11575         and cond temporaries don't have reference type if iterator has
11576         pointer type.  For init use &for_pre_body instead of pre_p if
11577         for_pre_body is non-empty.
11579 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11581         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
11582         double-double modes to SFmode directly directly.
11583         (trunc<mode>sf2_fprs): Delete.
11585 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11587         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
11588         for conversions between IFmode and the decimal floating point modes.
11589         (init_float128_ieee): Use the correct names for conversions between
11590         KFmode and the decimal floating point modes.
11592 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11594         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
11595         for the conversions between TDmode and IFmode.
11596         (init_float128_ieee): Use more correct names for the conversions
11597         between TDmode and KFmode.
11599 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
11601         PR tree-optimization/86526
11602         * builtins.c (expand_builtin_memcmp): Formatting fixes.
11603         (inline_expand_builtin_string_cmp): Likewise.
11604         (inline_string_cmp): Likewise.  Use c_readstr instead of
11605         builtin_memcpy_read_str.  Add unit_mode temporary.
11607 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11609         PR middle-end/86528
11610         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
11611         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
11613 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11615         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
11616         Alphabetize prototypes of built-in functions, separating out
11617         built-in functions that are listed in this section but should be
11618         described elsewhere.
11620 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
11622         PR target/86511
11623         * expmed.c (emit_store_flag): Do not emit setcc followed by a
11624         conditional move when trapping comparison was split to a
11625         non-trapping one (and vice versa).
11627 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11629         * config/s390/s390.c (s390_function_profiler): Generate nops
11630         instead of profiler call sequences.
11631         * config/s390/s390.opt: Add the new option.
11633 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11635         * config/s390/s390.c (s390_function_profiler): Generate
11636         __mcount_loc section.
11637         * config/s390/s390.opt: Add the new option.
11639 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11641         * common.opt: Add the new warning.
11642         * config/s390/s390.c (s390_function_profiler): Emit "brasl
11643         %r0,__fentry__" when -mfentry is specified.
11644         (s390_option_override_internal): Disallow -mfentry for 31-bit
11645         CPUs.
11646         * config/s390/s390.opt: Add the new option.
11648 2018-07-16  Richard Biener  <rguenther@suse.de>
11650         PR lto/86523
11651         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
11652         for function-local FUNCTION_DECL and RESULT_DECL immediately.
11654 2018-07-16  Martin Liska  <mliska@suse.cz>
11656         PR ipa/86529
11657         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
11658         to ::get_create.
11660 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11662         * config/arc/arcHS.md: Update ARCHS scheduling rules.
11664 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11666         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
11667         for ARCHS4x.
11668         * config/arc/arc-cpus.def (hs4x): New cpu.
11669         (hs4xd): Likewise.
11670         * config/arc/arc-tables.opt: Regenerate.
11671         * config/arc/arc.c (arc_sched_issue_rate): New function.
11672         (TARGET_SCHED_ISSUE_RATE): Define.
11673         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
11674         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
11675         fpu_cvt.
11676         (attr tune): Add ARCHS4x tune values.
11677         (attr tune_dspmpy): Define.
11678         (*tst): Correct instruction type.
11679         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
11680         * config/arc/arcHS4x.md: New file.
11681         * config/arc/fpu.md: Update instruction type attributes.
11682         * config/arc/t-multilib: Regenerate.
11684 2018-07-16  Tom de Vries  <tdevries@suse.de>
11686         PR debug/86455
11687         * var-tracking.c (vt_initialize): Fix pre_dec handling.
11689 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
11691         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
11692         early clobber.
11694 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
11696         PR tree-optimization/86514
11697         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
11698         conversion to a boolean type from a type with greater precision.
11700 2018-07-16  Tom de Vries  <tdevries@suse.de>
11702         * var-tracking.c (vt_initialize): Print adjusted insn slim if
11703         dump_flags request TDF_SLIM.
11705 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
11707         * fold-const.c (int_const_binop_1): Abstract...
11708         (wide_int_binop): ...wide int code here.
11709         (poly_int_binop): ...poly int code here.
11710         Abstract the rest of int_const_binop_1 into int_const_binop.
11711         * fold-const.h (wide_int_binop): New.
11712         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
11713         Remove useless PLUS/MINUS_EXPR case.
11714         (zero_nonzero_bits_from_vr): Move wide int code...
11715         (zero_nonzero_bits_from_bounds): ...here.
11716         (extract_range_from_binary_expr_1): Move mask optimization code...
11717         (range_easy_mask_min_max): ...here.
11718         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
11719         (range_easy_mask_min_max): New.
11721 2018-07-15  Jeff Law  <law@redhat.com>
11723         PR target/85993
11724         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
11725         block.
11727 2018-07-14  Jim Wilson  <jimw@sifive.com>
11729         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
11731 2018-07-14  Paul Koning  <ni1d@arrl.net>
11733         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
11735 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
11737         * lto-streamer-out.c (copy_function_or_variable): Dump info about
11738         copying section.
11740 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
11741             Steve Munroe  <munroesj52@gmail.com>
11743         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
11744         (_mm_andnot_si128): Likewise.
11745         (_mm_or_si128): Likewise.
11746         (_mm_xor_si128): Likewise.
11748 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
11750         PR middle-end/78809
11751         * builtins.c (expand_builtin_memcmp): Inline the calls first
11752         when result_eq is false.
11753         (expand_builtin_strcmp): Inline the calls first.
11754         (expand_builtin_strncmp): Likewise.
11755         (inline_string_cmp): New routine. Expand a string compare
11756         call by using a sequence of char comparison.
11757         (inline_expand_builtin_string_cmp): New routine. Inline expansion
11758         a call to str(n)cmp/memcmp.
11759         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
11760         New option.
11761         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
11763 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
11765         * config/arm/driver-arm.c: Include arm-native.h.
11766         (host_detect_local_cpu): Use auto-generated data tables.
11767         (vendors, arm_cpu_table): Delete.  Move part information to ...
11768         * config/arm/arm-cpus.in: ... here.
11769         * config/arm/parsecpu.awk (gen_native): New function.
11770         (vendor, part): New CPU fields.
11771         (END): Add support for building the native CPU detection tables.
11772         * config/arm/t-arm (arm-native.h): Add build rule.
11773         (driver-arm.o): Add dependency on arm-native.h.
11775 2018-07-13  Richard Biener  <rguenther@suse.de>
11777         PR middle-end/85974
11778         * match.pd (addr1 - addr2): Allow either of the operand to
11779         have a conversion.
11781 2018-07-13  Tom de Vries  <tdevries@suse.de>
11783         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
11784         in remap_ssa_name.
11786 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
11788         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
11789         arrays instead of numbered variables.
11791 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
11793         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
11794         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
11796 2018-07-13  Richard Biener  <rguenther@suse.de>
11798         PR debug/86452
11799         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
11800         instead of get_context_die.
11802 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
11803             Richard Biener  <rguenther@suse.de>
11805         PR middle-end/86489
11806         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
11807         that the loop latch destination where phi is defined.
11809 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
11811         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
11812         (riscv_expand_epilogue): Add assertion to check interrupt mode.
11813         (riscv_set_current_function): Extract getting interrupt type to new
11814         function.
11815         (riscv_get_interrupt_type): New function.
11816         (riscv_merge_decl_attributes): New function, checking interrupt type is
11817         same.
11818         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
11820 2018-07-12  Paul Koning  <ni1d@arrl.net>
11822         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
11823         directive.
11825 2018-07-12  Paul Koning  <ni1d@arrl.net>
11827         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
11828         zero reference, add doloop_end instead.
11829         * doc/md.texi (decrement_and_branch_until_zero): Remove.
11830         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
11831         detail for doloop_end.
11833 2018-07-12  Martin Sebor  <msebor@redhat.com>
11835         PR c/86453
11836         * attribs.c (decl_attributes): Reject conflicting attributes before
11837         calling attribute handlers.
11839 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
11841         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
11842          parameter.
11843         (gcc::dump_manager::get_dump_file_name): likewise.
11844         (dump_begin): Likewise.
11845         * dumpfile.h (dump_begin): Update prototype.
11846         (gcc::dump_manager::get_dump_file_name,
11847         gcc::dump_manager::get_dump_file_name): Update prototype.
11849 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11851         * internal-fn.h (vectorizable_internal_fn_p): New function.
11852         * tree-vect-slp.c (compatible_calls_p): Likewise.
11853         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
11854         to internal functions.
11855         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
11857 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11859         * fold-const.h (inverse_conditions_p): Declare.
11860         * fold-const.c (inverse_conditions_p): New function.
11861         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
11862         that test the inverse condition of a conditional internal function.
11863         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
11864         * internal-fn.c (internal_fn_mask_index): Handle conditional
11865         internal functions.
11866         (vectorized_internal_fn_supported_p): New function.
11867         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
11868         (any_pred_load_store): Replace with...
11869         (need_to_predicate): ...this new variable.
11870         (redundant_ssa_names): New variable.
11871         (ifcvt_can_use_mask_load_store): Move initial checks to...
11872         (ifcvt_can_predicate): ...this new function.  Handle tree codes
11873         for which a conditional internal function exists.
11874         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
11875         instead of ifcvt_can_use_mask_load_store.  Update after variable
11876         name change.
11877         (predicate_load_or_store): New function, split out from
11878         predicate_mem_writes.
11879         (check_redundant_cond_expr): New function.
11880         (value_available_p): Likewise.
11881         (predicate_rhs_code): Likewise.
11882         (predicate_mem_writes): Rename to...
11883         (predicate_statements): ...this.  Use predicate_load_or_store
11884         and predicate_rhs_code.
11885         (combine_blocks, tree_if_conversion): Update after above name changes.
11886         (ifcvt_local_dce): Handle redundant_ssa_names.
11887         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
11888         general conditional functions.
11889         * tree-vect-stmts.c (vectorizable_call): Likewise.
11891 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11892             Alan Hayward  <alan.hayward@arm.com>
11893             David Sherwood  <david.sherwood@arm.com>
11895         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
11896         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
11897         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
11898         plus and minus and convert them into IFN_COND_FMA-based sequences.
11899         (convert_mult_to_fma): Handle conditional plus and minus.
11901 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11903         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
11904         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
11905         (cond_fnms_optab): New optabs.
11906         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
11907         internal functions.
11908         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
11909         * internal-fn.h (get_conditional_internal_fn): Declare.
11910         (get_unconditional_internal_fn): Likewise.
11911         * internal-fn.c (cond_ternary_direct): New macro.
11912         (expand_cond_ternary_optab_fn): Likewise.
11913         (direct_cond_ternary_optab_supported_p): Likewise.
11914         (FOR_EACH_COND_FN_PAIR): Likewise.
11915         (get_conditional_internal_fn): New function.
11916         (get_unconditional_internal_fn): Likewise.
11917         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
11918         (gimple_match_op::gimple_match_op): Add a new overload for 5
11919         operands.
11920         (gimple_match_op::set_op): Likewise.
11921         (gimple_resimplify5): Declare.
11922         * genmatch.c (decision_tree::gen): Generate simplifications for
11923         5 operands.
11924         * gimple-match-head.c (gimple_simplify): Define an overload for
11925         5 operands.  Handle calls with 5 arguments in the top-level overload.
11926         (convert_conditional_op): Handle conversions from unconditional
11927         internal functions to conditional ones.
11928         (gimple_resimplify5): New function.
11929         (build_call_internal): Pass a fifth operand.
11930         (maybe_push_res_to_seq): Likewise.
11931         (try_conditional_simplification): Try converting conditional
11932         internal functions to unconditional internal functions.
11933         Handle 3-operand unconditional forms.
11934         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
11935         Define ternary equivalents of the current rules for binary conditional
11936         internal functions.
11937         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
11938         ternary operations.
11939         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
11940         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
11941         (optab): Handle them.
11942         (SVE_COND_FP_TERNARY): New int iterator.
11943         (sve_fmla_op, sve_fmad_op): New int attributes.
11944         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
11945         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
11946         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
11948 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11950         * target.def (preferred_else_value): New target hook.
11951         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
11952         * doc/tm.texi: Regenerate.
11953         * targhooks.h (default_preferred_else_value): Declare.
11954         * targhooks.c (default_preferred_else_value): New function.
11955         * internal-fn.h (conditional_internal_fn_code): Declare.
11956         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
11957         (get_conditional_internal_fn): Use it.
11958         (conditional_internal_fn_code): New function.
11959         * gimple-match.h (gimple_match_cond): New struct.
11960         (gimple_match_op): Add a cond member function.
11961         (gimple_match_op::gimple_match_op): Update all forms to take a
11962         gimple_match_cond.
11963         * genmatch.c (expr::gen_transform): Use the same condition as res_op
11964         for the suboperation, but don't specify a particular else_value.
11965         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
11966         (visit_nary_op, visit_reference_op_load): Pass
11967         gimple_match_cond::UNCOND to the gimple_match_op constructor.
11968         * gimple-match-head.c: Include tree-eh.h
11969         (convert_conditional_op): New function.
11970         (maybe_resimplify_conditional_op): Likewise.
11971         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
11972         (gimple_resimplify2): Likewise.
11973         (gimple_resimplify3): Likewise.
11974         (gimple_resimplify4): Likewise.
11975         (maybe_push_res_to_seq): Return null for conditional operations.
11976         (try_conditional_simplification): New function.
11977         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
11978         constructor.
11979         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
11980         IFN_COND_* call.
11981         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
11982         function.
11983         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
11985 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
11987         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11988         DECL_FCONTEXT
11989         (hash_tree): Do not hash DECL_FCONTEXT
11990         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
11991         Do not stream DECL_FCONTEXT.
11992         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
11993         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
11995 2018-07-12  Richard Biener  <rguenther@suse.de>
11997         PR debug/86462
11998         * dwarf2out.c (gen_block_die): Only output blocks when they have
11999         at least one !DECL_IGNORED_P variable.
12001 2018-07-12  Richard Biener  <rguenther@suse.de>
12003         PR target/84829
12004         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12005         Remove -mieee-fp handling.
12007 2018-07-12  Richard Biener  <rguenther@suse.de>
12009         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12010         left-over from last patch.
12012 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12014         PR tree-optimization/86492
12015         * gimple-ssa-store-merging.c
12016         (imm_store_chain_info::coalesce_immediate_stores): Call
12017         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12019 2018-07-12  Richard Biener  <rguenther@suse.de>
12021         PR middle-end/86479
12022         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12023         move possibly trapping operations into the conditional.
12025 2018-07-12  Richard Biener  <rguenther@suse.de>
12027         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12028         (vn_lookup_simplify_result): Remove recursion limit applied
12029         here.
12030         (vn_nary_build_or_lookup_1): Adjust.
12031         (try_to_simplify): Likewise.
12032         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12033         here.
12034         (gimple_resimplify2): Likewise.
12035         (gimple_resimplify3): Likewise.
12036         (gimple_resimplify4): Likewise.
12038 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12040         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12041         Use __mmask64 type instead of __mmask8 for __M argument.
12042         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12043         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12044         __U argument.
12045         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12046         __mmask16 for __M argument.
12047         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12048         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12049         to __mmask16 instead of __mmask8.
12050         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12051         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12052         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12053         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12054         instead of __mmask16 for __U argument.
12055         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12056         __mmask16 instead of __mmask8 for __U argument.
12057         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12058         __U argument.
12059         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12060         __mmask16.
12061         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12062         argument.
12063         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12064         __U argument.
12065         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12066         __mmask16.
12067         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12068         of __mmask16.
12069         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12070         __U argument.
12071         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12072         __U argument.
12073         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12074         __U argument.
12075         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12076         __U argument.
12077         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12078         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12079         return type as well as __M argument type and all casts from __mmask8
12080         to __mmask32.
12081         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12082         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12083         return type as well as __M argument type and all casts from __mmask8
12084         to __mmask16.
12085         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12086         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12087         return type as well as __M argument type and all casts from __mmask8
12088         to __mmask32.
12089         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12090         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12091         return type as well as __M argument type and all casts from __mmask8
12092         to __mmask16.
12093         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12094         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12095         __mmask16.
12097 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
12099         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12100         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12101         for __U argument.
12103 2018-07-11  Paul Koning  <ni1d@arrl.net>
12105         * doc/md.texi (define_subst): Document how multiple occurrences of
12106         the same argument in the replacement pattern are handled.
12108 2018-07-11  Paul Koning  <ni1d@arrl.net>
12110         * doc/extend.texi (Common Variable Attributes): Move "mode" into
12111         alphabetical order.
12112         (Common Type Attributes): Add "mode" attribute.
12114 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
12116         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12117         stream DECL_ORIGINAL_TYPE.
12118         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12119         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12120         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12121         Do not walk original type.
12122         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12123         external decls.
12124         (write_ts_decl_non_common_tree_pointers): Do not stream
12125         DECL_ORIGINAL_TYPE
12126         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12127         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12129 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12131         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12132         thread twice from the same starting edge.
12134 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
12136         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12137         * gimple.c (gimple_call_nonnull_result_p): ...here...
12138         (gimple_call_nonnull_arg): ...and here.
12139         * gimple.h (gimple_call_nonnull_result_p): New.
12140         (gimple_call_nonnull_arg): New.
12142 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
12144         * config/arm/arm-cpus.in: Move information from fpu field of each
12145         cpu definition to the isa field.
12146         * config/arm/parsecpu.awk (fpu): Delete match rule.
12147         (gen_comm_data): Don't add bits from the CPU's FPU entry.
12149 2018-07-11  Richard Biener  <rguenther@suse.de>
12151         PR debug/86457
12152         * dwarf2out.c (init_sections_and_labels): Use
12153         output_asm_line_debug_info consistently.
12154         (dwarf2out_early_finish): Likewise.
12155         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12156         type units.
12158 2018-07-11  Richard Biener  <rguenther@suse.de>
12160         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12161         Rework father_bb setting in a way to avoid propagating constants
12162         multiple times on a loop body.
12164 2018-07-10  Mark Wielaard  <mark@klomp.org>
12166         PR debug/86459
12167         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12169 2018-07-10  Richard Biener  <rguenther@suse.de>
12171         * hash-map.h (hash_map::iterator::operator*): Return
12172         references to key and value.
12174 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
12176         PR c++/86443
12177         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12178         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12179         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12180         (gimplify_omp_for): For composite loops, move outer
12181         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12182         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
12183         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
12184         TREE_LIST for both the original class iterator and the "last" helper
12185         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
12186         loop, remember has_decl_expr from outer composite loops for the
12187         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12189 2018-07-09  Martin Sebor  <msebor@redhat.com>
12191         PR middle-end/77357
12192         PR middle-end/86428
12193         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12194         accessing implicitly initialized array elements.
12195         * expr.c (string_constant): Handle string initializers of
12196         character arrays within aggregates.
12197         * gimple-fold.c (fold_array_ctor_reference): Add argument.
12198         Store element offset.  As a special case, handle zero size.
12199         (fold_nonarray_ctor_reference): Same.
12200         (fold_ctor_reference): Add argument.  Store subobject offset.
12201         * gimple-fold.h (fold_ctor_reference): Add argument.
12203 2018-07-09  Paul Koning  <ni1d@arrl.net>
12205         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12206         (pdp11_insn_cost): New function.
12207         (pdp11_md_asm_adjust): New function.
12208         (TARGET_INVALID_WITHIN_DOLOOP): Define.
12209         (pdp11_rtx_costs): Update to match machine better.
12210         (output_addr_const_pdp11): Correct format mismatch warnings.
12211         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12212         * config/pdp11/pdp11.md: General change to add base_cost and/or
12213         length attributes for use by new pdp11_insn_cost function.
12214         (MIN_BRANCH): Correct definition.
12215         (MIN_SOB): Ditto.
12216         (doloop_end): Use standard pattern name for looping pattern.
12217         (doloop_end_nocc): New.
12218         (movsf): Add another constraint alternative.
12219         (zero_extendqihi2): Add constraint alternatives for not in place
12220         extend.
12221         (zero_extendhisi2): Remove.
12222         (shift patterns): Add CC handling variants.
12223         (bswaphi2): New.
12224         (bswapsi2): New.
12225         (rothi3): New.
12226         (define_peephole2): New peephole to recognize mov that sets CC for
12227         subsequent test.
12229 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12231         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12232         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12233         wi::add.
12235 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
12237         PR c/86420
12238         * real.c (real_nextafter): Return true if result is denormal.
12240 2018-07-09  Martin Liska  <mliska@suse.cz>
12242         * common.opt: Add back wrongly removed attribute.
12244 2018-07-09  Richard Biener  <rguenther@suse.de>
12246         PR debug/86413
12247         * dwarf2out.c (gen_block_die): For an early generated DIE
12248         always output high/low PC attributes.
12250 2018-07-09  Tom de Vries  <tdevries@suse.de>
12252         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12253         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12254         onto VAR_DECL with abstract origin.
12256 2018-07-07  Jim Wilson  <jimw@sifive.com>
12258         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12260 2018-07-07  Tom de Vries  <tdevries@suse.de>
12262         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12264 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
12266         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12267         overflow_type.
12268         (combine_bound): Use wide-int overflow calculation instead of
12269         rolling our own.
12270         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12271         overflow_type.
12272         * fold-const.c (int_const_binop_2): Same.
12273         (extract_muldiv_1): Same.
12274         (fold_div_compare): Same.
12275         (fold_abs_const): Same.
12276         * match.pd: Same.
12277         * poly-int.h (add): Same.
12278         (sub): Same.
12279         (neg): Same.
12280         (mul): Same.
12281         * predict.c (predict_iv_comparison): Same.
12282         * profile-count.c (slow_safe_scale_64bit): Same.
12283         * simplify-rtx.c (simplify_const_binary_operation): Same.
12284         * tree-chrec.c (tree_fold_binomial): Same.
12285         * tree-data-ref.c (split_constant_offset_1): Same.
12286         * tree-if-conv.c (idx_within_array_bound): Same.
12287         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12288         * tree-ssa-phiopt.c (minmax_replacement): Same.
12289         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12290         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12291         * vr-values.c (vr_values::adjust_range_with_scev): Same.
12292         * wide-int.cc (wi::add_large): Same.
12293         (wi::mul_internal): Same.
12294         (wi::sub_large): Same.
12295         (wi::divmod_internal): Same.
12296         * wide-int.h: Change overflow type to overflow_type for neg, add,
12297         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12298         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12299         mul_internal, divmod_internal.
12300         (overflow_type): New enum.
12301         (accumulate_overflow): New.
12303 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
12305         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12306         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12308 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12310         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12311         argument is checked for zero before entering loop, avoid checking again.
12313 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12315         * gimplify.h (generic_expr_could_trap_p): Set as global function.
12316         * gimplify.h (generic_expr_could_trap_p): Likwise.
12317         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
12319 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
12321         PR tree-optimization/86401
12322         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
12323         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
12324         (fold_bit_and_mask): ... here.  New helper function for match.pd.
12325         * fold-const.h (fold_bit_and_mask): Declare.
12326         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
12328 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
12330         PR target/86324
12331         * target.def (translate_mode_attribute): New hook.
12332         * targhooks.h (default_translate_mode_attribute): Declare.
12333         * targhooks.c (default_translate_mode_attribute): New function.
12334         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
12335         * doc/tm.texi: Regenerate.
12336         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
12337         (rs6000_translate_mode_attribute): New function.
12339 2018-07-06  Paul Koning  <ni1d@arrl.net>
12341         * doc/md.texi (define_split): Document DONE and FAIL.
12342         (define_peephole2): Ditto.
12344 2018-07-05  Jeff Law  <law@redhat.com>
12346         PR tree-optimization/86010
12347         * tree-ssa-dse.c (compute_trims): More aggressively trim at
12348         both the head and tail of mem* and str* calls.
12350 2018-07-05  Jim Wilson  <jimw@sifive.com>
12352         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
12354 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
12356         * config/aarch64/aarch64-simd.md: correct flags text for
12357         MIN_EXPR replacement.
12359 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
12361         * configure: Regenerated.
12363 2018-07-05  Carl Love  <cel@us.ibm.com>
12365         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
12366         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
12367         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
12368         VSX_BUILTIN_DOUBLEL_V4SF.
12370 2018-07-05  Martin Sebor  <msebor@redhat.com>
12372         PR c++/86400
12373         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
12374         than its domain to compute its the upper bound of a char array.
12376 2018-07-05  Nathan Sidwell  <nathan@acm.org>
12378         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
12379         * doc/cpp.texi: Update comment.
12380         * doc/tm.texi: Rebuilt.
12381         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
12382         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
12383         * doc/extend.texi (Backwards Compatibility): Clarify it is system
12384         headers affected by extern "C".
12385         * system.h: Poison NO_IMPLICIT_EXTERN_C.
12386         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
12387         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
12388         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
12389         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
12390         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
12391         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
12392         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
12393         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
12394         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
12395         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
12396         config/sparc/sp64-elf.h, config/spu/spu.h,
12397         config/stormy16/stormy16.h, config/v850/v850.h,
12398         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
12399         define NO_IMPLICIT_EXTERN_C.
12400         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
12402 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12404         PR target/84711
12405         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12406         instead of GET_MODE_SIZE when comparing Units.
12408 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12410         PR target/84711
12411         * rtlanal.c (set_noop_p): Constrain on mode change,
12412         include hard-reg-set.h
12414 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12416         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
12418 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
12420         Revert
12421         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12423         PR sanitizer/84250
12424         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12425         libasan.
12426         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12428 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12430         PR sanitizer/84250
12431         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12432         libasan.
12433         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12435 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
12437         PR middle-end/86380
12438         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
12440 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
12442         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
12443         neg_*_op* variables.
12445 2018-07-04  Martin Liska  <mliska@suse.cz>
12447         * tree-switch-conversion.c: Define
12448         max_ratio_for_speed and max_ratio_for_size constants.
12450 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12451             Martin Liska  <mliska@suse.cz>
12453         PR middle-end/66240
12454         PR target/45996
12455         PR c/84100
12456         * common.opt: Rename align options with 'str_' prefix.
12457         * common/config/i386/i386-common.c (set_malign_value): New
12458         function.
12459         (ix86_handle_option): Use it to set -falign-* options/
12460         * config/aarch64/aarch64-protos.h (struct tune_params): Change
12461         type from int to string.
12462         * config/aarch64/aarch64.c: Update default values from int
12463         to string.
12464         * config/alpha/alpha.c (alpha_override_options_after_change):
12465         Likewise.
12466         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
12467         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12468         max skip conditionally.
12469         * config/i386/freebsd.h (SUBALIGN_LOG): New.
12470         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12471         max skip conditionally.
12472         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12473         max skip conditionally.
12474         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
12475         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12476         max skip conditionally.
12477         * config/i386/i386.c (struct ptt): Change type from int to
12478         string.
12479         (ix86_default_align): Set default values.
12480         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
12481         max skip conditionally.
12482         * config/i386/iamcu.h (SUBALIGN_LOG): New.
12483         (ASM_OUTPUT_MAX_SKIP_ALIGN):
12484         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
12485         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12486         max skip conditionally.
12487         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
12488         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
12489         * config/i386/x86-64.h (SUBALIGN_LOG): New.
12490         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12491         max skip conditionally.
12492         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12493         * config/ia64/ia64.c (ia64_option_override): Set default values
12494         for alignment options.
12495         * config/m68k/m68k.c: Handle new str_align_* options.
12496         * config/mips/mips.c (mips_set_compression_mode): Change
12497         type of constants.
12498         (mips_option_override): Set default values for options.
12499         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
12500         Likewise.
12501         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12502         Likewise.
12503         * config/rx/rx.c (rx_option_override): Likewise.
12504         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
12505         (LABEL_ALIGN): Use align_labels_log.
12506         (LOOP_ALIGN): Use align_loops_align.
12507         * config/s390/s390.c (s390_asm_output_function_label): Use new
12508         macros.
12509         * config/sh/sh.c (sh_override_options_after_change):
12510         Change type of constants.
12511         * config/spu/spu.c (spu_sched_init): Likewise.
12512         * config/sparc/sparc.c (sparc_option_override): Set default
12513         values for options.
12514         * config/visium/visium.c (visium_option_override): Likewise.
12515         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
12516         emit p2align format with last argument if it's not needed.
12517         * doc/invoke.texi: Document extended format of -falign-*.
12518         * final.c: Use align_labels alignment.
12519         * flags.h (struct target_flag_state): Change type to use
12520         align_flags.
12521         (struct align_flags_tuple): New.
12522         (struct align_flags): Likewise.
12523         (align_loops_log): Redefine macro to use new types.
12524         (align_loops_max_skip): Redefine macro to use new types.
12525         (align_jumps_log): Redefine macro to use new types.
12526         (align_jumps_max_skip): Redefine macro to use new types.
12527         (align_labels_log): Redefine macro to use new types.
12528         (align_labels_max_skip): Redefine macro to use new types.
12529         (align_functions_log): Redefine macro to use new types.
12530         (align_loops): Redefine macro to use new types.
12531         (align_jumps): Redefine macro to use new types.
12532         (align_labels): Redefine macro to use new types.
12533         (align_functions): Redefine macro to use new types.
12534         (align_functions_max_skip): Redefine macro to use new types.
12535         (align_loops_value): New macro.
12536         (align_jumps_value): New macro.
12537         (align_labels_value): New macro.
12538         (align_functions_value): New macro.
12539         * function.c (invoke_set_current_function_hook): Propagate
12540         alignment values from flags to global variables default in
12541         topleev.h.
12542         * ipa-icf.c (sem_function::equals_wpa): Use
12543         cl_optimization_option_eq instead of memcmp.
12544         * lto-streamer.h (cl_optimization_stream_out): Support streaming
12545         of string types.
12546         (cl_optimization_stream_in): Likewise.
12547         * optc-save-gen.awk: Support strings in cl_optimization.
12548         * opth-gen.awk: Likewise.
12549         * opts.c (finish_options): Remove error checking of invalid
12550         value ranges.
12551         (MAX_CODE_ALIGN): Remove.
12552         (MAX_CODE_ALIGN_VALUE): Likewise.
12553         (parse_and_check_align_values): New function.
12554         (check_alignment_argument): Likewise.
12555         (common_handle_option): Use check_alignment_argument.
12556         * opts.h (parse_and_check_align_values): Declare.
12557         * toplev.c (init_alignments): Remove.
12558         (read_log_maxskip): New.
12559         (parse_N_M): Likewise.
12560         (parse_alignment_opts): Likewise.
12561         (backend_init_target): Remove usage of init_alignments.
12562         * toplev.h (parse_alignment_opts): Declare.
12563         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
12564         argument.
12565         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
12566         * tree.c (cl_option_hasher::equal): New.
12567         * varasm.c: Use new global macros.
12569 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12571         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
12572         Use a simpler align directive also if MAXSKIP = ALIGN-1.
12573         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12574         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12575         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12576         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12577         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
12578         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
12579         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
12580         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12581         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12582         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12583         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12585 2018-07-04  Martin Liska  <mliska@suse.cz>
12586             Jonathan Wakely  <jwakely@redhat.com>
12588         * coverage.c: Use correct type.
12589         * doc/invoke.texi: Language correction.
12591 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
12593         PR target/85620
12594         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
12595         ENDBRANCH for non-tail call which may return via indirect branch.
12596         * doc/extend.texi: Document indirect_return attribute.
12598 2018-07-03  Martin Sebor  <msebor@redhat.com>
12600         PR tree-optimization/86274
12601         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
12602         precondition.
12603         (format_floating): Correct handling of infinities and NaNs.
12605 2018-07-03  Martin Sebor  <msebor@redhat.com>
12607         * print-tree.c (print_real_cst): New function.
12608         (print_node_brief): Call it.
12609         (print_node): Ditto.
12611 2018-07-03  Jeff Law  <law@redhat.com>
12613         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
12614         into a single pattern.
12616         * config/h8300/h8300.md (ors code_iterator): New.
12617         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
12618         a single pattern and single splitter.
12619         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
12620         (iorqi3_1, xorqi3_1): Likewise.
12621         (iorqi3, xorqi3 expanders): Similarly.
12623         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
12624         (movmd_internal) into a single pattern using the P mode iterator.
12625         (movmd splitters): Similarly.
12626         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
12627         (movsd splitters): Similarly.
12629         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
12630         ADDB, ADDW and ADDL into a single ADD attribute which selects the
12631         right table based on the size of the operand.
12632         * config/h8300/h8300.md (length_table): Corresponding changes. All
12633         references to "addb", "addw" and "addl" changed to "add".
12634         (btst patterns): Merge two variants into a single pattern.
12635         (tstqi, tsthi): Likewise.
12636         (addhi3_incdec, addsi3_incdec): Likewise.
12637         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
12638         (mulhi3, mulsi3): Likewise.
12639         (udivhi3, udivsi3): Likewise.
12640         (divhi3, divsi3): Likewise.
12641         (andorqi3, andorhi3, andorsi3): Likewise.
12643 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
12645         PR target/85694
12646         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
12647         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
12649 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12651         PR tree-optimization/85694
12652         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
12653         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
12654         UNSPEC_URHADD.
12655         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
12656         (<u>avg<mode>3_ceil): New patterns.
12658 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
12660         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
12661         scan-tree-dump directive.
12662         * gcc.dg/vect/slp-perm-2.c: Likewise.
12663         * gcc.dg/vect/slp-perm-3.c: Likewise.
12664         * gcc.dg/vect/slp-perm-5.c: Likewise.
12665         * gcc.dg/vect/slp-perm-6.c: Likewise.
12666         * gcc.dg/vect/slp-perm-7.c: Likewise.
12667         * gcc.dg/vect/slp-perm-8.c: Likewise.
12669 2018-07-03  Marek Polacek  <polacek@redhat.com>
12671         PR middle-end/86202
12672         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
12674 2018-07-03  Richard Biener  <rguenther@suse.de>
12676         PR ipa/86389
12677         * tree-ssa-structalias.c (find_func_clobbers): Properly
12678         handle indirect calls.
12680 2018-07-03  Jeff Law  <law@redhat.com>
12682         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
12683         (shifts): New code iterator.
12684         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
12685         expander.  Fix HImode handling on H8/SX.
12686         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
12687         (subqi3, subhi3, subsi3 expanders): Likewise.
12688         (andqi3, andhi3, andsi3 expanders): Likewise.
12689         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
12690         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
12691         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
12692         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
12693         (zero_extendqihi2, zero_extendqisi2): Likewise.
12694         (extendqihi2, extendqisi2): Likewise.
12695         (rotlqi3, rotlhi3, rotlsi3): Likewise.
12696         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
12697         (rotlqi3_1, rotlhi3_1): Likewise.
12698         (logicalhi3_sn, logicalsi3_sn): Likewise.
12699         (logicalhi3, logicalsi3): Likewise.
12701 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12703         * tree-vect-patterns.c (vect_recog_rotate_pattern)
12704         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12705         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
12706         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
12707         type to append_pattern_def_seq instead of creating a stmt_vec_info
12708         directly.
12709         (build_mask_conversion): Likewise.  Remove vinfo argument.
12710         (vect_add_conversion_to_patterm): Likewise, renaming to...
12711         (vect_add_conversion_to_pattern): ...this.
12712         (vect_recog_mask_conversion_pattern): Update call to
12713         build_mask_conversion.  Pass the vector type to
12714         append_pattern_def_seq here too.
12715         (vect_recog_gather_scatter_pattern): Update call to
12716         vect_add_conversion_to_pattern.
12718 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12720         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
12721         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12722         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
12723         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
12724         STMT_VINFO_PATTERN_DEF_SEQ to null here.
12725         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
12726         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
12727         append_pattern_def_seq instead of new_pattern_def_seq.
12728         (vect_recog_divmod_pattern): Do both of the above.
12729         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
12730         is null.
12732 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12734         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
12735         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
12736         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
12737         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
12738         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
12739         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12740         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
12741         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12742         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12743         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
12744         parameter with a single stmt_vec_info.
12745         (vect_recog_func_ptr): Likewise.
12746         (vect_recog_gather_scatter_pattern): Likewise, folding in...
12747         (vect_try_gather_scatter_pattern): ...this.
12748         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
12749         the stmt_vec_info of the statement to be matched.  Don't clear
12750         STMT_VINFO_RELATED_STMT.
12751         (vect_pattern_recog): Update call accordingly.
12753 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12755         PR tree-optimization/85694
12756         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
12757         (uavgM3_ceil): Document new optabs.
12758         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
12759         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
12760         functions.
12761         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
12762         (savg_ceil_optab): New optabs.
12763         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
12764         (vect_vect_recog_func_ptrs): Add it.
12765         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
12766         constant directly from the associated lhs.
12768 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12770         * tree-vect-patterns.c (vect_split_statement): New function.
12771         (vect_convert_input): Use it to try to split an existing cast.
12773 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12775         * poly-int.h (print_hex): New function.
12776         * dumpfile.h (dump_dec, dump_hex): Declare.
12777         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
12778         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
12779         min_input_precision, operation_precision and operation_sign.
12780         * tree-vect-patterns.c (vect_get_range_info): New function.
12781         (vect_same_loop_or_bb_p, vect_single_imm_use)
12782         (vect_operation_fits_smaller_type): Delete.
12783         (vect_look_through_possible_promotion): Add an optional
12784         single_use_p parameter.
12785         (vect_recog_over_widening_pattern): Rewrite to use new
12786         stmt_vec_info infomration.  Handle one operation at a time.
12787         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
12788         (vect_truncatable_operation_p, vect_set_operation_type)
12789         (vect_set_min_input_precision): New functions.
12790         (vect_determine_min_output_precision_1): Likewise.
12791         (vect_determine_min_output_precision): Likewise.
12792         (vect_determine_precisions_from_range): Likewise.
12793         (vect_determine_precisions_from_users): Likewise.
12794         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
12795         (vect_vect_recog_func_ptrs): Put over_widening first.
12796         Add cast_forwprop.
12797         (vect_pattern_recog): Call vect_determine_precisions.
12799 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12801         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
12802         statements that have been replaced by further pattern statements.
12803         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
12805 2018-07-03  Richard Biener  <rguenther@suse.de>
12807         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
12808         always set *dt.  Dump vectype in vectype overload.
12809         * dumpfile.h (dump_gimple_expr): New function.
12810         (dump_gimple_expr_loc): Likewise.
12811         * dumpfile.c (dump_gimple_expr): New function.
12812         (dump_gimple_expr_loc): Likewise.
12814 2018-07-02  Jeff Law  <law@redhat.com>
12816         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
12817         the H8/300, H8/300H and H8/S variants into a single pattern.
12818         (movhi_h8300, movqi_h8300hs): Similarly.
12819         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
12820         (QHI mode iterator): New.
12822         * config/h8300/h8300.md: Remove trailing whitespace.
12824 2018-07-02  Jim Wilson  <jimw@sifive.com>
12826         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
12827         instead of emit_insn for interrupt returns.
12828         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
12829         (riscv_sret, riscv_uret): Likewise.
12831 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
12833         * pretty-print.c (selftest::test_pp_format): Move save and restore
12834         of quotes to class auto_fix_quotes, and add an instance.
12835         * selftest.c: Include "intl.h".
12836         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
12837         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
12838         * selftest.h (selftest::auto_fix_quotes): New class.
12840 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
12842         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
12843         (aarch64_sve_prepare_conditional_op): Remove.
12844         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
12845         Allow aarch64_simd_reg_or_zero as select operand; remove
12846         the aarch64_sve_prepare_conditional_op call.
12847         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
12848         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12849         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
12850         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
12851         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
12852         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
12853         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
12854         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
12855         and a splitters to match all of the *_any patterns.
12856         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
12858         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
12859         (SVE_COND_FP_BINARY_REV): Remove.
12860         (sve_int_op_rev, sve_fp_op_rev): New.
12861         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
12862         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
12863         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
12864         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
12865         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
12866         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
12867         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
12868         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
12869         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
12871         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
12872         Remove match_dup 1 from the inner unspec.
12873         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12875         * config/aarch64/aarch64.md (movprfx): New attr.
12876         (length): Default movprfx to 8.
12877         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
12878         (*madd<SVE_I>, *msub<SVE_I): Likewise.
12879         (*<su>mul<SVE_I>3_highpart): Likewise.
12880         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
12881         (*v<ASHIFT><SVE_I>3): Likewise.
12882         (*<su><MAXMIN><SVE_I>3): Likewise.
12883         (*<su><MAXMIN><SVE_F>3): Likewise.
12884         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
12885         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
12886         (*div<SVE_F>4): Likewise.
12888 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
12890         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
12891         in dump string.
12893 2018-07-02  Richard Biener  <rguenther@suse.de>
12895         PR tree-optimization/86363
12896         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
12897         memset argument refers to a non-variable address.
12899 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
12901         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
12902         {PLUS,MINUS}_EXPR code to...
12903         (adjust_symbolic_bound): ...here,
12904         (combine_bound): ...here,
12905         (set_value_range_with_overflow): ...and here.
12907 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
12909         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
12910         code...
12911         (extract_range_from_abs_expr): ...here.
12913 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
12915         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
12916         -fno-omit-frame-pointer when not optimizing.
12918 2018-07-02  Martin Liska  <mliska@suse.cz>
12920         PR ipa/86279
12921         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
12922         (propagate_nothrow): Likewise.
12924 2018-07-02  Martin Liska  <mliska@suse.cz>
12926         PR ipa/86323
12927         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
12929 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
12931         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
12932         function in r262149, changing "loc" param from source_location to
12933         const dump_location_t &.
12934         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
12935         declaration, as above.
12937 2018-07-01  Paul Koning  <ni1d@arrl.net>
12939         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12940         -munit-asm, -mgnu-asm, -mdec-asm.
12941         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
12942         (pdp11_output_labelref): New.
12943         (pdp11_output_def): New.
12944         (pdp11_output_addr_vec_elt): New.
12945         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
12946         %# and %@ format codes.
12947         (pdp11_option_override): New.
12948         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
12949         (pdp11_output_ident): New.
12950         (pdp11_asm_named_section): New.
12951         (pdp11_asm_init_sections): New.
12952         (pdp11_file_start): New.
12953         (pdp11_file_end): New.
12954         (output_ascii): Use .ascii/.asciz for -mdec-asm.
12955         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
12956         %o, like %c but octal.
12957         (pdp11_option_override): New.
12958         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
12959         -mdec-asm.
12960         (DATA_SECTION_ASM_OP): Ditto.
12961         (READONLY_DATA_SECTION_ASM_OP): New.
12962         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
12963         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
12964         (ASM_OUTPUT_LABELREF): Ditto.
12965         (ASM_OUTPUT_DEF): Ditto.
12966         (ASM_OUTPUT_EXTERNAL): New.
12967         (ASM_OUTPUT_SOURCE_FILENAME): New.
12968         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
12969         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
12970         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
12971         %# and %@ format codes.
12972         * config/pdp11/pdp11.opt (mgnu-asm): New.
12973         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
12974         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
12975         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
12977 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
12979         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
12980         dereferencing path[] beyond its length.
12981         (debug_path): New.
12982         (debug_all_paths): New.
12983         (rewire_first_differing_edge): New.
12984         (adjust_paths_after_duplication): New.
12985         (duplicate_thread_path): Call adjust_paths_after_duplication.
12986         Add new argument.
12987         (thread_through_all_blocks): Add new argument to
12988         duplicate_thread_path.
12990 2018-06-30  Jim Wilson  <jimw@sifive.com>
12992         * config/riscv/predicates.md (p2m1_shift_operand): New.
12993         (high_mask_shift_operand): New.
12994         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
12995         pattern using p2m1_shift_operand.
12996         (lshsi3_zero_extend_3+2): New combiner pattern using
12997         high_mask_shift_operand.
12999 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13001         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13002         split out from...
13003         (vect_recog_rotate_pattern): ...here.
13004         (vect_convert_input): Try to insert casts of invariants in the
13005         preheader.
13006         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13007         preheader to be empty.
13009 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13011         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13012         vector type.  If given, install it in the new statement's
13013         STMT_VINFO_VECTYPE.
13014         (vect_element_precision): New function.
13015         (vect_unpromoted_value): New struct.
13016         (vect_unpromoted_value::vect_unpromoted_value): New function.
13017         (vect_unpromoted_value::set_op): Likewise.
13018         (vect_look_through_possible_promotion): Likewise.
13019         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13020         (vect_widened_op_tree, vect_convert_input): Likewise.
13021         (vect_convert_inputs, vect_convert_output): Likewise.
13022         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13023         to handle the optional cast of the multiplication result and
13024         vect_widened_op_tree to detect the widened multiplication itself.
13025         Do not require the input and output of promotion casts to have
13026         the same sign, but base the signedness of the operation on the
13027         input rather than the result.  If the pattern includes two
13028         promotions, check that those promotions have the same sign.
13029         Do not restrict the MULT_EXPR handling to a double-width result;
13030         handle quadruple-width results and wider.  Use vect_convert_inputs
13031         to convert the inputs to the common type.
13032         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13033         to handle the optional cast of the ABS result.  Also allow a sign
13034         change or a sign extension between the ABS and MINUS.
13035         Use vect_widened_op_tree to detect the widened subtraction and use
13036         vect_convert_inputs to convert the inputs to the common type.
13037         (vect_handle_widen_op_by_const): Delete.
13038         (vect_recog_widen_op_pattern): New function.
13039         (vect_recog_widen_mult_pattern): Use it.
13040         (vect_recog_widen_shift_pattern): Likewise.
13041         (vect_recog_widen_sum_pattern): Use
13042         vect_look_through_possible_promotion to handle the promoted
13043         PLUS_EXPR operand.
13045 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13047         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13048         the containing gimple_seq *.
13049         * gimple-iterator.h (gsi_for_stmt): Declare it.
13050         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13051         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13052         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13053         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13054         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13055         checks.
13056         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13057         split out from...
13058         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
13059         statement being replaced is part of an existing pattern
13060         definition sequence, inserting the new pattern statements before
13061         the original one.
13062         (vect_pattern_recog_1): Don't return a bool.  If the statement
13063         is already part of a pattern, instead apply pattern matching
13064         to the pattern definition statements.  Don't clear the
13065         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13066         (vect_pattern_recog): Don't break after the first match;
13067         continue processing the pattern definition statements instead.
13068         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13070 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13072         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13073         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13074         (vect_recog_widen_sum_pattern): Use it.
13076 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13078         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13079         phi is not a pattern statement and has not been replaced by
13080         a pattern statement.
13081         * tree-vect-patterns.c (type_conversion_p): Don't check
13082         STMT_VINFO_IN_PATTERN_P.
13083         (vect_recog_vector_vector_shift_pattern): Likewise.
13084         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13085         the pattern statement rather than the original statement; check
13086         directly for a WIDEN_MULT_EXPR here.
13087         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13088         vect_is_simple_use to return the pattern statement rather
13089         than the original statement; use is_pattern_stmt_p to check
13090         for such a pattern statement.
13091         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13092         to return the pattern statement rather than the original statement;
13093         don't do the same transformation here.
13094         (vect_is_simple_use): If the defining statement has been replaced
13095         by a pattern statement, return the pattern statement instead.
13096         Remove the corresponding (local) transformation from the vectype
13097         overload.
13099 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13101         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13102         end and default to null.
13103         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13104         (vectorizable_reduction): Update calls accordingly, dropping the
13105         gimple ** argument if the passed-back statement isn't needed.
13106         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13107         (vect_recog_rotate_pattern): Likewise.
13108         (vect_recog_mask_conversion_pattern): Likewise.
13109         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13110         (vect_mask_constant_operand_p): Likewise.
13111         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13112         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13113         (get_group_load_store_type, get_load_store_type): Likewise.
13114         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13115         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13116         (vectorizable_conversion, vectorizable_assignment): Likewise.
13117         (vectorizable_shift, vectorizable_operation): Likewise.
13118         (vectorizable_store, vect_is_simple_cond): Likewise.
13119         (vectorizable_condition, vectorizable_comparison): Likewise.
13120         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13121         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13122         and move it to the end.  Cope with null def_stmt_outs.
13124 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13126         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13128 2018-06-29  Jeff Law  <law@redhat.com>
13130         * config/v850/v850.c (v850_legitimate_address_p): Handle large
13131         displacements for TARGET_V850E2V3 and newer.
13132         (TARGET_LRA_P): Remove.  Defaults to LRA now.
13133         * config/v850/v850.md (sign23byte_load): Remove.
13134         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13135         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13137 2018-06-29  Martin Liska  <mliska@suse.cz>
13139         PR lto/85759
13140         * coverage.c (coverage_init): Mangle full path name.
13141         * doc/invoke.texi: Document the change.
13142         * gcov-io.c (mangle_path): New.
13143         * gcov-io.h (mangle_path): Likewise.
13144         * gcov.c (mangle_name): Use mangle_path for path mangling.
13146 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13148         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13149         if starting source register is not even.
13151 2018-06-29  Martin Liska  <mliska@suse.cz>
13153         PR tree-optimization/86263
13154         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13155         Make edge redirection.
13157 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13159         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13160         (dump_scope_depth): New variable.
13161         (get_dump_scope_depth): New function.
13162         (dump_begin_scope): New function.
13163         (dump_end_scope): New function.
13164         * dumpfile.h (get_dump_scope_depth): New declaration.
13165         (dump_begin_scope): New declaration.
13166         (dump_end_scope): New declaration.
13167         (class auto_dump_scope): New class.
13168         (AUTO_DUMP_SCOPE): New macro.
13169         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13170         AUTO_DUMP_SCOPE.
13172 2018-06-29  Richard Biener  <rguenther@suse.de>
13174         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13175         compute_all_dependences succeeds.
13176         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13177         exceed --param loop-max-datarefs-for-datadeps.
13179 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
13181         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13183 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
13185         PR target/86348
13186         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13187         alternative 0 in preferred_for_speed attribute.
13189 2018-06-28  Paul Koning  <ni1d@arrl.net>
13191         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13192         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13193         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13194         * config/pdp11/pdp11.md: Correct "length" attribute calculation
13195         for shift insn patterns.
13197 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
13199         * cgraph.c (cgraph_node::get_body): Replace assignments to
13200         "dump_file" with calls to set_dump_file.
13201         * dumpfile.c (alt_dump_file): Make static, and group with...
13202         (alt_flags): ...this definition.
13203         (dumps_are_enabled): New variable.
13204         (refresh_dumps_are_enabled): New function.
13205         (set_dump_file): New function.
13206         (set_alt_dump_file): New function.
13207         (gcc::dump_manager::dump_start): Replace assignments to
13208         "dump_file" and "alt_dump_file" with calls to set_dump_file and
13209         set_alt_dump_file.
13210         (gcc::dump_manager::dump_finish): Likewise.
13211         * dumpfile.h (alt_dump_file): Delete decl.
13212         (dumps_are_enabled): New variable decl.
13213         (set_dump_file): New function decl.
13214         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13215         global.
13216         * tree-nested.c (lower_nested_functions): Replace assignments to
13217         "dump_file" with calls to set_dump_file.
13219 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13221         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13222         goto_locus of each outgoing edge of each basic block.
13224 2018-06-28  Richard Biener  <rguenther@suse.de>
13226         * dwarf2out.c (decl_scope_table): Remove.
13227         (push_decl_scope): Likewise.
13228         (pop_decl_scope): Likewise.
13229         (gen_type_die_for_member): Do not call push/pop_decl_scope.
13230         (gen_struct_or_union_type_die): Likewise.
13231         (gen_tagged_type_die): Likewise.
13232         (dwarf2out_init): Do not initialize decl_scope_table.
13233         (dwarf2out_c_finalize): Do not free it.
13235 2018-06-28  Richard Biener  <rguenther@suse.de>
13237         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13238         deciding whether to not re-use a DIE.
13240 2018-06-28  Richard Biener  <rguenther@suse.de>
13242         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13243         DW_AT_abstract_origin attribute.
13245 2018-06-28  Martin Liska  <mliska@suse.cz>
13247         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13248         Use newly introduced constants.
13249         * tree-switch-conversion.h (struct jump_table_cluster):
13250         Define max_ratio_for_size and max_ratio_for_speed.
13252 2018-06-28  Martin Liska  <mliska@suse.cz>
13254         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13255         Add new checking assert to catch invalid state.
13256         (jump_table_cluster::can_be_handled): Handle single case
13257         clusters.
13258         (jump_table_cluster::is_beneficial): Bail out for such case.
13259         (bit_test_cluster::find_bit_tests):
13260         Add new checking assert to catch invalid state.
13261         (bit_test_cluster::can_be_handled): Handle single case
13262         clusters.
13263         (bit_test_cluster::is_beneficial): Bail out for such case.
13264         (switch_decision_tree::analyze_switch_statement):
13265         Fix comment.
13267 2018-06-28  Martin Liska  <mliska@suse.cz>
13269         * common.opt: Introduce -completion option.
13270         * gcc.c (driver_handle_option): Handle it.
13271         (driver::main): Print completions if completion
13272         is set.
13273         * opt-suggestions.c (option_proposer::get_completions):
13274         New function.
13275         (option_proposer::suggest_completion): Likewise.
13276         (option_proposer::find_param_completions): Likewise.
13277         (verify_autocompletions): Likewise.
13278         (test_completion_valid_options): Likewise.
13279         (test_completion_valid_params): Likewise.
13280         (in_completion_p): Likewise.
13281         (empty_completion_p): Likewise.
13282         (test_completion_partial_match): Likewise.
13283         (test_completion_garbage): Likewise.
13284         (opt_proposer_c_tests): Likewise.
13285         * opt-suggestions.h: Declare new functions.
13286         * opts.c (common_handle_option): Handle OPT__completion_.
13287         * selftest-run-tests.c (selftest::run_tests): Add
13288         opt_proposer_c_tests.
13289         * selftest.c (assert_str_startswith): New.
13290         * selftest.h (assert_str_startswith): Likewise.
13291         (opt_proposer_c_tests): New.
13292         (ASSERT_STR_STARTSWITH): Likewise.
13294 2018-06-28  Martin Liska  <mliska@suse.cz>
13296         * Makefile.in: Add opt-suggestions.o.
13297         * gcc-main.c: Include opt-suggestions.h.
13298         * gcc.c (driver::driver): Likewise.
13299         (driver::~driver): Remove m_option_suggestions.
13300         (driver::build_option_suggestions): Moved to option_proposer.
13301         (driver::suggest_option): Likewise.
13302         (driver::handle_unrecognized_options): Use option_proposer.
13303         * gcc.h (class driver): Add new memver m_option_proposer.
13304         * opt-suggestions.c: New file.
13305         * opt-suggestions.h: New file.
13307 2018-06-28  Martin Liska  <mliska@suse.cz>
13309         * vec.h (class auto_string_vec): New (moved from auto_argvec).
13310         (auto_string_vec::~auto_string_vec): Likewise.
13312 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13314         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
13315         prevent_decl_creation_for_types fields up and add reset_location field.
13316         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
13317         statement if id->reset_location is true.
13318         (copy_edges_for_bb): Do not set goto_locus on the new edges if
13319         id->reset_location is true.
13320         (copy_phis_for_bb): Force input_location on the arguments if
13321         id->reset_location is true.
13322         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
13323         is set on the function to be inlined.
13325 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
13327         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
13329 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
13331         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
13332         registers for Pmode.
13333         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
13334         hard registers for the clobbered pseudo.
13336 2018-06-27  Paul Koning  <ni1d@arrl.net>
13338         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13339         mutually exclusive options.
13340         * config/pdp11/constraints.md (h): New constraint.
13341         (O): Update definition to match shift code generation.
13342         (D): New constraint.
13343         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
13344         (CCFP): Remove.
13345         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
13346         function.
13347         (output_jump): Change arguments.
13348         (pdp11_fixed_cc_regs): New function.
13349         (pdp11_cc_mode): Ditto.
13350         (pdp11_expand_shift): Ditto.
13351         (pdp11_assemble_shift): Ditto.
13352         (pdp11_small_shift): Ditto.
13353         (pdp11_branch_cost): Remove.
13354         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
13355         from output.
13356         (pdp11_register_move_cost): Update for CC registers.
13357         (pdp11_rtx_costs): Add case for LSHIFTRT.
13358         (pdp11_output_jump): Add CCNZ mode conditional branches.
13359         (notice_update_cc_on_set): Remove.
13360         (pdp11_cc_mode): New function.
13361         (simple_memory_operand): Correct pre/post decrement case.
13362         (no_side_effect_operand): New function.
13363         (pdp11_regno_reg_class): Add CC_REGS class.
13364         (pdp11_fixed_cc_regs): New function.
13365         (pdp11_small_shift): New function.
13366         (pdp11_expand_shift): New function to expand shift insns.
13367         (pdp11_assemble_shift): New function to output shifts.
13368         (pdp11_branch_cost): Remove.
13369         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
13370         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
13371         (WCHAR_TYPE): Ditto.
13372         (PTRDIFF_TYPE): Ditto.
13373         (ADJUST_INSN_LENGTH): New macro.
13374         (FIXED_REGISTERS): Add CC registers.
13375         (CALL_USED_REGISTERS): Ditto.
13376         (reg_class): Ditto.
13377         (REG_CLASS_NAMES): Ditto.
13378         (REG_CLASS_CONTENTS): Ditto.
13379         (SELECT_CC_MODE): Use new function.
13380         (TARGET_FLAGS_REGNUM): New macro.
13381         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
13382         (cc0_reg_rtx): Remove.
13383         (CC_STATUS_MDEP): Remove.
13384         (CC_STATUS_MDEFP_INIT): Remove.
13385         (CC_IN_FPU): Remove.
13386         (NOTICE_UPDATE_CC): Remove.
13387         (REGISTER_NAMES): Add CC registers.
13388         (BRANCH_COST): Change to constant 1.
13389         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
13390         handling.
13391         * config/pdp11/pdp11.opt (mbcopy): Remove.
13392         (mbcopy-builtin): Remove.
13393         (mbranch-cheap): Remove.
13394         (mbranch-expensive): Remove.
13395         * config/pdp11/predicates.md (expand_shift_operand): Update to
13396         match shift code generation.
13397         (ccnz_operator): New predicate.
13398         * doc/invoke.texi (PDP-11 Options): Remove deleted options
13399         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
13400         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
13401         exclusive options.
13402         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
13403         description of O constraint.
13405 2018-06-27  Jeff Law  <law@redhat.com>
13406             Austin Law  <austinklaw@gmail.com>
13408         * config/v850/v850.md (addsi3_set_flags): New pattern.
13409         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
13410         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
13411         (zero_extendhisi2_v850_set_flags): Likewise.
13412         (zero_extendqisi2_v850_set_flags): Likewise.
13413         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
13414         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
13415         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
13417         * config/v850/v850-protos.h (notice_update_cc): Remove.
13418         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
13419         (v850_print_operand): Handle 'D' and "d".
13420         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
13421         Add handling of arithmetic/logical operations compared against zero.
13422         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
13423         Do not look at v850_compare_op, instead get mode from last argument.
13424         (v850_gen_compare): Remove
13425         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
13426         after reload for prologue insns.
13427         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
13428         patterns.
13429         (construct_save_jarl): Likewise.
13430         (TARGET_FLAGS_REGNUM): Define.
13431         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
13432         (NOTICE_UPDATE_CC): Remove.
13433         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
13434         than cc0.  Conditionalize on reload_completed.
13435         (cmpsi_insn, setfcc_insn): Likewise.
13436         (tst1 splitter): Turn into define_and_split which sets the flags
13437         after reload.
13438         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
13439         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
13440         (cstoresf4, cstoredf4): Clobber the flags.
13441         (cmpsi, cmpsf, cmpdf): Remove expanders.
13442         (setf_insn): Remove pattern.
13443         (addsi3): Turn into define_and_split which clobbers the flags after
13444         reload and a suitable pattern (addsi3_clobber_flags) for use after
13445         reload.
13446         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
13447         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
13448         (ashrsi3, ashrsi3_v850e2): Likewise.
13449         (bins): Clobber the flags.
13450         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
13451         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
13452         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
13453         (call_value_internal_short, call_value_internal_long): Likewise.
13454         (callt_save_interrupt, callt_return_interrupt): Likewise.
13455         (save_interrupt, return_interrupt): Likewise.
13456         (callt_save_all_interrupt, save_all_interrupt): Likewise.
13457         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
13458         (restore_all_interrupt, _restore_all_interrupt): Likewise.
13459         (All FP comparisons): Only allow after reload has completed.
13460         (trfsr): Likewise.
13461         (divh, divhu): Tweak output template.
13462         (branch_z_normal, branch_z_invert): Remove
13463         (branch_nz_normal, branch_nz_invert): Likewise.
13464         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
13466         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
13467         * config/v850/v850.c (notice_update_cc): Remove.
13468         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
13469         (CC_NO_CARRY): Likewise.
13470         (NOTICE_UPDATE_CC): Define to nothing.
13471         * config/v850/v850.md: Remove block comment on cc0 handling
13472         Remove "cc" attribute from all patterns.  Remove cc_status handling
13473         from all patterns.  Minor formatting fixes.
13475 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13477         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
13478         (cortex-a76.cortex-a55): Likewise.
13479         * config/aarch64/aarch64-tune.md: Regenerate.
13480         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
13481         cortex-a76.cortex-a55.
13483 2018-06-27  Jeff Law  <law@redhat.com>
13485         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
13486         (MULTILIB_DIRNAMES): Similarly.
13488 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
13490         * gimple.h (gimple_return_retbnd): Delete.
13491         (gimple_return_set_retbnd): Likewise.
13492         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
13493         gimple_return_set_retbnd.
13494         * gimple-pretty-print.c (dump_gimple_return): Remove call to
13495         gimple_return_retbnd and adjust.
13496         * tree-inline.h (struct copy_body_data): Remove retbnd field.
13497         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
13498         Explicitly return NULL in a couple more cases.  Move assertion
13499         on debug statements and remove unreachable code.
13500         (reset_debug_binding): Do not test id->retbnd.
13501         (expand_call_inline): Do not set it.
13503 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
13505         * configure.ac: Add --disable-gcov option.
13506         * configure: Regenerate.
13507         * Makefile.in: Honour @enable_gcov@.
13508         * doc/install.texi: Document --disable-gcov.
13510 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13512         * config/arm/arm-cpus.in (cortex-a76): New entry.
13513         (cortex-a76.cortex-a55): Likewise.
13514         * config/arm/arm-tables.opt: Regenerate.
13515         * config/arm/arm-tune.md: Likewise.
13516         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
13517         * doc/invoke.texi (ARM Options): Document cortex-a76 and
13518         cortex-a76.cortex-a55.
13520 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
13522         PR target/85769
13523         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
13525 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
13527         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
13528         comment.
13529         (EPILOGUE_USES): Likewise.
13531 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
13533         * tree-inline.c (remap_location): New function extracted from...
13534         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
13535         (copy_phis_for_bb): ...here.  Call remap_location.
13536         (copy_cfg_body): Adjust call to copy_edges_for_bb.
13538 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
13540         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
13541         unaligned vsx for 16B memset.
13543 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
13545         PR target/86285
13546         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
13547         ieee128_float_type_node to long_double_type_node unless
13548         TARGET_LONG_DOUBLE_128 is set.
13550 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
13552         * cfgloop.c (get_loop_location): Convert return type from
13553         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
13554         by implicit construction from rtx_insn *, and using
13555         dump_user_location_t::from_function_decl for the fallback case.
13556         * cfgloop.h (get_loop_location): Convert return type from
13557         location_t to dump_user_location_t.
13558         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
13559         dump_printf_loc to pass in a dump_location_t rather than a
13560         location_t, via the gimple stmt.
13561         * coverage.c (get_coverage_counts): Update calls to
13562         dump_printf_loc to pass in dump_location_t rather than a
13563         location_t.
13564         * doc/optinfo.texi (Dump types): Convert example of
13565         dump_printf_loc from taking "locus" to taking "insn".  Update
13566         description of the "_loc" calls to cover dump_location_t.
13567         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
13568         "selftest.h".
13569         (dump_user_location_t::dump_user_location_t): New constructors,
13570         from gimple *stmt and rtx_insn *.
13571         (dump_user_location_t::from_function_decl): New function.
13572         (dump_loc): Make static.
13573         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
13574         const dump_location_t &.
13575         (dump_generic_expr_loc): Delete.
13576         (dump_printf_loc): Convert param "loc" from location_t to
13577         const dump_location_t &.
13578         (selftest::test_impl_location): New function.
13579         (selftest::dumpfile_c_tests): New function.
13580         * dumpfile.h: Include "profile-count.h".
13581         (class dump_user_location_t): New class.
13582         (struct dump_impl_location_t): New struct.
13583         (class dump_location_t): New class.
13584         (dump_printf_loc): Convert 2nd param from source_location to
13585         const dump_location_t &.
13586         (dump_generic_expr_loc): Delete.
13587         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
13588         const dump_location_t &.
13589         * gimple-fold.c (fold_gimple_assign): Update call to
13590         dump_printf_loc to pass in a dump_location_t rather than a
13591         location_t, via the gimple stmt.
13592         (gimple_fold_call): Likewise.
13593         * gimple-loop-interchange.cc
13594         (loop_cand::analyze_iloop_reduction_var): Update for change to
13595         check_reduction_path.
13596         (tree_loop_interchange::interchange): Update for change to
13597         find_loop_location.
13598         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
13599         change in return-type of find_loop_location.
13600         (graphite_regenerate_ast_isl): Likewise.
13601         * graphite-optimize-isl.c (optimize_isl): Likewise.
13602         * graphite.c (graphite_transform_loops): Likewise.
13603         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
13604         pass in a dump_location_t rather than a location_t, via the
13605         gimple stmt.
13606         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13607         * ipa.c (walk_polymorphic_call_targets): Likewise.
13608         * loop-unroll.c (report_unroll): Convert "locus" param from
13609         location_t to dump_location_t.
13610         (decide_unrolling): Update for change to get_loop_location's
13611         return type.
13612         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
13613         location_t to dump_user_location_t.
13614         (grid_find_single_omp_among_assignments_1): Updates calls to
13615         dump_printf_loc to pass in a dump_location_t rather than a
13616         location_t, via the gimple stmt.
13617         (grid_parallel_clauses_gridifiable): Convert "tloc" from
13618         location_t to dump_location_t.  Updates calls to dump_printf_loc
13619         to pass in a dump_location_t rather than a location_t, via the
13620         gimple stmt.
13621         (grid_inner_loop_gridifiable_p): Likewise.
13622         (grid_dist_follows_simple_pattern): Likewise.
13623         (grid_gfor_follows_tiling_pattern): Likewise.
13624         (grid_target_follows_gridifiable_pattern): Likewise.
13625         (grid_attempt_target_gridification): Convert initialization
13626         of local "grid" from memset to zero-initialization; FIXME: does
13627         this require C++11?  Update call to dump_printf_loc to pass in a
13628         optinfo_location rather than a location_t, via the gimple stmt.
13629         * profile.c (read_profile_edge_counts): Updates call to
13630         dump_printf_loc to pass in a dump_location_t rather than a
13631         location_t
13632         (compute_branch_probabilities): Likewise.
13633         * selftest-run-tests.c (selftest::run_tests): Call
13634         dumpfile_c_tests.
13635         * selftest.h (dumpfile_c_tests): New decl.
13636         * tree-loop-distribution.c (pass_loop_distribution::execute):
13637         Update for change in return type of find_loop_location.
13638         * tree-parloops.c (parallelize_loops): Likewise.
13639         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
13640         "locus" from location_t to dump_user_location_t.
13641         (canonicalize_loop_induction_variables): Likewise.
13642         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
13643         for change in return type of find_loop_location.
13644         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
13645         to dump_printf_loc to pass in a dump_location_t rather than a
13646         location_t, via the stmt.
13647         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
13648         Likewise.
13649         * tree-vect-loop-manip.c (find_loop_location): Convert return
13650         type from source_location to dump_user_location_t.
13651         (vect_do_peeling): Update for above change.
13652         (vect_loop_versioning): Update for change in type of
13653         vect_location.
13654         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
13655         from location_t to dump_user_location_t.
13656         (vect_estimate_min_profitable_iters): Update for change in type
13657         of vect_location.
13658         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
13659         location_t to dump_location_t.
13660         (vect_slp_bb): Update for change in type of vect_location.
13661         * tree-vectorizer.c (vect_location): Convert from source_location
13662         to dump_user_location_t.
13663         (try_vectorize_loop_1): Update for change in vect_location's type.
13664         (vectorize_loops): Likewise.
13665         (increase_alignment): Likewise.
13666         * tree-vectorizer.h (vect_location): Convert from source_location
13667         to dump_user_location_t.
13668         (find_loop_location): Convert return type from source_location to
13669         dump_user_location_t.
13670         (check_reduction_path): Convert 1st param from location_t to
13671         dump_user_location_t.
13672         * value-prof.c (check_counter): Update call to dump_printf_loc to
13673         pass in a dump_user_location_t rather than a location_t; update
13674         call to error_at for change in type of "locus".
13675         (check_ic_target): Update call to dump_printf_loc to
13676         pass in a dump_user_location_t rather than a location_t, via the
13677         call_stmt.
13679 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
13681         * config/s390/s390.h (enum processor_flags): Do not use
13682         default tune parameter when -march was specified.
13684 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
13686         PR target/86314
13687         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
13688         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
13689         operands.
13691 2018-06-26  Richard Biener  <rguenther@suse.de>
13693         PR tree-optimization/86287
13694         PR bootstrap/86316
13695         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
13696         (vect_analyze_loop): Initialize n_stmts.
13698 2018-06-26  Richard Biener  <rguenther@suse.de>
13700         PR middle-end/86271
13701         * fold-const.c (fold_convertible_p): Pointer extension
13702         isn't valid.
13704 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
13706         PR debug/86064
13707         * dwarf2out.c (loc_list_has_views): Adjust comments.
13708         (dw_loc_list): Split single cross-partition range with
13709         nonzero locview.
13711 2018-06-25  Jeff Law  <law@redhat.com>
13713         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
13714         on -mbig-switch by default.
13716         * config/v850/predicates.md (const_float_1_operand): Fix match_code
13717         test.
13718         (const_float_0_operand): Remove unused predicate.
13719         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
13720         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
13721         (recipsf2): New expander.  Original pattern now called
13722         (recipsf2_insn).
13723         (recipdf2, recipdf2_insn): Similarly.
13724         (rsqrtsf2, rsqrtsf2_insn): Similarly
13725         (rsqrtdf2, rsqrtdf2_insn): Similarly
13727 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
13729         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
13730         Simplify logic for FreeBSD (twice).
13732 2018-06-25  Martin Sebor  <msebor@redhat.com>
13734         PR tree-optimization/86204
13735         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
13736         a strnlen result if it's less than the length of the string.
13738 2018-06-25  Martin Sebor  <msebor@redhat.com>
13740         PR tree-optimization/85700
13741         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
13742         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
13743         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
13745 2018-06-25  Martin Sebor  <msebor@redhat.com>
13747         * doc/extend.texi (Zero-length arrays): Update and clarify.
13749 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
13751         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
13752         added IEEE/IBM long double multilib support on PowerPC little
13753         endian Linux systems.
13754         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
13755         (MULTILIB_DEFAULTS): Likewise.
13756         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13757         Likewise.
13758         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
13759         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
13760         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
13762 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
13764         PR middle-end/86311
13765         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
13766         (REORDER_45): Likewise.
13768 2018-06-25  Jeff Law  <law@redhat.com>
13770         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
13771         dividend to 32 bits.  Adjust length.
13772         (udivmodhi4): Cleanup output template.  Fix length.
13774 2018-06-25  Carl Love  <cel@us.ibm.com>
13776         * config/rs6000/vsx.md: Change word selector to prefered location.
13778 2018-06-25  Richard Biener  <rguenther@suse.de>
13780         PR tree-optimization/86304
13781         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
13782         epilogue-if-converted loops as well.
13784 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
13786         * lto-section-out.c (lto_begin_section): Do not print section
13787         name for noaddr and unnumbered dumps.
13789 2018-06-25  Richard Biener  <rguenther@suse.de>
13791         * tree-vectorizer.h (struct vec_info_shared): New structure
13792         with parts split out from struct vec_info and loop_nest from
13793         struct _loop_vec_info.
13794         (struct vec_info): Adjust accordingly.
13795         (struct _loop_vec_info): Likewise.
13796         (LOOP_VINFO_LOOP_NEST): Adjust.
13797         (LOOP_VINFO_DATAREFS): Likewise.
13798         (LOOP_VINFO_DDRS): Likewise.
13799         (struct _bb_vec_info): Likewise.
13800         (BB_VINFO_DATAREFS): Likewise.
13801         (BB_VINFO_DDRS): Likewise.
13802         (struct _stmt_vec_info): Add dr_aux member.
13803         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
13804         (DR_MISALIGNMENT_UNINITIALIZED): New.
13805         (set_dr_misalignment): Adjust.
13806         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
13807         (vect_analyze_loop): Adjust prototype.
13808         (vect_analyze_loop_form): Likewise.
13809         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13810         Compute dependences lazily.
13811         (vect_record_base_alignments): Use shared datarefs/ddrs.
13812         (vect_verify_datarefs_alignment): Likewise.
13813         (vect_analyze_data_refs_alignment): Likewise.
13814         (vect_analyze_data_ref_accesses): Likewise.
13815         (vect_analyze_data_refs): Likewise.
13816         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
13817         constructor parameter for shared part.
13818         (vect_analyze_loop_form): Pass in shared part and adjust.
13819         (vect_analyze_loop_2): Pass in storage for the number of
13820         stmts.  Move loop nest finding to the caller.  Compute
13821         datarefs lazily.
13822         (vect_analyze_loop): Pass in shared part.
13823         (vect_transform_loop): Verify shared datarefs are unchanged.
13824         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
13825         constructor parameter for shared part.
13826         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
13827         (vect_slp_bb): Verify shared datarefs are unchanged before
13828         transform.
13829         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
13830         change.
13831         (new_stmt_vec_info): Initialize DR_AUX misalignment to
13832         DR_MISALIGNMENT_UNINITIALIZED.
13833         * tree-vectorizer.c (vec_info::vec_info): Add constructor
13834         parameter for shared part.
13835         (vec_info::~vec_info): Adjust.
13836         (vec_info_shared::vec_info_shared): New.
13837         (vec_info_shared::~vec_info_shared): Likewise.
13838         (vec_info_shared::save_datarefs): Likewise.
13839         (vec_info_shared::check_datarefs): Likewise.
13840         (try_vectorize_loop_1): Construct shared part live for analyses
13841         of a single loop for multiple vector sizes.
13842         * tree-parloops.c (gather_scalar_reductions): Adjust.
13844 2018-06-25  Richard Biener  <rguenther@suse.de>
13846         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
13847         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
13848         (vect_analyze_data_refs): Remove similar code from here and
13849         simplify accordingly.
13851 2018-06-25  Richard Biener  <rguenther@suse.de>
13853         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
13854         for reverse storage order accesses rather than asserting
13855         they cannot happen here.
13857 2018-06-25  Tom de Vries  <tdevries@suse.de>
13859         PR debug/86257
13860         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
13861         Use data16 instead of .byte for insn prefix.
13863 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
13865         PR C++/86082
13866         * parser.c (make_char_string_pack): Pass this literal chars
13867         through cpp_interpret_string.
13868         (cp_parser_userdef_numeric_literal): Check the result of
13869         make_char_string_pack.
13871 2018-06-24  Maya Rashish  <coypu@sdf.org>
13873         * ginclude/stddef.h: Simplify conditions around avoiding
13874         re-definition of __size_t.
13876 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13878         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
13879         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
13881 2018-06-22  Maya Rashish  <coypu@sdf.org>
13883         * doc/invoke.texi (mno-fancy-math-387): Update for changes
13884         made to OpenBSD and NetBSD through the years.
13886 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13888         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
13889         behavior of vec_pack (vector double, vector double) to match
13890         behavior of vec_float2 (vector double, vector double).
13892 2018-06-22  Olivier Hainque  <hainque@adacore.com>
13894         * gimplify.c (gimplify_function_tree): Prevent creation
13895         of a trampoline for the address of the current function
13896         passed to entry/exit instrumentation hooks.
13898 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
13900         PR target/86222
13901         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
13902         correctly.
13904 2018-06-22  Martin Liska  <mliska@suse.cz>
13906         PR tree-optimization/86263
13907         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13908         Bail out if is_enabled is false.
13909         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
13910         New declaration.
13911         (jump_table_cluster::is_enabled): New function.
13913 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13915         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13916         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
13917         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
13918         (lto_input_ts_binfo_tree_pointers): Likewise.
13919         * tree-streamer-out.c (streamer_write_tree_bitfields,
13920         write_ts_binfo_tree_pointers): Likewise.
13921         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
13923 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13925         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
13927 2018-06-22  Martin Liska  <mliska@suse.cz>
13929         * symbol-summary.h (get): Make it pure and inline move
13930         functionality from ::get function.
13931         (get): Remove and inline into ::get and ::get_create.
13932         (get_create): Move code from ::get function.
13934 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13936         PR target/85994
13937         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
13938         -x assembler-with-cpp.
13940 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13942         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
13943         _FILE_OFFSET_BITS=64 for C++.
13945 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
13947         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
13948         conversion insn that shows up when pr85657-3.c is compiled using
13949         IEEE 128-bit long double.
13950         (neg<mode>2_internal): Use the correct mode to check whether the
13951         mode is IBM extended.
13952         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
13953         multiply and divide external functions from being created more
13954         than once.
13956 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13958         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
13959         functions.
13960         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
13961         the edge can be forwarded.
13962         (cfg_layout_merge_blocks): Likewise.
13964 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13966         * except.c (finish_eh_generation): Commit edge insertions only after
13967         the EH edges have been redirected from post-landing to landing pads.
13969 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13971         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
13972         create_tmp_var_for to create the FRAME decl.
13973         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
13975 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13977         * tree-inline.c (copy_edges_for_bb): Minor tweak.
13978         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
13979         debug statement when resetting its value.
13980         (expand_call_inline): Copy the locus of the call onto the assignment
13981         of the return value, if any.  Use local variable in more cases.
13983 2018-06-21  Martin Liska  <mliska@suse.cz>
13985         * ipa-pure-const.c (propagate_nothrow): Use
13986         funct_state_summaries->get.
13987         (dump_malloc_lattice): Likewise.
13988         (propagate_malloc): Likewise.
13990 2018-06-21  Richard Biener  <rguenther@suse.de>
13992         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
13993         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
13994         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
13995         comment.
13996         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
13997         BLOCK_ABSTRACT_ORIGIN unconditionally.
13999 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14001         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14002         deleting it.
14003         * ipa-reference.c (ipa_reference_c_finalize): Delete
14004         ipa_ref_opt_sum_summaries and set it to NULL.
14006 2018-06-21  Tom de Vries  <tdevries@suse.de>
14008         PR tree-optimization/85859
14009         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14010         test with comment from bb_no_side_effects_p.
14012 2018-06-21  Richard Biener  <rguenther@suse.de>
14014         PR tree-optimization/86232
14015         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14016         max for constant niter.
14018 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14020         * config/aarch64/aarch64-simd.md
14021         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14023 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14025         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14026         Make opernads of the unspec commutative.
14028 2018-06-21  Richard Biener  <rguenther@suse.de>
14030         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14031         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14032         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14033         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14034         NULL DR_STEP.
14035         (vect_record_base_alignments): Do not record base alignment
14036         for gathers or scatters.
14037         (vect_compute_data_ref_alignment): Drop return value that is always
14038         true.  Bail out early for gathers or scatters.
14039         (vect_enhance_data_refs_alignment): Bail out early for gathers
14040         or scatters.
14041         (vect_find_same_alignment_drs): Likewise.
14042         (vect_analyze_data_refs_alignment): Remove dead code.
14043         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14044         (vect_analyze_data_refs): For possible gathers or scatters do
14045         not create an alternate DR, just check their possible validity
14046         and mark them.  Adjust DECL_NONALIASED handling to not rely
14047         on DR_BASE_ADDRESS.
14048         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14049         update inits of gathers or scatters.
14050         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14051         Also copy gather/scatter flag to pattern vinfo.
14053 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14055         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14056         behavior of vec_packsu (vector unsigned long long, vector unsigned
14057         long long) to match behavior of vec_packs with same signature.
14059 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
14060             Thomas Schwinge <thomas@codesourcery.com>
14061             Cesar Philippidis  <cesar@codesourcery.com>
14063         * gimplify.c (gimplify_scan_omp_clauses): Add support for
14064         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14065         (gimplify_adjust_omp_clauses): Likewise.
14066         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14067         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14068         (gimplify_omp_target_update): Update handling of acc update and
14069         enter/exit data.
14070         * omp-low.c (install_var_field): Remove unused parameter
14071         base_pointers_restrict.
14072         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14073         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14074         FINALIZE}
14075         (omp_target_base_pointers_restrict_p): Delete.
14076         (scan_omp_target): Update call to scan_sharing_clauses.
14077         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14078         FINALIZE}.
14079         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14080         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14081         (convert_local_omp_clauses): Likewise.
14082         * tree-pretty-print.c (dump_omp_clause): Likewise.
14083         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
14084         FINALIZE}.
14085         (omp_clause_code_name): Likewise.
14087 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14089         PR debug/86194
14090         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14091         be narrowed.
14093         PR tree-optimization/86231
14094         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
14095         anti-range don't overwrite *vr0min before using it to compute *vr0max.
14097 2018-06-20  Tom de Vries  <tdevries@suse.de>
14099         PR tree-optimization/86097
14100         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14101         iv type if signedness of iv type is not the same as that of *nit.
14103 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
14105         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
14106         EDGE_EH edges, verify they are all EDGE_EH.
14108 2018-06-20  Maya Rashish  <coypu@sdf.org>
14110         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14112 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14114         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14115         * config/aarch64/aarch64.c (xgene1_tunings): Add
14116         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14117         (aarch64_mode_valid_for_sched_fusion_p):
14118         Allow 16-byte modes.
14119         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14120         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14121         128-bit modes.
14122         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14123         New pattern.
14124         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14125         * config/aarch64/iterators.md (VQ2): New mode iterator.
14127 2018-06-20  Martin Liska  <mliska@suse.cz>
14129         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14130         Change default ratio from 10 to 8.
14132 2018-06-20  Martin Liska  <mliska@suse.cz>
14134         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14135         New.
14136         (bit_test_cluster::find_bit_tests): Likewise.
14137         (switch_decision_tree::analyze_switch_statement): Find clusters.
14138         * tree-switch-conversion.h (struct jump_table_cluster): Document
14139         hierarchy.
14141 2018-06-20  Martin Liska  <mliska@suse.cz>
14143         * tree-switch-conversion.c (switch_conversion::collect):
14144         Record m_uniq property.
14145         (switch_conversion::expand): Bail out for special conditions.
14146         (group_cluster::~group_cluster): New.
14147         (group_cluster::group_cluster): Likewise.
14148         (group_cluster::dump): Likewise.
14149         (jump_table_cluster::emit): New.
14150         (switch_decision_tree::fix_phi_operands_for_edges): New.
14151         (struct case_node): Remove struct.
14152         (jump_table_cluster::can_be_handled): New.
14153         (case_values_threshold): Moved to header.
14154         (reset_out_edges_aux): Likewise.
14155         (jump_table_cluster::is_beneficial): New.
14156         (bit_test_cluster::can_be_handled): Likewise.
14157         (add_case_node): Remove.
14158         (bit_test_cluster::is_beneficial): New.
14159         (case_bit_test::cmp): New.
14160         (bit_test_cluster::emit): New.
14161         (expand_switch_as_decision_tree_p): Remove.
14162         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14163         (fix_phi_operands_for_edge): Likewise.
14164         (switch_decision_tree::analyze_switch_statement): New.
14165         (compute_cases_per_edge): Move ...
14166         (switch_decision_tree::compute_cases_per_edge): ... here.
14167         (try_switch_expansion): Likewise.
14168         (switch_decision_tree::try_switch_expansion): Likewise.
14169         (record_phi_operand_mapping): Likewise.
14170         (switch_decision_tree::record_phi_operand_mapping): Likewise.
14171         (emit_case_decision_tree): Likewise.
14172         (switch_decision_tree::emit): Likewise.
14173         (balance_case_nodes): Likewise.
14174         (switch_decision_tree::balance_case_nodes): Likewise.
14175         (dump_case_nodes): Likewise.
14176         (switch_decision_tree::dump_case_nodes): Likewise.
14177         (emit_jump): Likewise.
14178         (switch_decision_tree::emit_jump): Likewise.
14179         (emit_cmp_and_jump_insns): Likewise.
14180         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14181         (emit_case_nodes): Likewise.
14182         (switch_decision_tree::emit_case_nodes): Likewise.
14183         (conditional_probability): Remove.
14184         * tree-switch-conversion.h (enum cluster_type): New.
14185         (PRINT_CASE): New.
14186         (struct cluster): Likewise.
14187         (cluster::cluster): Likewise.
14188         (struct simple_cluster): Likewise.
14189         (simple_cluster::simple_cluster): Likewise.
14190         (struct group_cluster): Likewise.
14191         (struct jump_table_cluster): Likewise.
14192         (struct bit_test_cluster): Likewise.
14193         (struct min_cluster_item): Likewise.
14194         (struct case_tree_node): Likewise.
14195         (case_tree_node::case_tree_node): Likewise.
14196         (jump_table_cluster::case_values_threshold): Likewise.
14197         (struct case_bit_test): Likewise.
14198         (struct switch_decision_tree): Likewise.
14199         (struct switch_conversion): Likewise.
14200         (switch_decision_tree::reset_out_edges_aux): Likewise.
14202 2018-06-20  Martin Liska  <mliska@suse.cz>
14204         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14205         (hoist_edge_and_branch_if_true): Likewise.
14206         (expand_switch_using_bit_tests_p): Likewise.
14207         (struct case_bit_test): Likewise.
14208         (case_bit_test_cmp): Likewise.
14209         (emit_case_bit_tests): Likewise.
14210         (switch_conversion::switch_conversion): New class.
14211         (struct switch_conv_info): Remove old struct.
14212         (collect_switch_conv_info): More to ...
14213         (switch_conversion::collect): ... this.
14214         (check_range): Likewise.
14215         (switch_conversion::check_range): Likewise.
14216         (check_all_empty_except_final): Likewise.
14217         (switch_conversion::check_all_empty_except_final): Likewise.
14218         (check_final_bb): Likewise.
14219         (switch_conversion::check_final_bb): Likewise.
14220         (create_temp_arrays): Likewise.
14221         (switch_conversion::create_temp_arrays): Likewise.
14222         (free_temp_arrays): Likewise.
14223         (gather_default_values): Likewise.
14224         (switch_conversion::gather_default_values): Likewise.
14225         (build_constructors): Likewise.
14226         (switch_conversion::build_constructors): Likewise.
14227         (constructor_contains_same_values_p): Likewise.
14228         (switch_conversion::contains_same_values_p): Likewise.
14229         (array_value_type): Likewise.
14230         (switch_conversion::array_value_type): Likewise.
14231         (build_one_array): Likewise.
14232         (switch_conversion::build_one_array): Likewise.
14233         (build_arrays): Likewise.
14234         (switch_conversion::build_arrays): Likewise.
14235         (gen_def_assigns): Likewise.
14236         (switch_conversion::gen_def_assigns): Likewise.
14237         (prune_bbs): Likewise.
14238         (switch_conversion::prune_bbs): Likewise.
14239         (fix_phi_nodes): Likewise.
14240         (switch_conversion::fix_phi_nodes): Likewise.
14241         (gen_inbound_check): Likewise.
14242         (switch_conversion::gen_inbound_check): Likewise.
14243         (process_switch): Use the newly created class.
14244         (switch_conversion::expand): New.
14245         (switch_conversion::~switch_conversion): New.
14246         * tree-switch-conversion.h: New file.
14248 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14250         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14251         tree-vect-patterns.c.
14252         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14253         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
14254         (vect_recog_sad_pattern): Likewise.
14255         (vect_recog_widen_sum_pattern): Likewise.
14256         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
14257         (vect_recog_widen_shift_pattern): Remove the type_in argument.
14258         (vect_recog_rotate_pattern): Likewise.
14259         (vect_recog_mult_pattern): Likewise.
14260         (vect_recog_vector_vector_shift_pattern): Likewise.
14261         (vect_recog_divmod_pattern): Likewise.
14262         (vect_recog_mixed_size_cond_pattern): Likewise.
14263         (vect_recog_bool_pattern): Likewise.
14264         (vect_recog_mask_conversion_pattern): Likewise.
14265         (vect_try_gather_scatter_pattern): Likewise.
14266         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
14267         (vect_recog_over_widening_pattern): Likewise.
14268         (vect_recog_gather_scatter_pattern): Likewise.
14269         (vect_recog_func_ptr): Move from tree-vectorizer.h
14270         (vect_vect_recog_func_ptrs): Move further down the file.
14271         (vect_recog_func): Likewise.  Remove the third argument.
14272         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14273         (vect_pattern_recog_1): Expect the pattern function to do any
14274         necessary target tests.  Also expect it to provide a vector type.
14275         Remove the type_in handling.
14277 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14279         * tree-vect-patterns.c (vect_pattern_detected): New function.
14280         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14281         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14282         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14283         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14284         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14285         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14286         (vect_recog_mask_conversion_pattern)
14287         (vect_try_gather_scatter_pattern): Likewise.
14289 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14291         * tree-vect-patterns.c (vect_get_internal_def): New function.
14292         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14293         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14294         (search_type_for_mask_1): Use it.
14296 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14298         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14299         redundant WIDEN_SUM_EXPR handling.
14300         (vect_recog_sad_pattern): Likewise.
14302 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14304         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14305         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14306         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14307         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14308         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14310 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14312         * tree-vect-stmts.c (vectorizable_call): Make sure that we
14313         use the stmt_vec_info of the original bb statement for the
14314         new zero assignment, even if the call is part of a pattern.
14316 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14318         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
14319         that the sequence is attached to the original statement rather
14320         than the pattern statement.
14321         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
14322         PATTERN_DEF_SEQ from the original statement rather than
14323         the main pattern statement.
14324         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
14325         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
14326         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
14328 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14330         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
14331         definition statements before the early exit for statements that aren't
14332         live or relevant.
14333         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
14334         split out from...
14335         (vect_transform_loop): ...here.  Process pattern definition
14336         statements without first checking whether the main pattern
14337         statement is live or relevant.
14339 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14341         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
14342         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
14344 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
14346         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
14347         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
14348         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
14349         (expand_block_compare): Change select_block_compare_mode call.
14350         (expand_strncmp_align_check): Use new functions, fix comment.
14351         (emit_final_str_compare_gpr): New function.
14352         (expand_strn_compare): Refactor and clean up code.
14353         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
14355 2018-06-19  Tony Reix  <tony.reix@atos.com>
14356             Damien Bergamini  <damien.bergamini@atos.com>
14357             David Edelsohn  <dje.gcc@gmail.com>
14359         * collect2.c (static_obj): New variable.
14360         (static_libs): New variable.
14361         (is_in_list): Uncomment declaration.
14362         (main): Track AIX libraries linked statically.
14363         (is_in_list): Uncomment definition.
14364         (scan_prog_file): Don't add AIX shared libraries initializer
14365         to constructor list if linking statically.
14367 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
14369         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
14370         constant.
14371         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
14373 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
14375         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
14376         blocks.
14378 2018-06-19  Martin Liska  <mliska@suse.cz>
14380         * config/i386/i386.c (ix86_can_inline_p): Do not use
14381         ipa_fn_summaries::get_create.
14382         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
14383         get.
14384         (devirtualization_time_bonus): Likewise.
14385         (ipcp_propagate_stage): Likewise.
14386         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14387         (edge_set_predicate): Likewise.
14388         (evaluate_conditions_for_known_args): Likewise.
14389         (evaluate_properties_for_edge): Likewise.
14390         (ipa_call_summary::reset): Tranform to ...
14391         (ipa_call_summary::~ipa_call_summary): ... this.
14392         (ipa_fn_summary::reset): Transform to ...
14393         (ipa_fn_summary::~ipa_fn_summary): ... this.
14394         (ipa_fn_summary_t::remove): Rename to ...
14395         (ipa_fn_summary_t::remove_callees): ... this.
14396         (ipa_fn_summary_t::duplicate): Use placement new
14397         instead of memory copy.
14398         (ipa_call_summary_t::duplicate): Likewise.
14399         (ipa_call_summary_t::remove): Remove.
14400         (dump_ipa_call_summary): Change get_create to get.
14401         (ipa_dump_fn_summary): Dump only when summary exists.
14402         (analyze_function_body): Use symbol_summary::get instead
14403         of get_create.
14404         (compute_fn_summary): Likewise.
14405         (estimate_edge_devirt_benefit): Likewise.
14406         (estimate_edge_size_and_time): Likewise.
14407         (inline_update_callee_summaries): Likewise.
14408         (remap_edge_change_prob): Likewise.
14409         (remap_edge_summaries): Likewise.
14410         (ipa_merge_fn_summary_after_inlining): Likewise.
14411         (write_ipa_call_summary): Likewise.
14412         (ipa_fn_summary_write): Likewise.
14413         (ipa_free_fn_summary): Likewise.
14414         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
14415         (struct ipa_call_summary): Likewise.
14416         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
14417         of get_create.
14418         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
14419         (estimate_size_after_inlining): Likewise.
14420         (estimate_growth): Likewise.
14421         (growth_likely_positive): Likewise.
14422         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14423         (inline_call): Likewise.
14424         * ipa-inline.c (caller_growth_limits): Likewise.
14425         (can_inline_edge_p): Likewise.
14426         (can_inline_edge_by_limits_p): Likewise.
14427         (compute_uninlined_call_time): Likewise.
14428         (compute_inlined_call_time): Likewise.
14429         (want_inline_small_function_p): Likewise.
14430         (edge_badness): Likewise.
14431         (update_caller_keys): Likewise.
14432         (update_callee_keys): Likewise.
14433         (inline_small_functions): Likewise.
14434         (inline_to_all_callers_1): Likewise.
14435         (dump_overall_stats): Likewise.
14436         (early_inline_small_functions): Likewise.
14437         (early_inliner): Likewise.
14438         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14439         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14440         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14441         * ipa-split.c (execute_split_functions): Likewise.
14442         * symbol-summary.h: Likewise.
14443         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14445 2018-06-19  Richard Biener  <rguenther@suse.de>
14447         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
14448         (vectorize_loops): ... here.  Fix dbgcnt handling.
14449         (try_vectorize_loop): Wrap try_vectorize_loop_1.
14451 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
14453         PR target/86197
14454         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
14455         ieee128 argument takes up only one (vector) register, not two (floating
14456         point) registers.
14458 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14460         * gimplify.c (gimplify_init_constructor): Really never clear for an
14461         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
14463 2018-06-19  Richard Biener  <rguenther@suse.de>
14465         PR tree-optimization/86179
14466         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
14467         after failed recognition.
14469 2018-06-18  Martin Sebor  <msebor@redhat.com>
14471         PR middle-end/85602
14472         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
14473         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
14474         Handle integer subtraction.
14475         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
14476         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
14478 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14480         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
14481         param from rtx to rtx_insn *.
14482         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
14483         param.
14484         (frv_ifcvt_modify_insn): Likwise.
14485         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
14486         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
14487         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
14488         as_a <rtx_insn *> cast to local "unprotected_region" once
14489         it's been established that it's not NULL or pc_rtx.
14490         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
14491         param "sethi" from rtx to rtx_insn *.
14492         (nds32_group_float_insns): Likewise for param "insn".
14493         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
14494         param.
14495         (vax_output_int_subtract): Likewise.
14496         * config/vax/vax.c (vax_output_int_add): Likewise for param
14497         "insn".
14498         (vax_output_int_subtract): Likewise.
14499         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
14500         (emit_pattern_after): Likewise for param "after".
14501         (emit_insn_after): Likewise.
14502         (emit_jump_insn_after): Likewise.
14503         (emit_call_insn_after): Likewise.
14504         (emit_debug_insn_after): Likewise.
14505         (emit_pattern_before): Likewise for param "before".
14506         (emit_insn_before): Likewise.
14507         (emit_jump_insn_before): Likewise.
14508         * final.c (get_insn_template): Likewise for param "insn", removing
14509         a cast.
14510         * output.h (get_insn_template): Likewise for 2nd param.
14511         * rtl.h (emit_insn_before): Likewise.
14512         (emit_jump_insn_before): Likewise.
14513         (emit_debug_insn_before_noloc): Likewise.
14514         (emit_insn_after): Likewise.
14515         (emit_jump_insn_after): Likewise.
14516         (emit_call_insn_after): Likewise.
14517         (emit_debug_insn_after): Likewise.
14518         (set_insn_deleted): Likewise for param.
14520 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
14522         PR target/85358
14523         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
14524         floating point modes, so that IFmode is numerically greater than
14525         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
14526         to declare the ordering.  This prevents IFmode from being
14527         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
14528         machine.  Include rs6000-modes.h to share the fractional values
14529         between genmodes* and the rest of the compiler.
14530         (IFmode): Likewise.
14531         (KFmode): Likewise.
14532         (TFmode): Likewise.
14533         * config/rs6000/rs6000-modes.h: New file.
14534         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
14535         meaning of rs6000_long_double_size so that 126..128 selects an
14536         appropriate 128-bit floating point type.
14537         (rs6000_option_override_internal): Likewise.
14538         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
14539         (TARGET_LONG_DOUBLE_128): Change the meaning of
14540         rs6000_long_double_size so that 126..128 selects an appropriate
14541         128-bit floating point type.
14542         (LONG_DOUBLE_TYPE_SIZE): Update comment.
14543         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
14544         source and destination to match the standard usage.
14545         (truncifkf2): Likewise.
14546         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
14547         ISA 2.07 to use an explicit clobber, instead of passing in a
14548         temporary.
14549         (copysign<mode>3_soft): Likewise.
14551 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14553         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14554         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
14555         (vect_slp_analyze_instance_dependence): Likewise.
14556         (vect_enhance_data_refs_alignment): Likewise.
14557         (vect_analyze_data_refs_alignment): Likewise.
14558         (vect_slp_analyze_and_verify_instance_alignment
14559         (vect_analyze_data_ref_accesses): Likewise.
14560         (vect_prune_runtime_alias_test_list): Likewise.
14561         (vect_analyze_data_refs): Likewise.
14562         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
14563         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14564         (vect_analyze_scalar_cycles_1): Likewise.
14565         (vect_get_loop_niters): Likewise.
14566         (vect_analyze_loop_form_1): Likewise.
14567         (vect_update_vf_for_slp): Likewise.
14568         (vect_analyze_loop_operations): Likewise.
14569         (vect_analyze_loop): Likewise.
14570         (vectorizable_induction): Likewise.
14571         (vect_transform_loop): Likewise.
14572         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
14573         * tree-vect-slp.c (vect_analyze_slp): Likewise.
14574         (vect_make_slp_decision): Likewise.
14575         (vect_detect_hybrid_slp): Likewise.
14576         (vect_slp_analyze_operations): Likewise.
14577         (vect_slp_bb): Likewise.
14578         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
14579         (vectorizable_bswap): Likewise.
14580         (vectorizable_call): Likewise.
14581         (vectorizable_simd_clone_call): Likewise.
14582         (vectorizable_conversion): Likewise.
14583         (vectorizable_assignment): Likewise.
14584         (vectorizable_shift): Likewise.
14585         (vectorizable_operation): Likewise.
14586         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
14588 2018-06-18  Martin Sebor  <msebor@redhat.com>
14590         PR tree-optimization/81384
14591         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
14592         * builtins.c (expand_builtin_strnlen): New function.
14593         (expand_builtin): Call it.
14594         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
14595         * builtins.def (BUILT_IN_STRNLEN): New.
14596         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
14597         Warn for bounds in excess of maximum object size.
14598         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
14599         single-value ranges.  Handle strnlen.
14600         (handle_builtin_strlen): Handle strnlen.
14601         (strlen_check_and_optimize_stmt): Same.
14602         * doc/extend.texi (Other Builtins): Document strnlen.
14604 2018-06-18  Maya Rashish  <coypu@sdf.org>
14606         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
14607         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
14608         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
14610         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
14611         here to ...
14612         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
14614 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14616         * tree.c (escaped_string::escape): Replace cast to char * by
14617         const_cast<char *> (unescaped).
14619 2018-06-18  Nick Clifton  <nickc@redhat.com>
14621         PR 84195
14622         * tree.c (escaped_string): New class.  Converts an unescaped
14623         string into its escaped equivalent.
14624         (warn_deprecated_use): Use the new class to convert the
14625         deprecation message, if present.
14626         (test_escaped_strings): New self test.
14627         (test_c_tests): Add test_escaped_strings.
14628         * doc/extend.texi (deprecated): Add a note that the
14629         deprecation message is affected by the -fmessage-length
14630         option, and that control characters will be escaped.
14631         (#pragma GCC error): Document this pragma.
14632         (#pragma GCC warning): Likewise.
14633         * doc/invoke.texi (-fmessage-length): Document this option's
14634         effect on the #warning and #error preprocessor directives and
14635         the deprecated attribute.
14637 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
14639         * tree.c (decl_value_expr_lookup): Revert latest change.
14640         (decl_value_expr_insert): Likewise.
14642 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
14644         * gimplify.c (nonlocal_vlas): Delete.
14645         (nonlocal_vla_vars): Likewise.
14646         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
14647         referenced VLAs.
14648         (gimplify_body): Do not create and destroy nonlocal_vlas.
14649         * tree-nested.c: Include diagnostic.h.
14650         (use_pointer_in_frame): Tweak.
14651         (lookup_field_for_decl): Add assertion and declare the transformation.
14652         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
14653         internal error when the reference is in a wrong context.  Do not
14654         create a debug decl by default.
14655         (note_nonlocal_block_vlas): Delete.
14656         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
14657         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
14658         create a debug decl by default.
14659         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
14660         call is in a wrong context.
14661         (fixup_vla_decls): New function.
14662         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
14663         debug variables were created.
14664         * tree.c (decl_value_expr_lookup): Add checking assertion.
14665         (decl_value_expr_insert): Likewise.
14667 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14669         PR middle-end/82479
14670         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
14671         * tree-scalar-evolution.c (interpret_expr): Likewise.
14672         (expression_expensive_p): Likewise.
14673         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
14674         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
14675         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
14676         (ssa_defined_by_minus_one_stmt_p): New.
14678 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14680         PR middle-end/64946
14681         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
14682         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14683         * dojump.c (do_jump): Likewise.
14684         * expr.c (expand_expr_real_2): Check operand type's sign.
14685         * fold-const.c (const_unop): Handle ABSU_EXPR.
14686         (fold_abs_const): Likewise.
14687         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
14688         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
14689         (strip_sign_op_1): Likesise.
14690         * match.pd: Add new pattern to generate ABSU_EXPR.
14691         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
14692         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14693         * tree-eh.c (operation_could_trap_helper_p): Likewise.
14694         * tree-inline.c (estimate_operator_cost): Likewise.
14695         * tree-pretty-print.c (dump_generic_node): Likewise.
14696         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
14697         * tree.def (ABSU_EXPR): New.
14699 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
14701         PR middle-end/86095
14702         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
14703         documented as preserved for backward compatibility only.
14704         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
14706         PR rtl-optimization/86108
14707         * bb-reorder.c (create_forwarder_block): Renamed to ...
14708         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
14709         jump from new landing pad to the second part.
14710         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
14711         Adjust callers.
14713 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
14715         PR middle-end/85878
14716         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
14717         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
14718         Only call store_expr for halves if the mode is the same.
14720         PR middle-end/86123
14721         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
14722         Fix up comment formatting.
14724 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14726         * typed-splay-tree.h (typed_splay_tree::remove): New function.
14727         (typed_splay_tree::closure,
14728         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
14729         (typed_splay_tree::typed_splay_tree,
14730         typed_splay_tree::operator =): Declared private.
14731         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
14732         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
14733         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
14734         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
14735         typed_splay_tree::splay_tree_splay,
14736         typed_splay_tree::splay_tree_foreach_helper,
14737         typed_splay_tree::splay_tree_insert,
14738         typed_splay_tree::splay_tree_remove,
14739         typed_splay_tree::splay_tree_lookup,
14740         typed_splay_tree::splay_tree_predecessor,
14741         typed_splay_tree::splay_tree_successor,
14742         typed_splay_tree::splay_tree_min,
14743         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
14744         (typed_splay_tree::root, typed_splay_tree::comp,
14745         typed_splay_tree::delete_key,
14746         typed_splay_tree::delete_value): New data members.
14747         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
14748         typed_splay_tree::remove.
14750 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
14752         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
14753         -mginv and -mno-ginv to the assembler.
14754         * config/mips/mips.opt (-mcrc): New option.
14755         (-mginv): Likewise.
14756         * doc/invoke.text (-mcrc): Document.
14757         (-mginv): Likewise.
14759 2018-06-15  Nick Clifton  <nickc@redhat.com>
14761         PR 84195
14762         * tree.c (escaped_string): New class.  Converts an unescaped
14763         string into its escaped equivalent.
14764         (warn_deprecated_use): Use the new class to convert the
14765         deprecation message, if present.
14766         (test_escaped_strings): New self test.
14767         (test_c_tests): Add test_escaped_strings.
14768         * doc/extend.texi (deprecated): Add a note that the
14769         deprecation message is affected by the -fmessage-length
14770         option, and that control characters will be escaped.
14771         (#pragma GCC error): Document this pragma.
14772         (#pragma GCC warning): Likewise.
14773         * doc/invoke.texi (-fmessage-length): Document this option's
14774         effect on the #warning and #error preprocessor directives and
14775         the deprecated attribute.
14777 2018-06-15  Richard Biener  <rguenther@suse.de>
14779         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
14780         here, also noting vector size used.
14781         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
14782         size used in MSG_OPTIMIZED_LOCATIONS dump.
14783         (pass_slp_vectorize::execute): Adjust.
14785 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
14787         PR target/85968
14788         * config/arc/arc.c (arc_return_address_register): Fix
14789         if-condition.
14791 2018-06-15  Richard Biener  <rguenther@suse.de>
14793         PR middle-end/86159
14794         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
14795         leave useless conversion stripping to force_gimple_operand_gsi.
14796         (gimplify_build2): Likewise.
14797         (gimplify_build1): Likewise.
14799 2018-06-15  Richard Biener  <rguenther@suse.de>
14801         PR middle-end/86076
14802         * tree-cfg.c (move_stmt_op): unshare invariant addresses
14803         before adjusting their block.
14805 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14807         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
14808         multilibs for *-*-rtems*.
14809         * config/riscv/t-rtems: New file.
14811 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
14813         PR middle-end/86122
14814         * match.pd ((A +- CST1) +- CST2): Punt if last resort
14815         unsigned_type_for returns NULL.
14817         PR target/85945
14818         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
14819         subregs of multi-word pseudos unless the float mode has word size.
14821 2018-06-14  Richard Biener  <rguenther@suse.de>
14823         PR middle-end/86139
14824         * tree-vect-generic.c (build_word_mode_vector_type): Remove
14825         duplicate and harmful type_hash_canon.
14826         * tree.c (type_hash_canon): Assert we didn't find ourselves.
14828 2018-06-14  Richard Biener  <rguenther@suse.de>
14830         PR ipa/86124
14831         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
14832         NULL cgraph_node.
14834 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14836         * config/rtems.h (STDINT_LONG32): Define.
14838 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
14839             Prachi Godbole  <prachi.godbole@imgtec.com>
14841         * config/mips/mips-cpus.def: Define P6600.
14842         * config/mips/mips-tables.opt: Regenerate.
14843         * config/mips/mips.c (mips_ucbranch_type): New enum.
14844         (mips_rtx_cost_data): Add support for P6600.
14845         (mips_issue_rate): Likewise.
14846         (mips_multipass_dfa_lookahead): Likewise.
14847         (mips_avoid_hazard): Likewise.
14848         (mips_reorg_process_insns): Likewise.
14849         (mips_classify_branch_p6600): New function.
14850         * config/mips/mips.h (TUNE_P6600): New define.
14851         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
14852         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
14853         * config/mips/mips.md: Include p6600.md.
14854         (processor): Add p6600.
14855         * config/mips/p6600.md: New file.
14856         * doc/invoke.texi: Add p6600 to supported architectures.
14858 2018-06-13  Martin Sebor  <msebor@redhat.com>
14860         PR tree-optimization/86114
14861         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
14862         of integer types.
14863         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
14865 2018-06-13  Richard Biener  <rguenther@suse.de>
14867         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
14868         Properly set vector type of the intermediate stmt.
14869         * tree-vect-stmts.c (vectorizable_operation): The destination
14870         var always has vectype_out type.
14872 2018-06-13  Jeff Law  <law@redhat.com>
14874         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
14875         integer 0 for argument to print_rtl_with_bb.
14876         (rl78_reorg): Likewise.
14878 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
14880         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
14881         from rtx to rtx_insn *.
14882         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
14883         "label".
14884         (add_sched_insns_for_speculation): Likewise for local "target",
14885         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14886         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
14887         from rtx_insn ** to rtx_code_label **.
14888         (reorg_emit_nops): Likewise.
14889         (c6x_reorg): Likewise for local "call_labels".
14890         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
14891         rtx to rtx_insn *.
14892         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
14893         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
14894         the loops over LABEL_REFS.
14895         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
14896         braf_label.
14897         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14898         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
14899         (split_branches): Strengthen local "olabel" from rtx to
14900         rtx_insn *, adding a safe_as_a cast.
14901         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
14902         to "rtx_insn *".
14903         (add_insn_after): Likewise for first two params.
14904         (add_insn_before): Likewise.
14905         (remove_insn): Likewise for param.
14906         (emit_pattern_before_noloc): Likewise for second and third params.
14907         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
14908         (emit_call_insn_before_noloc): Likewise.
14909         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
14910         to "rtx_insn *".
14911         (emit_barrier_before): Likewise.
14912         (emit_label_before): Strengthen "label" param from "rtx" to
14913         "rtx_code_label *".  Strengthen "before" param from "rtx" to
14914         "rtx_insn *".
14915         (emit_insn_after_1): Strengthen "after" param from "rtx" to
14916         "rtx_insn *".
14917         (emit_pattern_after_noloc): Likewise.
14918         (emit_insn_after_noloc): Likewise.
14919         (emit_jump_insn_after_noloc): Likewise.
14920         (emit_call_insn_after_noloc): Likewise.
14921         (emit_debug_insn_after_noloc): Likewise.
14922         (emit_barrier_after): Likewise.
14923         (emit_label_after): Likewise for both params.
14924         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
14925         "loc" param from "int" to "location_t".
14926         (emit_insn_after_setloc): Likewise.
14927         (emit_jump_insn_after_setloc): Likewise.
14928         (emit_call_insn_after_setloc): Likewise.
14929         (emit_debug_insn_after_setloc): Likewise.
14930         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
14931         "loc" param from "int" to "location_t".
14932         (emit_pattern_before): Convert NULL_RTX to NULL.
14933         (emit_insn_before_setloc): Convert "loc" param from "int" to
14934         "location_t".
14935         (emit_jump_insn_before_setloc): Likewise.
14936         (emit_call_insn_before_setloc): Likewise.
14937         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
14938         rtx_insn *.  Convert "loc" param from "int" to "location_t".
14939         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
14940         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
14941         Convert 3rd param from "int" to "location_t".
14942         (emit_barrier_before, emit_barrier_after, next_real_insn):
14943         Strengthen param from rtx to rtx_insn *.
14944         (emit_label_before): Strengthen 1st param from "rtx" to
14945         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
14946         "rtx_insn *".
14947         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
14948         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
14949         Strengthen 2nd param from "rtx" to "rtx_insn *".
14950         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
14951         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
14952         Likewise. Convert 3rd param from "int" to "location_t".
14953         (emit_label_after): Strengthen 1st param from "rtx" to
14954         "rtx_code_label *".
14955         (next_real_insn, remove_insn): Strengthen param from "rtx" to
14956         "rtx_insn *".
14957         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
14958         from "rtx" to "rtx_insn *".
14960 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
14962         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
14963         bodies streamed in with -Q.
14964         * dumpfile.c (dump_files): Add lto-stream-out dump file.
14965         * dumpfile.h (tree_dump_index): Add lto_stream_out.
14966         * gimple-streamer-out.c: Include gimple-pretty-print.h
14967         (output_bb): Dump stmts streamed.
14968         * lto-section-out.c: Include print-tree.h
14969         (lto_begin_section): Dump sections created.
14970         (lto_output_decl_index): Dump decl encoded.
14971         * lto-streamer-out.c: Include print-tree.h
14972         (create_output_block): Dump output block created.
14973         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
14974         (output_function): Dump function output.
14975         (output_constructor): Dump constructor streamed.
14976         (write_global_stream): Output indexes encoded.
14977         (produce_asm_for_decls): Dump streams encoded.
14978         * lto-streamer.c (streamer_dump_file): New global var.
14979         * lto-streamer.h (streamer_dump_file): Declare.
14980         * passes.c (ipa_write_summaries): Initialize streamer dump.
14981         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
14982         in.
14984 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
14986         PR target/86048
14987         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
14988         offsets for register save directives.  Emit a second batch of save
14989         directives, if need be, when the function accesses prior frames.
14991 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14993         * config/arc/fpu.md (fmasf4): Force operand to register.
14994         (fnmasf4): Likewise.
14996 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14998         * config/arc/arc-protos.h (arc_pad_return): Remove.
14999         * config/arc/arc.c (machine_function): Remove force_short_suffix
15000         and size_reason.
15001         (arc_print_operand): Adjust printing of '&'.
15002         (arc_verify_short): Remove conditional printing of short suffix.
15003         (arc_final_prescan_insn): Remove reference to size_reason.
15004         (pad_return): New function.
15005         (arc_reorg): Call pad_return.
15006         (arc_pad_return): Remove.
15007         (arc_init_machine_status): Remove reference to force_short_suffix.
15008         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15009         (attr length): When attribute iscompact is true force to 2
15010         regardless; in the case of maybe check if we want to force the
15011         instruction to have 4 bytes length.
15012         (nopv): Change it to generate 4 byte long nop as well.
15013         (blockage): New pattern.
15014         (simple_return): Remove call to arc_pad_return.
15015         (p_return_i): Likewise.
15017 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15019         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15021 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15023         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15024         ARC cores.
15026 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15028         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15029         for ARC700 and ARCv2.
15031 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15033         PR target/86076
15034         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15035         operands[2] instead of operands[1].
15038 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15040         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15041         case, check whether the outer register overlaps an unallocatable
15042         register, not just whether it fits the required class.
15044 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15046         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15047         are poly_ints.
15048         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15049         (memrefs_conflict_p): Likewise.
15050         (init_alias_analysis): Likewise.
15051         * cfgexpand.c (expand_debug_expr): Likewise.
15052         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15053         * cse.c (fold_rtx): Likewise.
15054         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15055         * expr.c (emit_block_move_hints): Likewise.
15056         (clear_storage_hints, push_block, emit_push_insn): Likewise.
15057         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15058         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15059         (emit_group_store): Likewise.
15060         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
15061         to read the PRE/POST_MODIFY increment.
15062         * calls.c (store_one_arg): Use strip_offset.
15063         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15064         poly_int_rtx_p.
15065         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15066         by a VEC_SELECT.
15067         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15068         (simplify_binary_operation_1): Extend CONST_INT handling to
15069         poly_int_rtx_p.
15070         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15071         than a HOST_WIDE_INT.
15072         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15073         poly_int64.
15074         (adjust_mems, add_stores): Update accodingly.
15075         (vt_canonicalize_addr): Track polynomial offsets.
15076         (emit_note_insn_var_location): Likewise.
15077         (vt_add_function_parameter): Likewise.
15078         (vt_initialize): Likewise.
15080 2018-06-12  Jeff Law  <law@redhat.com>
15082         * config.gcc (alpha*-*-freebsd*): Remove.
15083         * config/alpha/freebsd.h: Remove.
15085 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
15087         PR other/69968
15088         * spellcheck-tree.c (levenshtein_distance): Rename to...
15089         (get_edit_distance): ...this, and update for underlying renaming.
15090         * spellcheck-tree.h (levenshtein_distance): Rename to...
15091         (get_edit_distance): ...this.
15092         * spellcheck.c (levenshtein_distance): Rename to...
15093         (get_edit_distance): ...this.  Convert from Levenshtein distance
15094         to Damerau-Levenshtein distance by supporting transpositions of
15095         adjacent characters.  Rename "v1" to "v_next" and "v0" to
15096         "v_one_ago".
15097         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15098         (selftest::test_edit_distance_unit_test_oneway): ...this, and
15099         update for underlying renaming.
15100         (selftest::levenshtein_distance_unit_test): Rename to...
15101         (selftest::test_get_edit_distance_unit): ...this, and update for
15102         underlying renaming.
15103         (selftest::test_find_closest_string): Add example from PR 69968
15104         where transposition helps
15105         (selftest::test_metric_conditions): Update for renaming.
15106         (selftest::test_metric_conditions): Likewise.
15107         (selftest::spellcheck_c_tests): Likewise.
15108         * spellcheck.h (levenshtein_distance): Rename both overloads to...
15109         (get_edit_distance): ...this.
15110         (best_match::consider): Update for renaming.
15112 2018-06-12  Martin Sebor  <msebor@redhat.com>
15114         PR tree-optimization/85259
15115         * builtins.c (compute_objsize): Handle constant offsets.
15116         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15117         true iff a warning has been issued.
15118         * gimple.h (gimple_nonartificial_location): New function.
15119         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15120         gimple_nonartificial_location and handle -Wno-system-headers.
15121         (handle_builtin_stxncpy): Same.
15123 2018-06-12  Martin Sebor  <msebor@redhat.com>
15125         PR c/85931
15126         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15128 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
15130         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15131         BUILTIN_VEC_XST entries for pointer to double and long long.
15133 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
15135         PR target/85990
15136         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15137         Update comments.
15138         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15139         Likewise.
15141 2018-06-12  Martin Liska  <mliska@suse.cz>
15143         * doc/options.texi: Document IntegerRange.
15145 2018-06-12  Martin Liska  <mliska@suse.cz>
15147         * config/i386/i386.opt: Make MPX-related options as Deprecated.
15148         * opt-functions.awk: Handle Deprecated flag.
15149         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15150         and report error.
15151         (read_cmdline_option): Report warning for a deprecated option.
15152         * opts.h (struct cl_option): Add new field cl_deprecated.
15153         (CL_ERR_DEPRECATED): New.
15155 2018-06-12  Martin Liska  <mliska@suse.cz>
15157         * doc/options.texi: Document Deprecated option flag.
15159 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15161         * config/arc/arc-arch.h (arc_extras): New enum.
15162         (arc_cpu_t):Add field extra.
15163         (arc_cpu_types): Consider the extras.
15164         * config/arc/arc-cpus.def: Add extras info.
15165         * config/arc/arc-opts.h (processor_type): Consider extra field.
15166         * config/arc/arc.c (arc_override_options): Handle extra field.
15168 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15170         * config/arc/arc-arch.h: Update ARC_OPTX macro.
15171         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15172         field.
15173         * config/arc/arc.c (arc_init): Update pic warning.
15174         (irq_range): Update irq range parsing warnings.
15175         (arc_override_options): Update various warning messages.
15176         (arc_handle_aux_attribute): Likewise.
15178 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
15180         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15182 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15184         * doc/sourcebuild.texi: Document usage of line number 0 in verify
15185         compiler messages directives.
15187 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
15189         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15190         * config/mips/mips-tables.opt: Regenerate.
15191         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15192         mips64r6.
15193         * doc/invoke.texi: Document -march=i6500.
15195 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
15197         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15198         (i6400_gpmul): Add cpu_unit.
15199         (i6400_gpdiv): Likewise.
15200         (i6400_msa_add_d): Update reservations.
15201         (i6400_msa_int_add) Likewise.
15202         (i6400_msa_short_logic3) Likewise.
15203         (i6400_msa_short_logic2) Likewise.
15204         (i6400_msa_short_logic) Likewise.
15205         (i6400_msa_move) Likewise.
15206         (i6400_msa_cmp) Likewise.
15207         (i6400_msa_short_float2) Likewise.
15208         (i6400_msa_div_d) Likewise.
15209         (i6400_msa_long_logic1) Likewise.
15210         (i6400_msa_long_logic2) Likewise.
15211         (i6400_msa_mult) Likewise.
15212         (i6400_msa_long_float2) Likewise.
15213         (i6400_msa_long_float4) Likewise.
15214         (i6400_msa_long_float5) Likewise.
15215         (i6400_msa_long_float8) Likewise.
15216         (i6400_fpu_fadd): Include frint type.
15217         (i6400_fpu_store): New define_insn_reservation.
15218         (i6400_fpu_load): Likewise.
15219         (i6400_fpu_move): Likewise.
15220         (i6400_fpu_fcmp): Likewise.
15221         (i6400_fpu_fmadd): Likewise.
15222         (i6400_int_mult): Include imul3nc type and update reservation.
15223         (i6400_int_div): Include idiv3 type and update reservation.
15224         (i6400_int_load): Update to check type not move_type.
15225         (i6400_int_store): Likewise.
15226         (i6400_int_prefetch): Set zero latency.
15228 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
15230         * gcc.c: Document new %@{...} sequence.
15231         (LINK_COMMAND_SPEC): Use it for the -L switches.
15232         (cpp_unique_options): Use it for the -I switches.
15233         (at_file_argbuf): New global variable.
15234         (in_at_file): Likewise.
15235         (alloc_args): Create at_file_argbuf.
15236         (clear_args): Truncate at_file_argbuf.
15237         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15238         (open_at_file): New function.
15239         (close_at_file): Likewise.
15240         (create_at_file): Delete.
15241         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15242         <'o'>: Likewise.
15243         <'@'>: New case.
15244         (validate_switches_from_spec): Deal with %@{...} sequence.
15245         (validate_switches): Likewise.
15246         (driver::finalize): Call clear_args.
15248 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
15250         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15252 2018-06-11  Martin Sebor  <msebor@redhat.com>
15254         * doc/invoke.texi (-Wall): List -Wc++17-compat.
15255         (Wno-class-memaccess): Add @opindex.
15256         (Wno-templates, Wno-multiple-inheritance): Same.
15257         (Wno-virtual-inheritance, Wno-namespaces): Same.
15258         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15259         (Wno-format-overflow, Wno-format-truncation): Same.
15260         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15261         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15262         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15263         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15264         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15265         (Wno-misspelled-isr): Same.
15267 2018-06-11  Martin Sebor  <msebor@redhat.com>
15269         * PR tree-optimization/86083
15270         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15272 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
15274         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15276 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
15278         PR target/85755
15279         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15280         on the correct operand.
15281         (*movdi_internal64): Ditto.
15283 2018-06-11  Martin Liska  <mliska@suse.cz>
15285         PR tree-optimization/86089
15286         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15288 2018-06-11  Julia Koval  <julia.koval@intel.com>
15290         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15291         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15292         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15294 2018-06-11  Olivier Hainque  <hainque@adacore.com>
15296         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15297         for Ada with strict dwarf2.
15299 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
15301         PR target/85755
15302         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15303         addresses.
15305 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
15307         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15309 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
15311         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15312         TARGET_ELF.
15314 2018-06-08  Martin Liska  <mliska@suse.cz>
15316         * tree-cfg.h (debug_function): Fix argument type to match
15317         implementation.
15319 2018-06-08  Martin Liska  <mliska@suse.cz>
15321         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
15322         Remove usage of MPX-related (and removed) fields.
15323         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
15325 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
15327         * cfg.c (debug): Use TDF_NONE rather than 0.
15328         * cfghooks.c (debug): Likewise.
15329         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
15330         (struct dump_option_value_info): Convert to...
15331         (struct kv_pair): ...this template type.
15332         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
15333         rather than 0.
15334         (optinfo_verbosity_options): Likewise.
15335         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
15336         OPTGROUP_NONE.
15337         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
15338         than int for "optgroup_flags" param.
15339         (dump_generic_expr_loc): Use dump_flags_t rather than int for
15340         "dump_kind" param.
15341         (dump_dec): Likewise.
15342         (dump_finish): Use TDF_NONE rather than 0.
15343         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
15344         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
15345         than 0.  Update for change to option_ptr.
15346         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
15347         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
15348         0.  Update for changes to optinfo_verbosity_options and
15349         optgroup_options.
15350         (opt_info_switch_p): Convert optgroup_flags from int to
15351         optgroup_flags_t.
15352         (dump_basic_block): Use dump_flags_t rather than int
15353         for "dump_kind" param.
15354         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
15355         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
15356         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
15357         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
15358         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
15359         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
15360         TDF_NONE): Convert from macros to...
15361         (enum dump_flag): ...this new enum.
15362         (dump_flags_t): Update to use enum.
15363         (operator|, operator&, operator~, operator|=, operator&=):
15364         Implement for dump_flags_t.
15365         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
15366         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
15367         Convert from macros to...
15368         (enum optgroup_flag): ...this new enum.
15369         (optgroup_flags_t): New typedef.
15370         (operator|, operator|=): Implement for optgroup_flags_t.
15371         (struct dump_file_info): Convert field "alt_flags" to
15372         dump_flags_t.  Convert field "optgroup_flags" to
15373         optgroup_flags_t.
15374         (dump_basic_block): Use dump_flags_t rather than int for param.
15375         (dump_generic_expr_loc): Likewise.
15376         (dump_dec): Likewise.
15377         (dump_register): Convert param "optgroup_flags" to
15378         optgroup_flags_t.
15379         (opt_info_enable_passes): Likewise.
15380         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
15381         than 0.
15382         * gimple-pretty-print.c (debug): Likewise.
15383         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
15384         (merged_store_group::apply_stores): Likewise.
15385         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
15386         * gimple.c (verify_gimple_pp): Likewise.
15387         * graphite-poly.c (print_pbb_body): Likewise.
15388         * passes.c (pass_manager::register_one_dump_file): Convert
15389         local "optgroup_flags" to optgroup_flags_t.
15390         * print-tree.c (print_node): Use TDF_NONE rather than 0.
15391         (debug): Likewise.
15392         (debug_body): Likewise.
15393         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
15394         to optgroup_flags_t.
15395         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
15396         than 0.
15397         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
15398         (convert_mult_to_fma): Likewise.
15399         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
15400         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
15401         * tree-vect-data-refs.c (dump_lower_bound): Convert param
15402         "dump_kind" to dump_flags_t.
15404 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15406         * config/rs6000/rs6000.c (min, max): Delete.
15408 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15410         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
15411         -mabi=spe and -mabi=no-spe.
15413 2018-06-08  Martin Liska  <mliska@suse.cz>
15415         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
15416         where we expect an existing summary.
15418 2018-06-08  Martin Liska  <mliska@suse.cz>
15420         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
15421         * ipa-inline.h (estimate_edge_growth): Likewise.
15423 2018-06-08  Martin Liska  <mliska@suse.cz>
15425         * cgraph.c (function_version_hasher::hash): Use
15426         cgraph_node::get_uid ().
15427         (function_version_hasher::equal):
15428         * cgraph.h (cgraph_node::get_uid): New method.
15429         * ipa-inline.c (update_caller_keys): Use
15430         cgraph_node::get_uid ().
15431         (update_callee_keys): Likewise.
15432         * ipa-utils.c (searchc): Likewise.
15433         (ipa_reduced_postorder): Likewise.
15434         * lto-cgraph.c (input_node): Likewise.
15435         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
15436         * symbol-summary.h (symtab_insertion): Likewise.
15437         (symtab_removal): Likewise.
15438         (symtab_duplication): Likewise.
15439         * tree-pretty-print.c (dump_function_header): Likewise.
15440         * tree-sra.c (convert_callers_for_node): Likewise.
15442 2018-06-08  Martin Liska  <mliska@suse.cz>
15444         * cgraph.c (symbol_table::create_edge): Always assign a new
15445         unique number.
15446         (symbol_table::free_edge): Do not recycle numbers.
15447         * cgraph.h (cgraph_edge::get): New method.
15448         * symbol-summary.h (symtab_removal): Use it.
15449         (symtab_duplication): Likewise.
15450         (call_summary::hashable_uid): Remove.
15452 2018-06-08  Martin Liska  <mliska@suse.cz>
15454         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
15455         (initialize_growth_caches): Remove.
15456         (free_growth_caches): Likewise.
15457         (do_estimate_edge_time): Use edge_growth_cache.
15458         (do_estimate_edge_size): Likewise.
15459         (do_estimate_edge_hints): Likewise.
15460         * ipa-inline.c (reset_edge_caches): Likewise.
15461         (recursive_inlining): Likewise.
15462         (inline_small_functions): Likewise.
15463         * ipa-inline.h (initialize_growth_caches): Remove.
15464         (estimate_edge_size): Likewise.
15465         (estimate_edge_time): Likewise.
15466         (estimate_edge_hints): Likewise.
15467         (reset_edge_growth_cache): Likewise.
15468         * symbol-summary.h (call_summary::remove): New method.
15470 2018-06-08  Martin Liska  <mliska@suse.cz>
15472         * ipa-cp.c (class edge_clone_summary): New summary.
15473         (grow_edge_clone_vectors): Remove.
15474         (ipcp_edge_duplication_hook): Remove.
15475         (class edge_clone_summary_t): New call_summary class.
15476         (ipcp_edge_removal_hook): Remove.
15477         (edge_clone_summary_t::duplicate): New function.
15478         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
15479         (create_specialized_node): Likewise.
15480         (ipcp_driver): Initialize edge_clone_summaries and do not
15481         register hooks.
15483 2018-06-08  Martin Liska  <mliska@suse.cz>
15485         * symbol-summary.h (get): New function.
15486         (call_summary::m_initialize_when_cloning): New class member.
15488 2018-06-08  Martin Liska  <mliska@suse.cz>
15490         * cgraph.c (cgraph_node::remove): Do not recycle uid.
15491         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
15492         (symbol_table::allocate_cgraph_symbol): Do not set uid.
15493         * passes.c (uid_hash_t): Record removed_nodes by their uids.
15494         (remove_cgraph_node_from_order): Use the removed_nodes set.
15495         (do_per_function_toporder): Likwise.
15496         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
15497         instead of summary_uid.
15498         (symtab_removal): Likewise.
15499         (symtab_duplication): Likewise.
15501 2018-06-08  Martin Liska  <mliska@suse.cz>
15503         * ipa-cp.c (ipcp_store_bits_results): Use
15504         ipcp_transformation_sum.
15505         (ipcp_store_vr_results): Likewise.
15506         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
15507         to ...
15508         (ipcp_transformation_initialize): ... this.
15509         (ipa_set_node_agg_value_chain):
15510         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
15511         (write_ipcp_transformation_info): Likewise.
15512         (read_ipcp_transformation_info): Likewise.
15513         (ipcp_update_bits): Likewise.
15514         (ipcp_update_vr): Likewise.
15515         (ipcp_transform_function): Likewise.
15516         * ipa-prop.h: Rename ipcp_transformation_summary to
15517         ipcp_transformation.
15518         (class ipcp_transformation_t): New function summary.
15519         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
15520         (ipa_get_agg_replacements_for_node): Likewise.
15522 2018-06-08  Martin Liska  <mliska@suse.cz>
15524         * ipa-pure-const.c (struct funct_state_d): Do it class instead
15525         of struct.
15526         (class funct_state_summary_t): New function_summary class.
15527         (has_function_state): Remove.
15528         (get_function_state): Likewise.
15529         (set_function_state): Likewise.
15530         (add_new_function): Likewise.
15531         (funct_state_summary_t::insert): New function.
15532         (duplicate_node_data): Remove.
15533         (remove_node_data): Remove.
15534         (funct_state_summary_t::duplicate): New function.
15535         (register_hooks): Create new funct_state_summaries.
15536         (pure_const_generate_summary): Use it.
15537         (pure_const_write_summary): Likewise.
15538         (pure_const_read_summary): Likewise.
15539         (propagate_pure_const): Likewise.
15540         (propagate_nothrow): Likewise.
15541         (dump_malloc_lattice): Likewise.
15542         (propagate_malloc): Likewise.
15543         (execute): Do not register hooks, just remove summary
15544         instead.
15545         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
15546         constructor.
15548 2018-06-08  Martin Liska  <mliska@suse.cz>
15550         * ipa-reference.c (remove_node_data): Remove.
15551         (duplicate_node_data): Likewise.
15552         (class ipa_ref_var_info_summary_t): New class.
15553         (class ipa_ref_opt_summary_t): Likewise.
15554         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
15555         (get_reference_optimization_summary): Use
15556         ipa_ref_opt_sum_summaries.
15557         (set_reference_vars_info): Remove.
15558         (set_reference_optimization_summary): Likewise.
15559         (ipa_init): Create summaries.
15560         (init_function_info): Use function summary.
15561         (ipa_ref_opt_summary_t::duplicate): New function.
15562         (ipa_ref_opt_summary_t::remove): New function.
15563         (get_read_write_all_from_node): Fix GNU coding style.
15564         (propagate): Use function summary.
15565         (write_node_summary_p): Fix GNU coding style.
15566         (stream_out_bitmap): Likewise.
15567         (ipa_reference_read_optimization_summary): Use function summary.
15568         (ipa_reference_c_finalize): Do not release hooks.
15570 2018-06-08  Martin Liska  <mliska@suse.cz>
15572         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
15573         (analyze_function_body): Extract multiple calls of get_create.
15574         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15575         * ipa-inline.c (recursive_inlining): Use ::get method.
15576         * ipa-inline.h (estimate_edge_growth): Likewise.
15578 2018-06-08  Martin Liska  <mliska@suse.cz>
15580         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
15581         HSA_INVALID.
15582         (hsa_function_summary::hsa_function_summary): Use the new enum
15583         value.
15584         (hsa_gpu_implementation_p): Use hsa_summaries::get.
15585         * hsa-gen.c (hsa_get_host_function): Likewise.
15586         (get_brig_function_name): Likewise.
15587         * ipa-hsa.c (process_hsa_functions): Likewise.
15588         (ipa_hsa_write_summary): Likewise.
15589         * symbol-summary.h (symtab_duplication): Use ::get function/
15590         (get): New function.
15592 2018-06-08  Martin Liska  <mliska@suse.cz>
15594         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
15595         of get.
15596         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
15597         (hsa_register_kernel): Likewise.
15598         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
15599         * hsa-gen.c (hsa_get_host_function): Likewise.
15600         (get_brig_function_name): Likewise.
15601         (generate_hsa): Likewise.
15602         (pass_gen_hsail::execute): Likewise.
15603         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
15604         (devirtualization_time_bonus): Likewise.
15605         (ipcp_propagate_stage): Likewise.
15606         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15607         (edge_set_predicate): Likewise.
15608         (evaluate_conditions_for_known_args): Likewise.
15609         (evaluate_properties_for_edge): Likewise.
15610         (ipa_fn_summary::reset): Likewise.
15611         (ipa_fn_summary_t::duplicate): Likewise.
15612         (dump_ipa_call_summary): Likewise.
15613         (ipa_dump_fn_summary): Likewise.
15614         (analyze_function_body): Likewise.
15615         (compute_fn_summary): Likewise.
15616         (estimate_edge_devirt_benefit): Likewise.
15617         (estimate_edge_size_and_time): Likewise.
15618         (estimate_calls_size_and_time): Likewise.
15619         (estimate_node_size_and_time): Likewise.
15620         (inline_update_callee_summaries): Likewise.
15621         (remap_edge_change_prob): Likewise.
15622         (remap_edge_summaries): Likewise.
15623         (ipa_merge_fn_summary_after_inlining): Likewise.
15624         (ipa_update_overall_fn_summary): Likewise.
15625         (read_ipa_call_summary): Likewise.
15626         (inline_read_section): Likewise.
15627         (write_ipa_call_summary): Likewise.
15628         (ipa_fn_summary_write): Likewise.
15629         (ipa_free_fn_summary): Likewise.
15630         * ipa-hsa.c (process_hsa_functions): Likewise.
15631         (ipa_hsa_write_summary): Likewise.
15632         (ipa_hsa_read_section): Likewise.
15633         * ipa-icf.c (sem_function::merge): Likewise.
15634         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15635         (do_estimate_edge_time): Likewise.
15636         (estimate_size_after_inlining): Likewise.
15637         (estimate_growth): Likewise.
15638         (growth_likely_positive): Likewise.
15639         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15640         (inline_call): Likewise.
15641         * ipa-inline.c (caller_growth_limits): Likewise.
15642         (can_inline_edge_p): Likewise.
15643         (can_inline_edge_by_limits_p): Likewise.
15644         (compute_uninlined_call_time): Likewise.
15645         (compute_inlined_call_time): Likewise.
15646         (want_inline_small_function_p): Likewise.
15647         (edge_badness): Likewise.
15648         (update_caller_keys): Likewise.
15649         (update_callee_keys): Likewise.
15650         (recursive_inlining): Likewise.
15651         (inline_small_functions): Likewise.
15652         (inline_to_all_callers_1): Likewise.
15653         (dump_overall_stats): Likewise.
15654         (early_inline_small_functions): Likewise.
15655         (early_inliner): Likewise.
15656         * ipa-inline.h (estimate_edge_growth): Likewise.
15657         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15658         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15659         * ipa-prop.h (IPA_NODE_REF): Likewise.
15660         (IPA_EDGE_REF): Likewise.
15661         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15662         (propagate_malloc): Likewise.
15663         * ipa-split.c (execute_split_functions): Likewise.
15664         * symbol-summary.h: Rename get to get_create.
15665         (get): Likewise.
15666         (get_create): Likewise.
15667         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15669 2018-06-08  Martin Liska  <mliska@suse.cz>
15671         * symbol-summary.h (release): Move definition out of class
15672         declaration.
15673         (symtab_removal): Likewise.
15674         (symtab_duplication): Likewise.
15676 2018-06-08  Martin Liska  <mliska@suse.cz>
15678         * symbol-summary.h (function_summary): Move constructor
15679         implementation out of class declaration.
15680         (release): Likewise.
15681         (symtab_insertion): Likewise.
15682         (symtab_removal): Likewise.
15683         (symtab_duplication): Likewise.
15684         (get): Likewise.
15686 2018-06-08  Martin Liska  <mliska@suse.cz>
15688         * Makefile.in: Remove support for MPX (macros, related functions,
15689         fields in cgraph_node, ...).
15690         * builtin-types.def (BT_BND): Likewise.
15691         (BT_FN_BND_CONST_PTR): Likewise.
15692         (BT_FN_CONST_PTR_BND): Likewise.
15693         (BT_FN_VOID_PTR_BND): Likewise.
15694         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
15695         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
15696         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
15697         (expand_builtin_mempcpy_with_bounds): Likewise.
15698         (expand_builtin_memset_with_bounds): Likewise.
15699         (expand_builtin_memset_args): Likewise.
15700         (std_expand_builtin_va_start): Likewise.
15701         (expand_builtin): Likewise.
15702         (expand_builtin_with_bounds): Likewise.
15703         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
15704         (DEF_LIB_BUILTIN_CHKP): Likewise.
15705         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
15706         (DEF_CHKP_BUILTIN): Likewise.
15707         (BUILT_IN_MEMCPY): Likewise.
15708         (BUILT_IN_MEMMOVE): Likewise.
15709         (BUILT_IN_MEMPCPY): Likewise.
15710         (BUILT_IN_MEMSET): Likewise.
15711         (BUILT_IN_STPCPY): Likewise.
15712         (BUILT_IN_STRCAT): Likewise.
15713         (BUILT_IN_STRCHR): Likewise.
15714         (BUILT_IN_STRCPY): Likewise.
15715         (BUILT_IN_STRLEN): Likewise.
15716         (BUILT_IN_MEMCPY_CHK): Likewise.
15717         (BUILT_IN_MEMMOVE_CHK): Likewise.
15718         (BUILT_IN_MEMPCPY_CHK): Likewise.
15719         (BUILT_IN_MEMSET_CHK): Likewise.
15720         (BUILT_IN_STPCPY_CHK): Likewise.
15721         (BUILT_IN_STRCAT_CHK): Likewise.
15722         (BUILT_IN_STRCPY_CHK): Likewise.
15723         * calls.c (store_bounds): Likewise.
15724         (emit_call_1): Likewise.
15725         (special_function_p): Likewise.
15726         (maybe_warn_nonstring_arg): Likewise.
15727         (initialize_argument_information): Likewise.
15728         (finalize_must_preallocate): Likewise.
15729         (compute_argument_addresses): Likewise.
15730         (expand_call): Likewise.
15731         * cfgexpand.c (expand_call_stmt): Likewise.
15732         (expand_return): Likewise.
15733         (expand_gimple_stmt_1): Likewise.
15734         (pass_expand::execute): Likewise.
15735         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
15736         (cgraph_node::remove): Likewise.
15737         (cgraph_node::dump): Likewise.
15738         (cgraph_node::verify_node): Likewise.
15739         * cgraph.h (chkp_function_instrumented_p): Likewise.
15740         (symtab_node::get_alias_target): Likewise.
15741         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
15742         (cgraph_local_p): Likewise.
15743         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
15744         (cgraph_edge::rebuild_references): Likewise.
15745         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
15746         (walk_polymorphic_call_targets): Likewise.
15747         (cgraph_node::expand_thunk): Likewise.
15748         (symbol_table::output_weakrefs): Likewise.
15749         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
15750         (ix86_handle_option): Likewise.
15751         * config/i386/constraints.md: Likewise.
15752         * config/i386/i386-builtin-types.def (BND): Likewise.
15753         (VOID): Likewise.
15754         (PVOID): Likewise.
15755         (ULONG): Likewise.
15756         * config/i386/i386-builtin.def (BDESC_END): Likewise.
15757         (BDESC_FIRST): Likewise.
15758         (BDESC): Likewise.
15759         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
15760         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
15761         * config/i386/i386.c (enum reg_class): Likewise.
15762         (ix86_target_string): Likewise.
15763         (ix86_option_override_internal): Likewise.
15764         (ix86_conditional_register_usage): Likewise.
15765         (ix86_valid_target_attribute_inner_p): Likewise.
15766         (ix86_set_indirect_branch_type): Likewise.
15767         (ix86_set_current_function): Likewise.
15768         (ix86_function_arg_regno_p): Likewise.
15769         (init_cumulative_args): Likewise.
15770         (ix86_function_arg_advance): Likewise.
15771         (ix86_function_arg): Likewise.
15772         (ix86_pass_by_reference): Likewise.
15773         (ix86_function_value_regno_p): Likewise.
15774         (ix86_function_value_1): Likewise.
15775         (ix86_function_value_bounds): Likewise.
15776         (ix86_return_in_memory): Likewise.
15777         (ix86_setup_incoming_vararg_bounds): Likewise.
15778         (ix86_va_start): Likewise.
15779         (indirect_thunk_need_prefix): Likewise.
15780         (print_reg): Likewise.
15781         (ix86_print_operand): Likewise.
15782         (ix86_expand_call): Likewise.
15783         (ix86_output_function_return): Likewise.
15784         (reg_encoded_number): Likewise.
15785         (BDESC_VERIFYS): Likewise.
15786         (ix86_init_mpx_builtins): Likewise.
15787         (ix86_init_builtins): Likewise.
15788         (ix86_emit_cmove): Likewise.
15789         (ix86_emit_move_max): Likewise.
15790         (ix86_expand_builtin): Likewise.
15791         (ix86_builtin_mpx_function): Likewise.
15792         (ix86_get_arg_address_for_bt): Likewise.
15793         (ix86_load_bounds): Likewise.
15794         (ix86_store_bounds): Likewise.
15795         (ix86_load_returned_bounds): Likewise.
15796         (ix86_store_returned_bounds): Likewise.
15797         (ix86_class_likely_spilled_p): Likewise.
15798         (ix86_hard_regno_mode_ok): Likewise.
15799         (x86_order_regs_for_local_alloc): Likewise.
15800         (ix86_mitigate_rop): Likewise.
15801         (ix86_bnd_prefixed_insn_p): Likewise.
15802         (ix86_mpx_bound_mode): Likewise.
15803         (ix86_make_bounds_constant): Likewise.
15804         (ix86_initialize_bounds): Likewise.
15805         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
15806         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
15807         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
15808         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
15809         (TARGET_CHKP_BOUND_MODE): Likewise.
15810         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
15811         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
15812         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
15813         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
15814         * config/i386/i386.h (TARGET_MPX): Likewise.
15815         (TARGET_MPX_P): Likewise.
15816         (VALID_BND_REG_MODE): Likewise.
15817         (FIRST_BND_REG): Likewise.
15818         (LAST_BND_REG): Likewise.
15819         (enum reg_class): Likewise.
15820         (BND_REG_P): Likewise.
15821         (BND_REGNO_P): Likewise.
15822         (BNDmode): Likewise.
15823         (ADJUST_INSN_LENGTH): Likewise.
15824         * config/i386/i386.md: Likewise.
15825         * config/i386/i386.opt: Likewise.
15826         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
15827         (defined): Likewise.
15828         (LINK_MPX): Likewise.
15829         (MPX_SPEC): Likewise.
15830         (LIBMPX_SPEC): Likewise.
15831         (LIBMPXWRAPPERS_SPEC): Likewise.
15832         (CHKP_SPEC): Likewise.
15833         * config/i386/predicates.md: Likewise.
15834         * dbxout.c (dbxout_type): Likewise.
15835         * doc/extend.texi: Likewise.
15836         * doc/invoke.texi: Likewise.
15837         * doc/md.texi: Likewise.
15838         * doc/tm.texi: Likewise.
15839         * doc/tm.texi.in: Likewise.
15840         * dwarf2out.c (is_base_type): Likewise.
15841         (gen_formal_types_die): Likewise.
15842         (gen_subprogram_die): Likewise.
15843         (gen_type_die_with_usage): Likewise.
15844         (gen_decl_die): Likewise.
15845         (dwarf2out_late_global_decl): Likewise.
15846         * expr.c (expand_assignment): Likewise.
15847         (emit_storent_insn): Likewise.
15848         (store_expr_with_bounds): Likewise.
15849         (store_expr): Likewise.
15850         (expand_expr_real_1): Likewise.
15851         * expr.h (store_expr_with_bounds): Likewise.
15852         * function.c (use_register_for_decl): Likewise.
15853         (struct bounds_parm_data): Likewise.
15854         (assign_parms_augmented_arg_list): Likewise.
15855         (assign_parm_find_entry_rtl): Likewise.
15856         (assign_parm_is_stack_parm): Likewise.
15857         (assign_parm_load_bounds): Likewise.
15858         (assign_bounds): Likewise.
15859         (assign_parms): Likewise.
15860         (expand_function_start): Likewise.
15861         * gcc.c (CHKP_SPEC): Likewise.
15862         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
15863         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
15864         (wrestrict_dom_walker::check_call): Likewise.
15865         * gimple.c (gimple_build_call_from_tree): Likewise.
15866         * gimple.h (enum gf_mask): Likewise.
15867         (gimple_call_with_bounds_p): Likewise.
15868         (gimple_call_set_with_bounds): Likewise.
15869         * gimplify.c (gimplify_init_constructor): Likewise.
15870         * ipa-cp.c (initialize_node_lattices): Likewise.
15871         (propagate_constants_across_call): Likewise.
15872         (find_more_scalar_values_for_callers_subset): Likewise.
15873         * ipa-hsa.c (process_hsa_functions): Likewise.
15874         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
15875         * ipa-icf.c (sem_function::merge): Likewise.
15876         * ipa-inline.c (early_inliner): Likewise.
15877         * ipa-pure-const.c (warn_function_noreturn): Likewise.
15878         (warn_function_cold): Likewise.
15879         (propagate_pure_const): Likewise.
15880         * ipa-ref.h (enum GTY): Likewise.
15881         * ipa-split.c (find_retbnd): Likewise.
15882         (consider_split): Likewise.
15883         (split_function): Likewise.
15884         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
15885         * ipa.c (walk_polymorphic_call_targets): Likewise.
15886         (symbol_table::remove_unreachable_nodes): Likewise.
15887         (process_references): Likewise.
15888         (cgraph_build_static_cdtor_1): Likewise.
15889         * lto-cgraph.c (lto_output_node): Likewise.
15890         (output_refs): Likewise.
15891         (compute_ltrans_boundary): Likewise.
15892         (input_overwrite_node): Likewise.
15893         (input_node): Likewise.
15894         (input_cgraph_1): Likewise.
15895         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
15896         * passes.c (pass_manager::execute_early_local_passes): Likewise.
15897         (class pass_chkp_instrumentation_passes): Likewise.
15898         (make_pass_chkp_instrumentation_passes): Likewise.
15899         * passes.def: Likewise.
15900         * rtl.h (struct GTY): Likewise.
15901         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
15902         * stor-layout.c (layout_type): Likewise.
15903         * symtab.c: Likewise.
15904         * target.def: Likewise.
15905         * targhooks.c (default_chkp_bound_type): Likewise.
15906         (default_chkp_bound_mode): Likewise.
15907         (default_builtin_chkp_function): Likewise.
15908         (default_chkp_function_value_bounds): Likewise.
15909         (default_chkp_make_bounds_constant): Likewise.
15910         (default_chkp_initialize_bounds): Likewise.
15911         * targhooks.h (default_chkp_bound_type): Likewise.
15912         (default_chkp_bound_mode): Likewise.
15913         (default_builtin_chkp_function): Likewise.
15914         (default_chkp_function_value_bounds): Likewise.
15915         (default_chkp_make_bounds_constant): Likewise.
15916         (default_chkp_initialize_bounds): Likewise.
15917         * toplev.c (compile_file): Likewise.
15918         (process_options): Likewise.
15919         * tree-core.h (DEF_BUILTIN): Likewise.
15920         (DEF_BUILTIN_CHKP): Likewise.
15921         * tree-inline.c (declare_return_variable): Likewise.
15922         (remap_gimple_stmt): Likewise.
15923         (copy_bb): Likewise.
15924         (initialize_inlined_parameters): Likewise.
15925         (expand_call_inline): Likewise.
15926         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
15927         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
15928         (make_pass_ipa_chkp_produce_thunks): Likewise.
15929         (make_pass_chkp): Likewise.
15930         (make_pass_chkp_opt): Likewise.
15931         (make_pass_chkp_instrumentation_passes): Likewise.
15932         * tree-pretty-print.c (dump_generic_node): Likewise.
15933         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
15934         * tree-ssa-dce.c (propagate_necessity): Likewise.
15935         (eliminate_unnecessary_stmts): Likewise.
15936         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
15937         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
15938         * tree-ssa-sccvn.h: Likewise.
15939         * tree-ssa-strlen.c (get_string_length): Likewise.
15940         (valid_builtin_call): Likewise.
15941         (adjust_last_stmt): Likewise.
15942         (handle_builtin_strchr): Likewise.
15943         (handle_builtin_strcpy): Likewise.
15944         (handle_builtin_stxncpy): Likewise.
15945         (handle_builtin_memcpy): Likewise.
15946         (handle_builtin_strcat): Likewise.
15947         (strlen_check_and_optimize_stmt): Likewise.
15948         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
15949         * tree-streamer-in.c: Likewise.
15950         * tree-streamer.c (record_common_node): Likewise.
15951         * tree.c (tree_code_size): Likewise.
15952         (wide_int_to_tree_1): Likewise.
15953         (type_contains_placeholder_1): Likewise.
15954         (build_common_tree_nodes): Likewise.
15955         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
15956         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
15957         (POINTER_BOUNDS_P): Likewise.
15958         (BOUNDED_TYPE_P): Likewise.
15959         (BOUNDED_P): Likewise.
15960         (CALL_WITH_BOUNDS_P): Likewise.
15961         (pointer_bounds_type_node): Likewise.
15962         * value-prof.c (gimple_ic): Likewise.
15963         * var-tracking.c (vt_add_function_parameters): Likewise.
15964         * varasm.c (make_decl_rtl): Likewise.
15965         (assemble_start_function): Likewise.
15966         (output_constant): Likewise.
15967         (maybe_assemble_visibility): Likewise.
15968         * varpool.c (ctor_for_folding): Likewise.
15969         * chkp-builtins.def: Remove.
15970         * ipa-chkp.c: Remove.
15971         * ipa-chkp.h: Remove.
15972         * rtl-chkp.c: Remove.
15973         * rtl-chkp.h: Remove.
15974         * tree-chkp-opt.c: Remove.
15975         * tree-chkp.c: Remove.
15976         * tree-chkp.h: Remove.
15978 2018-06-07  Carl Love  <cel@us.ibm.com>
15980         * config/rs6000/vsx.md (vextract_fp_from_shorth,
15981         vextract_fp_from_shortl): Add BE support.
15983 2018-06-07  Paul Koning  <ni1d@arrl.net>
15985         * compare-elim.c (try_merge_compare): Don't merge compare if
15986         address contains a side effect.
15987         (try_eliminate_compare): Likewise.
15989 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
15991         * config.gcc: Support "tremont".
15992         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
15993         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15994         PROCESSOR_TREMONT.
15995         * config/i386/i386.c (m_TREMONT): Define.
15996         (processor_target_table): Add "tremont".
15997         (PTA_TREMONT): Define.
15998         (ix86_lea_outperforms): Add TARGET_TREMONT.
15999         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16000         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16001         and M_INTEL_GOLDMONT_PLUS.
16002         (fold_builtin_cpu): Add "tremont".
16003         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16004         (ix86_option_override_internal): Add "tremont".
16005         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16006         (processor_type): Add PROCESSOR_TREMONT.
16007         * config/i386/x86-tune.def: Add m_TREMONT.
16008         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16010 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16012         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16013         symbol defined for msp430i* devices to be lower case.
16015 2018-06-07  Richard Biener  <rguenther@suse.de>
16017         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16018         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16019         Properly wrap signed arithmetic if overflow wraps.
16021 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16023         PR tree-optimization/69615
16024         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16025         of a cast from a same precision integral SSA_NAME in a bb dominated
16026         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16027         cast to utype if rhs2 has already a compatible type.
16029 2018-06-07  Richard Biener  <rguenther@suse.de>
16031         PR tree-optimization/85935
16032         * graphite-scop-detection.c (find_params_in_bb): Analyze
16033         condition operands with respect to the correct loop.  Assert
16034         the analysis doesn't fail.
16036 2018-06-04  Carl Love  <cel@us.ibm.com>
16038         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16039         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16040         as it is slightly cheaper.
16041         (first_match_or_eos_index_<mode>):
16042         Calculate index using natural element order.
16043         (first_match_index_<mode>):
16044         Calculate index using natural element order.
16045         (first_match_or_eos_index_<mode>):
16046         Calculate index using natural order.
16047         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16048         for BE and LE modes.
16049         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16050         P9V_BUILTIN_VCLZLSBB_V16QI.
16051         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16052         specific.
16054 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16056         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16057         indentation and line wrap for many prototypes.  Add missing
16058         @smallexample directives around block of prototypes for vec_xl and
16059         vec_xst.
16061 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
16063         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16064         track if we pass or return IEEE 128-bit floating point.
16065         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16066         C++ mangling that is compatible with GCC 8.1.
16067         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16068         (init_cumulative_args): Note if we pass or return IEEE 128-bit
16069         floating point types.
16070         (rs6000_function_arg_advance_1): Likewise.
16071         (rs6000_mangle_type): Optionally generate mangled names that match
16072         what GCC 8.1 generated for IEEE 128-bit floating point types.
16073         (rs6000_globalize_decl_name): If we have an external function that
16074         passes or returns IEEE 128-bit types, generate a weak reference
16075         from the mangled name used in GCC 8.1 to the current mangled
16076         name.
16077         (rs6000_init_builtins): Make __ibm128 use the long double type if
16078         long double is IBM extended double.  Make __float128 use the long
16079         double type if long double is IEEE 128-bit.
16081         PR target/85657
16082         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16083         macro for __ibm128 built-in functions.
16084         (PACK_IF): Add __ibm128 pack/unpack functions.
16085         (UNPACK_IF): Likewise.
16086         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16087         enable long double built-in functions if long double is IEEE
16088         128-bit floating point.
16089         (rs6000_invalid_builtin): Update long double built-in function
16090         error message.
16091         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16092         functions, adjust the built-in function to use the long double
16093         built-in function if __ibm128 and long double are the same type.
16094         * doc/extend.texi (PowerPC builtins): Update documention for
16095         __builtin_{,un}pack_longdouble.  Add documentation for
16096         __builtin_{,un}pack_ibm128.
16098 2018-06-06  Jim Wilson  <jimw@sifive.com>
16100         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16101         (struct machine_function): New field interrupt_mode.
16102         (riscv_handle_type_attribute): New function.  Add forward declaration.
16103         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16104         (riscv_expand_epilogue): Check interrupt_mode field.
16105         (riscv_set_current_function): Check interrupt attribute args and
16106         set interrupt_mode field.
16107         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16108         (riscv_sret, riscv_uret): New.
16109         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16110         new arguments to interrupt attribute.
16112 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
16114         PR target/63177
16115         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16116         Don't handle -mcpu=power8 if -mpower9-vector is also used.
16118 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16120         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16121         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16122         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16123         several redundant entries.
16125 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
16127         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16128         type from "rtx" to "rtx_insn *".
16129         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16130         for local "call_insn", removing cast.
16131         (ix86_expand_call): Likewise, introducing a "call_insn" local.
16133 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
16135         PR tree-optimization/86066
16136         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16137         for BIT_INSERT_EXPR stores.
16139 2018-06-06  Richard Biener  <rguenther@suse.de>
16141         PR tree-optimization/86062
16142         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16143         component refs ontop
16144         of to be offsetted base.
16146 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16148         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16149         to be static and remove check on interrupt attribute name.
16151 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16153         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16154         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16156 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16158         PR target/79924
16159         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16160         second argument.
16161         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16162         Remove second argument, change how error is called.
16163         (aarch64_layout_arg): Remove second argument from
16164         aarch64_err_no_fpadvsimd call.
16165         (aarch64_init_cumulative_args): Ditto.
16166         (aarch64_gimplify_va_arg_expr): Ditto.
16167         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16169 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
16171         * config/i386/i386.md (simple_return_indirect_internal): New expander.
16172         (*simple_return_indirect_internal<mode>): Rename from
16173         simple_return_indirect_internal.  Use W mode iterator.
16174         (rstorssp): New expander.
16175         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
16176         (clrssbsy): New expander.
16177         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
16179 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16181         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16182         __typeof__.
16183         (cmse_check_pointed_object): Likewise.
16185 2018-06-05  Martin Liska  <mliska@suse.cz>
16187         PR gcov-profile/47618
16188         * doc/invoke.texi: Document how -fprofile-dir format
16189         is extended.
16191 2018-06-05  Richard Biener  <rguenther@suse.de>
16193         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16194         removal pretend DOM info isn't available so we do not update
16195         it and only remove edges, not dominated blocks.  Actually free
16196         DOM info in case we removed something.  Remove unreachable blocks.
16197         (mfb_keep_latches): Work with either DOM info or marked backedges.
16198         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16199         first.  Mark backedges if DOM info isn't available.
16200         (Re-)compute DOM info after cleanup_control_flow_pre.
16202 2018-06-05  Richard Biener  <rguenther@suse.de>
16204         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16205         (locus_discrim_hasher::hash): Adjust.
16206         (locus_discrim_hasher::equal): Likewise.
16207         (next_discriminator_for_locus): Work on line directly.
16208         (same_line_p): Pass in expanded locus1 as well.
16209         (assign_discriminators): Avoid redundant location expansions.
16211 2018-06-05  Richard Biener  <rguenther@suse.de>
16213         PR tree-optimization/86046
16214         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16215         if required after clearing TREE_ADDRESSABLE.
16217 2018-06-05  Richard Biener  <rguenther@suse.de>
16219         PR tree-optimization/86047
16220         * tree-ssa-loop.c (for_each_index): Glob handling of all
16221         decls and constants and really handle all of them.
16223 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16225         PR target/81497
16226         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16227         qualifier_void_pointer and qualifier_const_void_pointer.
16228         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16229         (arm_init_builtins): Handle the above.
16230         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16231         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16232         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16233         void intrinsics.
16235 2018-06-05  Martin Liska  <mliska@suse.cz>
16237         * auto-profile.c (read_autofdo_file): Do not use
16238         gcov_ctr_summary struct.
16239         (afdo_callsite_hot_enough_for_early_inline): Likewise.
16240         * coverage.c (struct counts_entry): Likewise.
16241         (read_counts_file): Read just single summary entry.
16242         (get_coverage_counts): Use gcov_summary struct.
16243         * coverage.h (get_coverage_counts): Likewise.
16244         * gcov-dump.c (dump_working_sets): Likewise.
16245         (tag_summary): Dump just single summary.
16246         * gcov-io.c (gcov_write_summary): Write just histogram
16247         summary.
16248         (gcov_read_summary): Read just single summary.
16249         (compute_working_sets): Use gcov_summary struct.
16250         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16251         of GCOV_COUNTERS_SUMMABLE.
16252         (GCOV_COUNTERS_SUMMABLE): Remove.
16253         (GCOV_FIRST_VALUE_COUNTER): Replace with
16254         GCOV_COUNTER_V_INTERVAL.
16255         (struct gcov_ctr_summary): Remove.
16256         (struct gcov_summary): Directly use fields of former
16257         gcov_ctr_summary.
16258         (compute_working_sets): Use gcov_summary struct.
16259         * gcov.c (read_count_file): Do not use ctrs fields.
16260         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16261         struct.
16262         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16263         struct.
16264         * profile.c: Likewise.
16265         * profile.h: Likewise.
16267 2018-06-05  Martin Liska  <mliska@suse.cz>
16269         PR gcov-profile/84846
16270         * gcov.c (output_lines): Print working directory only
16271         in intermediate format.
16273 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
16275         * config/s390/s390-builtin-types.def: Add void function type.
16276         * config/s390/s390-builtins.def: Use the function type for the
16277         tbeginc builtin.
16279 2018-06-04  Jim Wilson  <jimw@sifive.com>
16281         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16282         to int.
16283         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16284         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
16285         handle EH_RETURN_DATA_REGNO registers properly.
16286         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16287         (riscv_expand_epilogue): Update comment.  Change argument name and
16288         type.  Update code to use new name and type.  Pass new args to
16289         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
16290         EXCEPTION_RETURN.
16291         * config/riscv/riscv.md (NORMAL_RETURN): New.
16292         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16293         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16294         (eh_return): Call gen_eh_return_internal and emit barrier.
16295         (eh_return_internal): Call riscv_expand_epilogue.
16297 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
16299         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16300         bit_insertion field and declare can_be_merged_into method.
16301         (merged_store_group::can_be_merged_into): New method.
16302         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16303         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
16304         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16306 2018-06-04  Richard Biener  <rguenther@suse.de>
16308         PR tree-optimization/85955
16309         * builtins.c (fold_builtin_sincos): Convert pointers to
16310         destination to appropriate type before dereferencing.
16312 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
16314         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
16316 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
16318         * expr.c (expand_expr_real_1): Force the operand into memory if
16319         its TYPE_MODE is BLKmode and if there is no integer mode for
16320         the number of bits being extracted.
16322 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16324         PR target/85832
16325         PR target/86036
16326         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16327         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
16329 2018-06-04  Richard Biener  <rguenther@suse.de>
16331         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
16332         (cleanup_tree_cfg_noloop): ... single caller.  Do
16333         start_recording_case_labels later.
16335 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
16337         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
16338         to _IMMINTRIN_H_INCLUDED.
16339         * config/i386/pconfigintrin.h: Ditto.
16340         * config/i386/waitpkgintrin.h: Ditto.
16341         * config/i386/immintrin.h: Add includes for sgxintrin.h,
16342         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
16343         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16344         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
16345         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
16346         waitpkgintrin.h and cldemoteintrin.h.
16348 2018-06-04  Richard Biener  <rguenther@suse.de>
16350         PR tree-optimization/86038
16351         * tracer.c (find_best_successor): Check probability for
16352         being initialized, bail out if not.
16354 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
16356         PR target/86003
16357         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
16358         of bits to ignore when comparing architectures.
16360 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16362         PR tree-optimization/69615
16363         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
16364         maximum or minimum of the type, try to merge it also as if
16365         range1 is + [-, x - 1] or + [x + 1, -].
16367         PR c++/86025
16368         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
16370 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
16372         PR tree-optimization/86034
16373         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
16374         the unsigned bitfield type in a bit insertion sequence if it does not
16375         have a larger precision than the bitfield size.
16376         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
16378 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
16380         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
16382 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16384         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
16385         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
16386         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
16387         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
16389 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16391         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
16392         Disable -fdelete-null-pointer-checks for ELF toolchain.
16394 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16395             Kito Cheng  <kito.cheng@gmail.com>
16397         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
16398         (nds32le-*-*, nds32be-*-*): Integrate checking process.
16399         (nds32*-*-*): Add glibc and uclibc conditions.
16400         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
16401         (TARGET_EXCEPT_UNWIND_INFO): Define.
16402         * config/nds32/elf.h: New file.
16403         * config/nds32/linux.h: New file.
16404         * config/nds32/nds32-elf.opt: New file.
16405         * config/nds32/nds32-linux.opt: New file.
16406         * config/nds32/nds32-fp-as-gp.c
16407         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
16408         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
16409         TARGET_LINUX_ABI.
16410         (nds32_asm_file_end): Ditto.
16411         (nds32_print_operand): Ditto.
16412         (nds32_insert_attributes): Ditto.
16413         (nds32_init_libfuncs): New function.
16414         (TARGET_HAVE_TLS): Define.
16415         (TARGET_INIT_LIBFUNCS): Define.
16416         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
16417         spec content.
16418         (TARGET_ELF): Apply different mcmodel setting.
16419         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
16420         been migrated into elf.h and linux.h files.
16421         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
16422         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
16423         (mcmodel): The content has been migrated into nds32-elf.opt and
16424         nds32-linux.opt files.
16425         * config/nds32/t-elf: New file.
16426         * config/nds32/t-linux: New file.
16428 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16429             Shiva Chen  <shiva0217@gmail.com>
16431         * config/nds32/constants.md (unspec_volatile_element): Add
16432         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
16433         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
16434         optimization.
16435         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
16436         (make_pass_nds32_fp_as_gp): Declare.
16437         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
16438         optmization pass.
16439         (nds32_asm_function_end_prologue): Remove unused asm output.
16440         (nds32_asm_function_begin_epilogue): Remove unused asm output.
16441         (nds32_asm_file_start): Output necessary fp_as_gp information.
16442         (nds32_option_override): Adjust register usage.
16443         (nds32_expand_prologue): Consider fp_as_gp situation.
16444         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
16445         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
16446         (epilogue): Ditto.
16447         (return): Ditto.
16448         (simple_return): Ditto.
16449         (omit_fp_begin): Output special directive for fp_as_gp.
16450         (omit_fp_end): Output special directive for fp_as_gp.
16451         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
16452         mforbid-fp-as-gp): New options.
16454 2018-06-01  Mark Wielaard  <mark@klomp.org>
16456         * dwarf2out.c (dwarf2out_finish): Remove generation of
16457         DW_AT_loclists_base.
16459 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
16461         * gimple-ssa-store-merging.c: Include gimple-fold.h.
16462         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
16463         (struct merged_store_group): Add bit_insertion field.
16464         (dump_char_array): Use standard hexadecimal format.
16465         (merged_store_group::merged_store_group): Set bit_insertion to false.
16466         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
16467         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
16468         also print the mask in the dump file.
16469         (pass_store_merging::gate): Minor tweak.
16470         (imm_store_chain_info::coalesce_immediate): Fix wrong association
16471         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
16472         stores with INTEGER_CST stores.
16473         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
16474         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
16475         and use it throughout.  Generate bit insertion sequences if need be.
16476         (pass_store_merging::process_store): Remove redundant condition.
16477         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
16479 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
16481         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
16482         the 128-bit floating point types.  Fix function comment.
16484 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16486         * config/aarch64/aarch64-simd.md
16487         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
16488         mnemonics.
16489         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
16490         mnemonics.
16492 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16494         PR tree-optimization/85989
16495         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
16496         variable.
16497         (backprop::intersect_uses): Check it when deciding whether this
16498         is a backedge reference.
16499         (backprop::process_block): Add each phi to m_visited_phis
16500         after visiting it, then clear it at the end.
16502 2018-06-01  Richard Biener  <rguenther@suse.de>
16504         * tree-vectorizer.h (vect_dr_stmt): New function.
16505         (vect_get_load_cost): Adjust.
16506         (vect_get_store_cost): Likewise.
16507         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
16508         Use vect_dr_stmt instead of DR_SMTT.
16509         (vect_record_base_alignments): Likewise.
16510         (vect_calculate_target_alignment): Likewise.
16511         (vect_compute_data_ref_alignment): Likewise and make static.
16512         (vect_update_misalignment_for_peel): Likewise.
16513         (vect_verify_datarefs_alignment): Likewise.
16514         (vector_alignment_reachable_p): Likewise.
16515         (vect_get_data_access_cost): Likewise.  Pass down
16516         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
16517         (vect_get_peeling_costs_all_drs): Likewise.
16518         (vect_peeling_hash_get_lowest_cost): Likewise.
16519         (vect_enhance_data_refs_alignment): Likewise.
16520         (vect_find_same_alignment_drs): Likewise.
16521         (vect_analyze_data_refs_alignment): Likewise.
16522         (vect_analyze_group_access_1): Likewise.
16523         (vect_analyze_group_access): Likewise.
16524         (vect_analyze_data_ref_access): Likewise.
16525         (vect_analyze_data_ref_accesses): Likewise.
16526         (vect_vfa_segment_size): Likewise.
16527         (vect_small_gap_p): Likewise.
16528         (vectorizable_with_step_bound_p): Likewise.
16529         (vect_prune_runtime_alias_test_list): Likewise.
16530         (vect_analyze_data_refs): Likewise.
16531         (vect_supportable_dr_alignment): Likewise.
16532         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
16533         (vect_gen_prolog_loop_niters): Likewise.
16534         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
16535         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
16536         modify DR_STMT.
16537         (vect_recog_mask_conversion_pattern): Likewise.
16538         (vect_try_gather_scatter_pattern): Likewise.
16539         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
16540         to vect_get_store_cost.
16541         (vect_get_store_cost): Get stmt_info instead of DR.
16542         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
16543         (vect_get_load_cost): Get stmt_info instead of DR.
16545 2018-06-01  Richard Biener  <rguenther@suse.de>
16547         PR middle-end/86017
16548         * gimple-fold.c (var_decl_component_p): Also allow offsetted
16549         vars wrapped in MEM_REFs.
16551 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16553         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16554         Fix subreg tests so that we only return a choice between
16555         GENERAL_REGS and FP_REGS if the original classes included both.
16557 2018-06-01  Richard Biener  <rguenther@suse.de>
16559         PR ipa/85960
16560         * tree-ssa-structalias.c (get_function_part_constraint):
16561         Handle NULL fi->decl.
16562         (find_func_aliases_for_call): Properly handle indirect
16563         fi from direct call.
16564         (find_func_clobbers): Likewise.
16565         (ipa_pta_execute): Likewise.
16566         (create_variable_info_for): For functions that are ifunc_resolver
16567         resolve to a varinfo that contains the result of the resolver call.
16568         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
16569         aliases.
16571 2018-05-31  Michael Collison  <michael.collison@arm.com>
16573         * config/aarch64/aarch64.md:
16574         (*fix_to_zero_extenddfdi2): New pattern.
16575         * gcc.target/aarch64/fix_extend1.c: New testcase.
16577 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
16579         PR middle-end/78809
16580         PR middle-end/83026
16581         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
16582         and BUILT_IN_STRNCMP_EQ.
16583         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
16584         BUILT_IN_STRNCMP_EQ.
16585         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
16586         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16587         (gimple_fold_builtin): Likewise.
16588         * tree-ssa-strlen.c (compute_string_length): New function.
16589         (determine_min_obsize): New function.
16590         (handle_builtin_string_cmp): New function to handle calls to
16591         string compare functions.
16592         (strlen_optimize_stmt): Add handling to builtin string compare
16593         calls.
16594         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
16595         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16596         * tree.c (build_common_builtin_nodes): Add new defines of
16597         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
16599 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
16601         PR target/85984
16602         * bb-reorder.c (pass_partition_blocks::gate): Return false for
16603         functions with naked attribute.
16605 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
16607         * config/i386/sse.md (avx_vec_concat<mode>):
16608         Substitute concat_tg_mode mode attribute with xtg_mode.
16609         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
16610         (concat_tg_mode): Remove mode attribute.
16612 2018-05-31  Martin Sebor  <msebor@redhat.com>
16614         PR c/82063
16615         * calls.c (alloc_max_size): Correct a logic error/typo.
16616         Treat excessive arguments as infinite.  Warn for invalid arguments.
16617         * doc/invoke.texi (-Walloc-size-larger-than): Update.
16619 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
16621         PR target/85829
16622         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
16623         and movx for Haswell.
16625 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
16626             Cesar Philippidis  <cesar@codesourcery.com>
16628         PR middle-end/85879
16629         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
16630         when emitting error on private/firstprivate reductions.
16631         * omp-low.c (lower_omp_target): Avoid reference-type processing
16632         on pointers for firstprivate clause.
16634 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16636         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
16637         (st1x2): Likewise.
16638         (st1x3): Likewise.
16639         * config/aarch64/aarch64-simd.md
16640         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
16641         (aarch64_ld1_x3_<mode>): Likewise
16642         (aarch64_st1x2<VALLDIF:mode>): Likewise
16643         (aarch64_st1_x2_<mode>): Likewise
16644         (aarch64_st1x3<VALLDIF:mode>): Likewise
16645         (aarch64_st1_x3_<mode>): Likewise
16646         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
16647         (vld1_s8_x3): Likewise.
16648         (vld1_u16_x3): Likewise.
16649         (vld1_s16_x3): Likewise.
16650         (vld1_u32_x3): Likewise.
16651         (vld1_s32_x3): Likewise.
16652         (vld1_u64_x3): Likewise.
16653         (vld1_s64_x3): Likewise.
16654         (vld1_f16_x3): Likewise.
16655         (vld1_f32_x3): Likewise.
16656         (vld1_f64_x3): Likewise.
16657         (vld1_p8_x3): Likewise.
16658         (vld1_p16_x3): Likewise.
16659         (vld1_p64_x3): Likewise.
16660         (vld1q_u8_x3): Likewise.
16661         (vld1q_s8_x3): Likewise.
16662         (vld1q_u16_x3): Likewise.
16663         (vld1q_s16_x3): Likewise.
16664         (vld1q_u32_x3): Likewise.
16665         (vld1q_s32_x3): Likewise.
16666         (vld1q_u64_x3): Likewise.
16667         (vld1q_s64_x3): Likewise.
16668         (vld1q_f16_x3): Likewise.
16669         (vld1q_f32_x3): Likewise.
16670         (vld1q_f64_x3): Likewise.
16671         (vld1q_p8_x3): Likewise.
16672         (vld1q_p16_x3): Likewise.
16673         (vld1q_p64_x3): Likewise.
16674         (vst1_s64_x2): Likewise.
16675         (vst1_u64_x2): Likewise.
16676         (vst1_f64_x2): Likewise.
16677         (vst1_s8_x2): Likewise.
16678         (vst1_p8_x2): Likewise.
16679         (vst1_s16_x2): Likewise.
16680         (vst1_p16_x2): Likewise.
16681         (vst1_s32_x2): Likewise.
16682         (vst1_u8_x2): Likewise.
16683         (vst1_u16_x2): Likewise.
16684         (vst1_u32_x2): Likewise.
16685         (vst1_f16_x2): Likewise.
16686         (vst1_f32_x2): Likewise.
16687         (vst1_p64_x2): Likewise.
16688         (vst1q_s8_x2): Likewise.
16689         (vst1q_p8_x2): Likewise.
16690         (vst1q_s16_x2): Likewise.
16691         (vst1q_p16_x2): Likewise.
16692         (vst1q_s32_x2): Likewise.
16693         (vst1q_s64_x2): Likewise.
16694         (vst1q_u8_x2): Likewise.
16695         (vst1q_u16_x2): Likewise.
16696         (vst1q_u32_x2): Likewise.
16697         (vst1q_u64_x2): Likewise.
16698         (vst1q_f16_x2): Likewise.
16699         (vst1q_f32_x2): Likewise.
16700         (vst1q_f64_x2): Likewise.
16701         (vst1q_p64_x2): Likewise.
16702         (vst1_s64_x3): Likewise.
16703         (vst1_u64_x3): Likewise.
16704         (vst1_f64_x3): Likewise.
16705         (vst1_s8_x3): Likewise.
16706         (vst1_p8_x3): Likewise.
16707         (vst1_s16_x3): Likewise.
16708         (vst1_p16_x3): Likewise.
16709         (vst1_s32_x3): Likewise.
16710         (vst1_u8_x3): Likewise.
16711         (vst1_u16_x3): Likewise.
16712         (vst1_u32_x3): Likewise.
16713         (vst1_f16_x3): Likewise.
16714         (vst1_f32_x3): Likewise.
16715         (vst1_p64_x3): Likewise.
16716         (vst1q_s8_x3): Likewise.
16717         (vst1q_p8_x3): Likewise.
16718         (vst1q_s16_x3): Likewise.
16719         (vst1q_p16_x3): Likewise.
16720         (vst1q_s32_x3): Likewise.
16721         (vst1q_s64_x3): Likewise.
16722         (vst1q_u8_x3): Likewise.
16723         (vst1q_u16_x3): Likewise.
16724         (vst1q_u32_x3): Likewise.
16725         (vst1q_u64_x3): Likewise.
16726         (vst1q_f16_x3): Likewise.
16727         (vst1q_f32_x3): Likewise.
16728         (vst1q_f64_x3): Likewise.
16729         (vst1q_p64_x3): Likewise.
16731 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16733         * config/msp430/msp430.c (msp430_output_labelref): Prepend
16734         user_label_prefix to name.
16736         * tree-core.h: Update comment about the format of NAME string
16737         passed to handler in attribute_spec.
16739         * config/msp430/msp430.md: Remove erroneous subreg expression from
16740         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
16741         zero_extend{q,h}isi2.
16743 2018-05-30  Borislav Petkov  <bp@suse.de>
16745         * doc/extend.texi: Document some architecture specific
16746         constraints and sort entries.
16748 2018-05-30  Martin Sebor  <msebor@redhat.com>
16750         PR middle-end/85369
16751         * builtins.c (expand_builtin_stpcpy_1): New function.
16752         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
16753         only if the former succeeds.
16755 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16757         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
16758         in saphira.
16760 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16762         * doc/invoke.texi (-flinker-output): Document
16764 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16766         * passes.c (ipa_write_summaries): Only modify statements if body
16767         is in memory.
16768         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
16769         incrementally linking.
16770         (ipa_passes): Likewise.
16771         * lto-cgraph.c (lto_output_node): When incrementally linking do not
16772         pass down resolution info.
16773         * common.opt (flag_incremental_link): Update info.
16774         * gcc.c (plugin specs): Turn flinker-output=* to
16775         -plugin-opt=-linker-output-known
16776         * toplev.c (compile_file): Also cut compilation when doing incremental
16777         link.
16778         * flag-types. (enum lto_partition_model): Add
16779         LTO_LINKER_OUTPUT_NOLTOREL.
16780         (invoke.texi): Add -flinker-output docs.
16781         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
16782         link same way as WPA; do not stream in dead initializers.
16784         * dwarf2out.c (dwarf2out_die_ref_for_decl,
16785         darf2out_register_external_decl): Support incremental link.
16787 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16789         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
16791 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16793         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
16794         it down to simple_object_copy_lto_debug_sections.
16795         (run_gcc): Determine incremental LTO link time and configure
16796         lto1 into non-wpa mode, disable renaming of debug sections.
16798 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16800         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16801         descriptions of various incorrectly documented functions.
16803 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16805         Revert:
16806         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16807         address check not strict.
16809 2018-05-30  Richard Biener  <rguenther@suse.de>
16811         PR tree-optimization/85964
16812         * tracer.c (better_p): Drop initialized count check, we only
16813         call the function with initialized counts now.
16814         (find_best_successor): Do find a best edge if one
16815         has uninitialized count.
16816         (find_best_predecessor): Likewise.  Do BB frequency check only
16817         if count is initialized.
16819 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
16821         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
16822         (aarch64_ldrstr_offset_compare): New.
16823         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
16824         load/store orderings.
16825         (aarch64_gen_adjusted_ldpstp): Likewise.
16827 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
16829         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16830         Check for subset of GENERAL_REGS and FP_REGS.
16831         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
16832         r=w alternative.
16834 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
16836         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
16837         and wi::to_poly_offset.  Add the current offset and then check
16838         whether the sum fits, rather than using an unchecked addition of
16839         a checked term.  Check for a shwi rather than a uhwi.
16840         * expr.c (get_bit_range): Use tree_to_poly_uint64.
16841         (store_constructor): Use poly_int_tree_p.
16842         (expand_expr_real_1): Likewise.
16843         * function.c (assign_temp): Likewise.
16844         * fold-const.c (const_binop): Use poly_int_tree_p and
16845         wi::to_poly_offset.
16846         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
16847         division.
16848         * ipa-icf-gimple.c (func_checker::compare_operand): Use
16849         to_poly_offset for MEM offsets.
16850         * ipa-icf.c (sem_variable::equals): Likewise.
16851         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
16852         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
16853         wi::to_poly_offset for BIT_FIELD_REF offsets.
16854         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
16855         wi::to_poly_offset.
16856         * var-tracking.c (emit_note_insn_var_location): Use
16857         tree_to_poly_uint64.
16859 2018-05-29  Jim Wilson  <jimw@sifive.com>
16861         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
16863 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
16865         PR target/85950
16866         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
16867         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
16868         sequence.
16869         (sse4_1_round<mode>2): Use nonimmediate_operand
16870         for operand 1 predicate.
16872 2018-05-29  Martin Sebor  <msebor@redhat.com>
16873             Richard Biener  <rguenther@suse.de>
16875         PR testsuite/85888
16876         * calls.c (get_size_range): Call determine_value_range instead
16877         of get_value_range..
16878         * tree-vrp.h (determine_value_range): Declared new function.
16879         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
16881 2018-05-29  Richard Biener  <rguenther@suse.de>
16883         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
16884         sure to use non-pattern stmts for get_earlier_stmt arguments.
16885         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
16886         called on pattern stmts.
16887         (get_later_stmt): Likewise.
16889 2018-05-29  Martin Liska  <mliska@suse.cz>
16891         PR gcov-profile/85759
16892         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
16893         env variables.
16895 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
16897         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
16898         VEC_UNPACK_*_EXPR.
16899         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
16900         VEC_PACK_*_EXPR.
16902         PR target/85918
16903         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16904         VEC_PACK_FLOAT_EXPR): New tree codes.
16905         * tree-pretty-print.c (op_code_prio): Handle
16906         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16907         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16908         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16909         * tree-inline.c (estimate_operator_cost): Likewise.
16910         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
16911         * fold-const.c (const_binop): Likewise.
16912         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
16913         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16914         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16915         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
16916         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16917         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16918         * expr.c (expand_expr_real_2): Likewise.
16919         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
16920         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
16921         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
16922         optabs.
16923         * optabs.c (expand_widen_pattern_expr): For
16924         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
16925         sign from result type rather than operand's type.
16926         (expand_binop_directly): For vec_packu_float_optab and
16927         vec_packs_float_optab allow result type to be different from operand's
16928         type.
16929         * optabs-tree.c (optab_for_tree_code): Handle
16930         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16931         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
16932         * tree-vect-generic.c (expand_vector_operations_1):  Handle
16933         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16934         VEC_PACK_FLOAT_EXPR.
16935         * tree-vect-stmts.c (supportable_widening_operation): Handle
16936         FIX_TRUNC_EXPR.
16937         (supportable_narrowing_operation): Handle FLOAT_EXPR.
16938         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
16939         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
16940         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
16941         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
16942         mode attributes.
16943         (vec_pack<floatprefix>_float_<mode>): New expander.
16944         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
16945         attributes.
16946         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
16947         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
16948         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
16949         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
16950         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
16951         Document.
16952         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
16953         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
16954         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16955         VEC_PACK_FLOAT_EXPR): Document.
16957 2018-05-29  Richard Biener  <rguenther@suse.de>
16959         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
16960         member.
16961         (stmt_vec_info_vec): Make pointer.
16962         (init_stmt_vec_info_vec): Remove.
16963         (free_stmt_vec_info_vec): Likewise.
16964         (set_stmt_vec_info_vec): New function.
16965         (free_stmt_vec_infos): Likewise.
16966         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
16967         (set_vinfo_for_stmt): Likewise.
16968         (get_earlier_stmt): Likewise.
16969         (get_later_stmt): Likewise.
16970         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
16971         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
16972         (vec_info::~vec_info): Free stmt_vec_infos.
16973         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
16974         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
16975         (pass_slp_vectorize::execute): Likewise.
16976         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
16977         (free_stmt_vec_info_vec): Likewise.
16978         (set_stmt_vec_info_vec): New function.
16979         (free_stmt_vec_infos): Likewise.
16980         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
16981         the global stmt_vec_info_vec.
16982         * tree-parloops.c (gather_scalar_reductions): Use
16983         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
16984         vector.
16986 2018-05-29  Richard Biener  <rguenther@suse.de>
16988         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
16990 2018-05-29  Martin Liska  <mliska@suse.cz>
16991             David Malcolm  <dmalcolm@redhat.com>
16993         * vec.c (test_reverse): New.
16994         (vec_c_tests): Add new test.
16995         * vec.h (vl_ptr>::reverse): New function.
16997 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
16999         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17001         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17002         and later.
17004 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17006         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17008 2018-05-28  Richard Biener  <rguenther@suse.de>
17010         PR tree-optimization/85933
17011         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17012         look at stmts marked as vectorizable.
17014 2018-05-28  Richard Biener  <rguenther@suse.de>
17016         PR tree-optimization/85934
17017         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17018         vector boolean check before scalar optimization.
17020 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17022         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17023         for armv5te.
17025 2018-05-28  Mark Wielaard  <mark@klomp.org>
17027         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17028         if it is an expression containing a minus sign.
17030 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17032         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17034 2018-05-27  Paul Koning  <ni1d@arrl.net>
17036         * config/pdp11/pdp11.md (truncsihi2): Remove.
17038 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17039             Chung-Ju Wu  <jasonwucj@gmail.com>
17041         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17042         implementation.
17043         (unaligned_store_dw): Ditto.
17044         * config/nds32/nds32-memory-manipulation.c
17045         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17046         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17047         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17048         (emit_setmem_word_loop): Rename to ...
17049         (emit_setmem_doubleword_loop): ... this.
17050         (nds32_gen_dup_4_byte_to_word_value): New function.
17051         (nds32_gen_dup_8_byte_to_double_word_value): New function.
17052         (nds32_expand_setmem_loop): Refine implementation.
17053         (nds32_expand_setmem_loop_v3m): Ditto.
17054         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17055         pattern.
17057 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17059         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17061 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17063         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17064         (nds32_init_machine_status): Initialize machine->attr_naked_p and
17065         machine->attr_no_prologue_p.
17066         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17067         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17068         (nds32_expand_epilogue): Consider attr_naked_p.
17069         (nds32_expand_epilogue_v3pop): Likewise.
17070         (nds32_can_use_return_insn): Likewise.
17071         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17072         attr_no_prologue_p fields.
17073         * config/nds32/nds32.opt (mret-in-naked-func): New option.
17075 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
17077         PR target/85918
17078         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17079         attributes.
17080         * config/i386/sse.md
17081         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17082         Rename to ...
17083         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17084         ... this.
17085         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17086         Rename to ...
17087         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17088         ... this.
17089         (*<floatsuffix>floatv2div2sf2): Rename to ...
17090         (*float<floatunssuffix>v2div2sf2): ... this.
17091         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17092         (float<floatunssuffix>v2div2sf2_mask): ... this.
17093         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17094         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17095         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17096         to ...
17097         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17098         ... this.
17099         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17100         Rename to ...
17101         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17102         ... this.
17103         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17104         Rename to ...
17105         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17106         ... this.
17107         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17108         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17109         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17110         gen_ufix_truncv8dfv8si2.
17111         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17112         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17113         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17114         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17115         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17116         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17117         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17118         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17120 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17122         PR target/85900
17123         PR target/85345
17124         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17126 2018-05-25  Jim Wilson  <jimw@sifive.com>
17128         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17129         * config/riscv/riscv.c (struct machine_function): Add
17130         interrupt_handler_p and attribute_checked_p fields.
17131         (riscv_attribute_table): Add interrupt.
17132         (riscv_interrupt_type_p): New.
17133         (riscv_save_reg_p): Save extra regs for interrupt handler.
17134         (riscv_use_save_libcall): Return false  for interrupt handler.
17135         (riscv_first_stack_step): Add forward declaration.
17136         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
17137         for interrupt handler with large frame.  Use it for saved reg list.
17138         (riscv_expand_prologue): Move flag_stack_usage_info support to
17139         eliminate duplication.
17140         (riscv_expand_epilogue): Generate mret for interrupt handler.
17141         (riscv_epilogue_uses): New.
17142         (riscv_can_use_return_insn): Return false for interrupt handler.
17143         (riscv_function_ok_for_sibcall): Likewise.
17144         (riscv_set_current_function): Add interrupt handler support.
17145         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17146         * config/riscv/riscv.md (UNSPECV_MRET): New.
17147         (GP_REGNUM): New.
17148         (riscv_frflags, riscv_fsflags): Use tab after opcode.
17149         (riscv_mret): New.
17150         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17152 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
17154         PR tree-optimization/85712
17155         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17156         this candidate has already been replaced in-situ by a copy.
17158 2018-05-25  Jason Merrill  <jason@redhat.com>
17160         PR c++/80485 - inline function non-zero address.
17161         * symtab.c (nonzero_address): Check DECL_COMDAT.
17163 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17165         PR target/83628
17166         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17167         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17168         extension of SImode operation.  Use const123_operand predicate.
17169         (*saddsi_1): Remove.
17170         (*saddl_se_1): Ditto.
17171         (*ssubsi_1): Ditto.
17172         (*ssubl_se_1): Ditto.
17173         * config/alpha/predicates.md (const123_operand): New predicate.
17174         * config/alpha/constraints.md (P): Use IN_RANGE.
17176 2018-05-25  Richard Biener  <rguenther@suse.de>
17178         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17179         defaulted to true.
17180         (ref_maybe_used_by_stmt_p): Likewise.
17181         (stmt_may_clobber_ref_p): Likewise.
17182         (stmt_may_clobber_ref_p_1): Likewise.
17183         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17184         and pass it along.
17185         (ref_maybe_used_by_stmt_p): Likewise.
17186         (stmt_may_clobber_ref_p): Likewise.
17187         (stmt_may_clobber_ref_p_1): Likewise.
17188         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17189         the alias oracle to disambiguate DRs with stmts DR analysis
17190         couldn't handle.
17191         (vect_analyze_data_refs): Do not give up on not analyzable
17192         DRs for BB vectorization.  Remove code truncating the dataref
17193         vector.
17195 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
17197         PR target/85832
17198         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17199         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
17200         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17202 2018-05-25  Richard Biener  <rguenther@suse.de>
17204         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17205         function, combining stmt data ref gathering and fatal analysis
17206         parts.
17207         (vect_analyze_data_refs): Remove now redudnant code and simplify.
17208         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17209         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17210         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17211         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17213 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17215         PR tree-optimization/85720
17216         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17217         SCC if all partitions are builtins.
17218         (version_loop_by_alias_check): New parameter.  Generate cancelable
17219         runtime alias check if all partitions are builtins.
17220         (distribute_loop): Update call to above function.
17222 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17224         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17225         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17226         (parm_default_def_partition_arg): Ditto.
17227         (set_parm_default_def_partition): Ditto.
17228         (get_parm_default_def_partitions): Ditto and make it static.
17229         (get_undefined_value_partitions): Ditto and make it static.
17230         (remove_ssa_form): Refactor call to init_var_map here.
17231         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17232         computation for loop region.
17233         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17234         (register_default_def): Delete.
17235         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17236         (parm_default_def_partition_arg): Ditto.
17237         (set_parm_default_def_partition): Ditto.
17238         (get_parm_default_def_partitions): Ditto and make it static.
17239         (get_undefined_value_partitions): Ditto and make it static.
17240         (coalesce_with_default, coalesce_with_default): Update comment.
17241         (create_coalesce_list_for_region): New func factored out from
17242         create_outofssa_var_map.
17243         (populate_coalesce_list_for_outofssa): New func factored out from
17244         create_outofssa_var_map and coalesce_ssa_name.
17245         (create_outofssa_var_map): Delete.
17246         (coalesce_ssa_name): Refactor to support live range computation.
17247         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17248         (get_parm_default_def_partitions): Delete.
17249         (get_undefined_value_partitions): Ditto.
17250         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17251         computation for loop region.
17252         (new_tree_live_info, loe_visit_block): Ditto.
17253         (live_worklist, set_var_live_on_entry): Ditto.
17254         (calculate_live_on_exit, verify_live_on_entry): Ditto.
17255         * tree-ssa-live.h (struct _var_map): New fields.
17256         (init_var_map): Change decl.
17257         (region_contains_p): New.
17259 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17261         * tree-ssa-live.h (live_merge_and_clear): Delete.
17263 2018-05-25  Richard Biener  <rguenther@suse.de>
17265         PR c++/85912
17266         * tree-dump.c (dequeue_and_dump): Remove access to removed
17267         operand 2 of a SWITCH_EXPR.
17269 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17271         * doc/sourcebuild.texi (vect_double_cond_arith): Include
17272         multiplication and division.
17273         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17274         (cond_udiv@var{m}, cond_umod@var{m}): Document.
17275         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17276         (cond_udiv_optab, cond_umod_optab): New optabs.
17277         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17278         (IFN_COND_RDIV): New internal functions.
17279         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17280         TRUNC_MOD_EXPR and RDIV_EXPR.
17281         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17282         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17283         New unspecs.
17284         (SVE_INT_BINARY): Include mult.
17285         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17286         (optab, sve_int_op): Handle mult.
17287         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17288         UNSPEC_COND_DIV.
17289         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17290         for SVE_INT_BINARY_SD.
17292 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17294         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17295         (optab, sve_int_op): Handle div and udiv.
17296         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17297         for SVE_INT_BINARY_SD.
17298         (*<optab><mode>3): New insn for the same.
17300 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17302         * tree-vect-patterns.c: Include predict.h.
17303         (vect_recog_divmod_pattern): Restrict check for division support
17304         to when optimizing for size.
17306 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17308         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17309         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17310         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17311         (gimple_match_op::set_op): Likewise.
17312         (gimple_resimplify4): Declare.
17313         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
17314         (expr::gen_transform): Likewise.
17315         (decision_tree::gen): Generate a simplification routine for 4 operands.
17316         * gimple-match-head.c (gimple_simplify): Add an overload for
17317         4 operands.  In the top-level function, handle up to 4 call
17318         arguments and call gimple_resimplify4.
17319         (gimple_resimplify4): New function.
17320         (build_call_internal): Pass a fourth operand.
17321         (maybe_push_to_seq): Likewise.
17322         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
17323         Fold VEC_COND_EXPRs of an operation and a default value into
17324         an IFN_COND_* function if possible.
17325         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
17326         New unspecs.
17327         (SVE_COND_FP_BINARY): Include them.
17328         (optab, sve_fp_op): Handle them.
17329         (SVE_INT_BINARY_REV): New code iterator.
17330         (SVE_COND_FP_BINARY_REV): New int iterator.
17331         (commutative): New int attribute.
17332         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
17333         Declare.
17334         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
17335         function.
17336         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
17337         (*cond_<optab><mode>): New patterns for reversed operands.
17339 2018-05-25  Richard Biener  <rguenther@suse.de>
17341         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
17342         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
17343         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
17344         (STMT_VINFO_GROUPED_ACCESS): Adjust.
17345         * tree-vect-data-refs.c (everywhere): Adjust users.
17346         * tree-vect-loop.c (everywhere): Likewise.
17347         * tree-vect-slp.c (everywhere): Likewise.
17348         * tree-vect-stmts.c (everywhere): Likewise.
17349         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
17351 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17353         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
17354         Rename to...
17355         (gcc_cv_as_section_exclude): ... this.
17356         Try Solaris as #exclude syntax.
17357         * configure: Regenerate.
17358         * config.in: Regenerate.
17359         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
17360         SECTION_EXCLUDE.
17361         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
17362         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
17364         * varasm.c (default_elf_asm_named_section): Don't check if
17365         HAVE_GAS_SECTION_EXCLUDE is defined.
17367 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17369         * doc/md.texi: Update the documentation of the cond_* optabs
17370         to mention the new final operand.  Fix GET_MODE_NUNITS call.
17371         Describe the scalar case too.
17372         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
17373         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
17374         instead of 2.
17375         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
17376         (get_conditional_internal_fn): Update comment.
17377         * tree-vect-loop.c (vectorizable_reduction): Pass the original
17378         accumulator value as a final argument to conditional functions.
17379         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
17380         a define_expand and add an "else" operand.  Assert for now that
17381         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
17382         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
17383         (*cond_<optab><mode>): New patterns.
17384         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
17385         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17386         (UNSPEC_COND_EOR): Delete.
17387         (optab): Remove associated mappings.
17388         (SVE_INT_BINARY): New code iterator.
17389         (sve_int_op): Remove int attribute and add "minus" to the code
17390         attribute.
17391         (SVE_COND_INT_OP): Delete.
17392         (SVE_COND_FP_OP): Rename to...
17393         (SVE_COND_FP_BINARY): ...this.
17395 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17397         * optabs.c (can_reuse_operands_p): New function.
17398         (maybe_legitimize_operands): Try to reuse the results for
17399         earlier operands.
17401 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17403         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
17404         Add {q} suffix to insn mnemonic.
17406 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17408         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
17409         (msp430_warn_func_return): New.
17411 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
17413         * fold-const.c (tree_nonzero_bits): New function.
17414         * fold-const.h (tree_nonzero_bits): Likewise.
17415         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
17416         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
17418 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17420         PR target/85900
17421         PR target/85345
17422         * varasm.c (assemble_alias): Check ifunc_resolver only on
17423         FUNCTION_DECL.
17425 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17427         PR target/85903
17428         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
17429         when memory input operand is handled.
17431 2018-05-24  Luis Machado  <luis.machado@linaro.org>
17433         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
17434         global.
17435         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
17437 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17439         * match.pd: Delay FMA folds until after vectorization.
17441 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17443         PR target/83009
17444         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17445         address check not strict.
17447 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17449         * gimple-match.h (gimple_match_op): New class.
17450         (mprts_hook): Replace parameters with a gimple_match_op *.
17451         (maybe_build_generic_op): Likewise.
17452         (gimple_simplified_result_is_gimple_val): Replace parameters with
17453         a const gimple_match_op *.
17454         (gimple_simplify): Replace code_helper * and tree * parameters with
17455         a gimple_match_op * parameter.
17456         (gimple_resimplify1): Replace code_helper *, tree and tree *
17457         parameters with a gimple_match_op * parameter.
17458         (gimple_resimplify2): Likewise.
17459         (gimple_resimplify3): Likewise.
17460         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
17461         parameters with a gimple_match_op * parameter.
17462         * gimple-match-head.c (gimple_simplify): Change prototypes of
17463         auto-generated functions to take a gimple_match_op * instead of
17464         separate code_helper * and tree * parameters.  Make the same
17465         change in the top-level overload and update calls to the
17466         gimple_resimplify routines.  Update calls to the auto-generated
17467         functions and to maybe_push_res_to_seq in the publicly-facing
17468         operation-specific gimple_simplify overloads.
17469         (gimple_match_op::MAX_NUM_OPS): Define.
17470         (gimple_resimplify1): Replace rcode and ops with a single res_op
17471         parameter.  Update call to gimple_simplify.
17472         (gimple_resimplify2): Likewise.
17473         (gimple_resimplify3): Likewise.
17474         (mprts_hook): Replace parameters with a gimple_match_op *.
17475         (maybe_build_generic_op): Likewise.
17476         (build_call_internal): Replace type, nargs and ops with
17477         a gimple_match_op *.
17478         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
17479         with a single gimple_match_op *.  Update calls to mprts_hook,
17480         build_call_internal and gimple_simplified_result_is_gimple_val.
17481         Factor out code that is common to the tree_code and combined_fn cases.
17482         * genmatch.c (expr::gen_transform): Replace tem_code and
17483         tem_ops with a gimple_match_op called tem_op.  Update calls
17484         to the gimple_resimplify functions and maybe_push_res_to_seq.
17485         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
17486         res_ops.  Update call to the gimple_resimplify functions.
17487         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
17488         (decision_tree::gen): Make the functions take a gimple_match_op *
17489         called res_op instead of separate res_code and res_ops parameters.
17490         Update call accordingly.
17491         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
17492         and ops with a single res_op parameter.  Update calls to
17493         maybe_build_generic_op and maybe_push_res_to_seq.
17494         (fold_stmt_1): Update calls to gimple_simplify and
17495         replace_stmt_with_simplification.
17496         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
17497         and gimple_simplified_result_is_gimple_val.
17498         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
17499         gimple_simplify.
17500         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
17501         with a gimple_match_op *.
17502         (vn_nary_build_or_lookup): Likewise.  Update call to
17503         vn_nary_build_or_lookup_1.
17504         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
17505         gimple_match_op *.  Update calls to the gimple_resimplify routines
17506         and to gimple_simplified_result_is_gimple_val.
17507         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
17508         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
17509         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
17510         (visit_nary_op): Likewise.
17511         (visit_reference_op_load): Likewise.
17513 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17515         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
17516         modifier for printing the step amount.
17518 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
17520         PR target/78849
17521         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
17522         types.
17524 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
17526         * doc/sourcebuild.texi (Endianness): New subsubsection.
17528 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17530         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17531         <prefetch_dynamic_strides>: New const bool field.
17532         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17533         prefetch_dynamic_strides.
17534         (exynosm1_prefetch_tune): Likewise.
17535         (thunderxt88_prefetch_tune): Likewise.
17536         (thunderx_prefetch_tune): Likewise.
17537         (thunderx2t99_prefetch_tune): Likewise.
17538         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
17539         false.
17540         (aarch64_override_options_internal): Update to set
17541         PARAM_PREFETCH_DYNAMIC_STRIDES.
17542         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17543         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17544         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17545         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17546         prefetch-dynamic-strides setting.
17548 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17550         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17551         <minimum_stride>: New const int field.
17552         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17553         minimum_stride field defaulting to -1.
17554         (exynosm1_prefetch_tune): Likewise.
17555         (thunderxt88_prefetch_tune): Likewise.
17556         (thunderx_prefetch_tune): Likewise.
17557         (thunderx2t99_prefetch_tune): Likewise.
17558         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
17559         <default_opt_level>: Set to 3.
17560         (aarch64_override_options_internal): Update to set
17561         PARAM_PREFETCH_MINIMUM_STRIDE.
17562         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17563         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17564         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17565         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17566         stride is constant and is below the minimum stride threshold.
17568 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17570         * config/arm/arm-cpus.in (mode26): Delete.
17571         (armv4): Delete mode26 reference.
17572         * config/arm/arm.c (arm_configure_build_target): Delete use of
17573         isa_bit_mode26.
17575 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17577         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
17578         New insn pattern.
17579         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
17580         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
17581         for non-SSE modes.
17582         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
17583         (floatunsdidf2): Ditto.
17585 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17587         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
17588         (fixuns_trunc<mode>si2_avx512f): Ditto.
17589         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
17590         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
17591         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
17593 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17595         PR rtl-optimization/79985
17596         * df-scan.c (df_insn_refs_collect): Remove special case for
17597         global registers and asm statements.
17599 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17601         * extend.texi (Global Register Variables): Rewrite the bullet list.
17602         Note that the register is available for allocation. Note that access
17603         via inline asm must use constraints. Add note about async-signal
17604         handlers. Remove paragraph about automagic register selection.
17606 2018-05-23  Richard Biener  <rguenther@suse.de>
17608         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
17609         of fixed offset from memset VN.
17611 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
17613         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
17614         first_interp field.
17615         (alloc_cand_and_find_basis): Initialize first_interp field.
17616         (slsr_process_mul): Modify first_interp field.
17617         (slsr_process_add): Likewise.
17618         (slsr_process_cast): Modify first_interp field for each new
17619         interpretation.
17620         (slsr_process_copy): Likewise.
17621         (dump_candidate): Dump first_interp field.
17622         (replace_mult_candidate): Process all interpretations, not just
17623         subsequent ones.
17624         (replace_rhs_if_not_dup): Likewise.
17625         (replace_one_candidate): Likewise.
17627 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
17629         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
17630         Add new boolean.
17631         (aarch64_needs_frame_chain): New function.
17632         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
17634 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
17636         PR target/84882
17637         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
17638         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
17639         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
17640         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
17641         as true for strict-align.
17642         (aarch64_can_inline_p): Perform checks even when callee has no
17643         attributes to check for strict alignment.
17644         * doc/extend.texi (AArch64 Function Attributes): Document
17645         no-strict-align.
17646         * doc/invoke.texi: (AArch64 Options): Likewise.
17648 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
17650         PR tree-optimization/85853
17651         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
17652         the handling of the root of the node to...
17653         (vect_slp_analyze_node_operations_1): ...this new function,
17654         and run the whole thing with the child nodes' def types
17655         set according to their SLP node's def type.
17657 2018-05-23  Richard Biener  <rguenther@suse.de>
17659         PR middle-end/85874
17660         * tree-data-ref.c (create_runtime_alias_checks): Defer
17661         and ignore overflow warnings.
17663 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
17665         PR tree-optimization/85822
17666         * tree-vrp.c (is_masked_range_test): Fix handling of negative
17667         constants.
17669 2018-05-23  Richard Biener  <rguenther@suse.de>
17671         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17672         memset constants via native_interpret_expr.
17674 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
17676         PR target/85345
17677         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
17678         attribute.
17679         (cgraph_node::create_alias): Likewise.
17680         (cgraph_node::get_availability): Check ifunc_resolver instead
17681         of looking up ifunc attribute.
17682         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
17683         * varasm.c (do_assemble_alias): Likewise.
17684         (assemble_alias): Likewise.
17685         (default_binds_local_p_3): Likewise.
17686         * cgraph.h (cgraph_node): Add ifunc_resolver.
17687         (cgraph_node::only_called_directly_or_aliased_p): Return false
17688         for IFUNC resolver.
17689         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
17690         attribute.
17691         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
17692         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
17693         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
17694         instead of looking up ifunc attribute.
17696 2018-05-22  Luis Machado  <luis.machado@linaro.org>
17698         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
17700 2018-05-22  Martin Sebor  <msebor@redhat.com>
17702         PR middle-end/85359
17703         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
17704         only when expasion succeeds.
17705         (expand_builtin_strcmp): Same.
17706         (expand_builtin_strncmp): Same.
17708 2018-05-22  Martin Sebor  <msebor@redhat.com>
17710         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
17712 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17713             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17715         * config/aarch64/aarch64-ldpstp.md: Replace uses of
17716         aarch64_mem_pair_operand with memory_operand and delete operand swapping
17717         code.
17718         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17719         Add check for legitimate_address.
17720         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
17721         (aarch64_swap_ldrstr_operands): New.
17722         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
17723         Define prototype.
17725 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17726             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17728         * config/aarch64/aarch64.md: New patterns to generate stp
17729         and ldp.
17730         (store_pair_sw, store_pair_dw): New patterns to generate stp for
17731         single words and double words.
17732         (load_pair_sw, load_pair_dw): Likewise.
17733         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
17734         Delete.
17735         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
17736         Delete.
17737         * config/aarch64/aarch64-ldpstp.md: Modify peephole
17738         for different mode ldpstp and add peephole for merged zero stores.
17739         Likewise for loads.
17740         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17741         Add size check.
17742         (aarch64_gen_store_pair): Rename calls to match new patterns.
17743         (aarch64_gen_load_pair): Rename calls to match new patterns.
17744         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
17745         (load_pair<DREG:mode><DREG2:mode>): ... This.
17746         (store_pair<mode>): Rename to...
17747         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
17748         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
17749         New mode iterators.
17750         (V_INT_EQUIV): Handle SImode.
17751         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
17752         New predicate.
17754 2018-05-22  Martin Sebor  <msebor@redhat.com>
17756         PR c/85623
17757         * calls.c (maybe_warn_nonstring_arg): Use string length to set
17758         or ajust the presumed bound on an operation to avoid unnecessary
17759         warnings.
17761 2018-05-22  Martin Sebor  <msebor@redhat.com>
17763         PR tree-optimization/85826
17764         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
17765         assuming that a DECL necesarily has a constant size.
17767 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
17769         PR middle-end/85862
17770         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
17772 2018-05-22  Richard Biener  <rguenther@suse.de>
17774         PR tree-optimization/85834
17775         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
17776         non-constant and non-zero memset arguments.
17778 2018-05-22  Martin Liska  <mliska@suse.cz>
17780         PR ipa/85607
17781         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
17783 2018-05-22  Richard Biener  <rguenther@suse.de>
17785         PR tree-optimization/85863
17786         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
17787         comparisons when vectype is specified.
17788         (vectorizable_condition): Do not specify vectype for
17789         vect_is_simple_cond when SLP vectorizing.
17791 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
17793         PR target/85657
17794         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
17795         define __ibm128 as long double.
17796         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
17797         as a distinct type when IEEE 128-bit support is enabled.
17798         (init_float128_ieee): Fix up conversions between IFmode and IEEE
17799         128-bit types to use the correct functions.
17800         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
17801         convert between 128-bit floating point types that have different
17802         modes but the same representation, instead of using gen_lowpart to
17803         makean alias.
17804         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
17805         KFmode.
17806         (IFKF_reg): New attributes to give the register constraints for
17807         IFmode and KFmode.
17808         (extend<mode>tf2_internal): New insns to mark an explicit
17809         conversion between 128-bit floating point types that have a
17810         different mode but share the same representation.
17812 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
17814         PR tree-optimization/85814
17815         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
17816         a null return from get_strinfo when unsharing the next
17817         strinfo in the chain.
17819 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
17821         PR gcc/84923
17822         * varasm.c (weak_finish): Clean up weak_decls.
17824 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17826         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
17827         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
17828         UNSPEC_UADALP values.
17829         * config/aarch64/iterators.md (ABAL): New int iterator.
17830         (ABDL2): Likewise.
17831         (ADALP): Likewise.
17832         (sur): Add mappings for the above.
17833         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
17834         New define_insn.
17835         (aarch64_<sur>abal<mode>_4): Likewise.
17836         (aarch64_<sur>adalp<mode>_3): Likewise.
17837         (<sur>sadv16qi): New define_expand.
17839 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
17841         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
17842         (*movdf_internal): Ditto.
17843         (*rcpsf2_sse): Ditto.
17844         (*rsqrtsf2_sse): Ditto.
17845         (*sqrt<mode>2_sse): Ditto.
17847 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
17849         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
17850         eor3q<mode>4.
17851         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
17852         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
17853         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
17854         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
17855         vbcaxq_s64): New.
17856         * config/aarch64/arm_neon.h: Likewise.
17857         * config/aarch64/iterators.md (VQ_I): New.
17859 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
17861         * config.gcc: Add arc/t-multilib-linux to tmake_file for
17862         arc*-*-linux*.
17863         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
17864         MULTILIB_DIRNAMES
17866 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
17868         * config/nds32/constraints.md (S): New constraint.
17869         * config/nds32/nds32.md (call_internal): Use constraint S.
17870         (call_value_internal): Likewise.
17871         (sibcall_internal): Likewise.
17872         (sibcall_value_internal): Likewise.
17874 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
17875             Chung-Ju Wu  <jasonwucj@gmail.com>
17877         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
17878         into consideration.
17880 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
17881             Chung-Ju Wu  <jasonwucj@gmail.com>
17883         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
17884         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
17885         (nds32_rtx_costs_impl): Simplify.
17886         (nds32_address_cost_impl): Simplify.
17887         (nds32_init_rtx_costs): New function.
17888         (nds32_rtx_costs_speed_prefer): Likewise.
17889         (nds32_rtx_costs_size_prefer): Likewise.
17890         (nds32_address_cost_speed_prefer): Likewise.
17891         (nds32_address_cost_speed_fwprop): Likewise.
17892         (nds32_address_cost_size_prefer): Likewise.
17893         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
17894         * config/nds32/nds32.c (nds32_option_override): Use
17895         nds32_init_rtx_costs function.
17897 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
17899         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
17900         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
17901         (TARGET_PIPELINE_N8): Likewise.
17902         (TARGET_PIPELINE_N10): Likewise.
17903         (TARGET_PIPELINE_N13): Likewise.
17904         (TARGET_PIPELINE_GRAYWOLF): Likewise.
17906 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
17908         * config/nds32/nds32-fpu.md: Update copyright year.
17910 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17912         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
17914 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17916         * config/nds32/nds32.c
17917         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
17918         * config/nds32/nds32.opt (minline-asm-r15): New option.
17920 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17922         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
17923         MASK_HW_ABS.
17924         * config/nds32/nds32.md (abssi2): New pattern.
17926 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
17928         * config/i386/i386.md (rex64namesuffix): New mode attribute.
17929         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
17930         Merge insn pattern from sse_cvtsi2ss<round_name> and
17931         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
17932         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
17933         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
17934         using SWI48 mode iterator.
17935         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
17936         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
17937         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
17938         pattern from sse_cvttss2si<round_saeonly_name>
17939         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
17940         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
17941         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
17942         using SWI48 mode iterator.
17943         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
17944         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
17945         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
17946         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
17947         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
17948         using SWI48 mode iterator.
17949         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
17950         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
17951         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
17952         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
17953         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
17954         SWI48 mode iterator.
17955         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
17956         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
17957         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
17958         pattern from sse_cvttsd2si<round_saeonly_name>
17959         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
17961 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17963         * config/nds32/nds32-md-auxiliary.c
17964         (nds32_valid_smw_lwm_base_p): Refine.
17965         (nds32_output_smw_single_word): Refine.
17966         (nds32_output_smw_double_word): New.
17967         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
17969 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17971         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
17972         (nds32_output_stack_pop): Refine.
17973         (nds32_expand_unaligned_load): Refine.
17974         (nds32_expand_unaligned_store): Refine.
17976 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
17977             Chung-Ju Wu  <jasonwucj@gmail.com>
17979         * config/nds32/constants.md: Add TP_REGNUM constant.
17980         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
17981         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
17982         UNSPEC_ADD32.
17983         * config/nds32/nds32-doubleword.md: Consider flag_pic.
17984         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
17985         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
17986         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
17987         and PIC code generation.
17988         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
17989         code generation.
17990         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
17991         optimization.
17992         * config/nds32/nds32.md: Support TLS and PIC.
17993         * config/nds32/nds32.c: Support TLS and PIC.
17994         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
17995         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
17996         predicate.
17998 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18000         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18001         mode with E_ prefix.
18003 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18004             Chung-Ju Wu  <jasonwucj@gmail.com>
18006         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18007         * config/nds32/nds32-md-auxiliary.c
18008         (symbolic_reference_mentioned_p): New.
18009         (nds32_legitimize_ict_address): New.
18010         (nds32_expand_ict_move): New.
18011         (nds32_indirect_call_referenced_p): New.
18012         (nds32_symbol_binds_local_p): Delete.
18013         (nds32_long_call_p): Modify.
18014         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18015         * config/nds32/nds32-protos.h
18016         (symbolic_reference_mentioned_p): Declare.
18017         (nds32_legitimize_ict_address): Declare.
18018         (nds32_expand_ict_move): Declare.
18019         (nds32_indirect_call_referenced_p): Declare.
18020         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18021         (nds32_relax_group): Use nds32_ict_const_p as condition.
18022         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18023         (nds32_asm_file_start): Output ict_model directive in asm code.
18024         (nds32_legitimate_address_p): Consider indirect call.
18025         (nds32_print_operand): Consider indirect call.
18026         (nds32_print_operand_address): Consider indirect call.
18027         (nds32_insert_attributes): Handle "indirect_call" attribute.
18028         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18029         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18030         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18031         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18032         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18033         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18034         (TARGET_ICT_MODEL_SMALL): Define.
18035         (TARGET_ICT_MODEL_LARGE): Define.
18036         * config/nds32/nds32.md (movsi): Consider ict model.
18037         (call, call_value): Consider ict model.
18038         (sibcall, sibcall_value): Consider ict model.
18039         * config/nds32/nds32.opt (mict-model): New option.
18040         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18041         model.
18043 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18044             Monk Chiang  <sh.chiang04@gmail.com>
18045             Jim Wilson <jimw@sifive.com>
18047         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18048         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18049         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18050         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18051         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
18052         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18053         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18054         compute save_libcall_adjustment properly.
18055         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18056         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18057         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18058         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18059         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18060         (ABI_SPEC): Handle mabi=ilp32e.
18061         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18062         (RVE): Add RVE mask.
18063         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18064         <-march>: Add rv32e as an example.
18066 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
18068         PR c++/82899
18069         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18070         (intra_create_variable_infos): Handle C++ constructors.
18072 2018-05-18  Martin Liska  <mliska@suse.cz>
18074         * passes.def: Remove a redundant pass.
18076 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
18078         PR bootstrap/85838
18079         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18081 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18083         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18084         (ARMv4): Update.
18085         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18086         (ARMv6m): Update.
18087         (armv2, armv2a, armv3, armv3m): Delete architectures.
18088         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18089         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18090         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18091         Delete cpus.
18092         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18093         (*mulsidi3adddi): Likewise.
18094         (mulsidi3): Likewise.
18095         (*mulsidi3_nov6): Likewise.
18096         (umulsidi3): Likewise.
18097         (umulsidi3_nov6): Likewise.
18098         (umaddsidi4): Likewise.
18099         (*umulsidi3adddi): Likewise.
18100         (smulsi3_highpart): Likewise.
18101         (*smulsi3_highpart_nov6): Likewise.
18102         (umulsi3_highpart): Likewise.
18103         (*umulsi3_highpart_nov6): Likewise.
18104         * config/arm/arm.h (arm_arch3m): Delete.
18105         * config/arm/arm.c (arm_arch3m): Delete.
18106         (arm_option_override_internal): Update armv3-related comment.
18107         (arm_configure_build_target): Delete use of isa_bit_mode32.
18108         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18109         (arm_rtx_costs_internal): Delete check of arm_arch3m.
18110         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18111         (mulsa3): Likewise.
18112         (mulusa3): Likewise.
18113         * config/arm/arm-protos.h (arm_arch3m): Delete.
18114         * config/arm/arm-tables.opt: Regenerate.
18115         * config/arm/arm-tune.md: Likewise.
18116         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18117         deleted architectures.
18119 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18121         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18122         (armv5t, armv5te): New features.
18123         (ARMv5, ARMv5e): Delete fgroups.
18124         (ARMv5t, ARMv5te): Adjust for above changes.
18125         (ARMv6m): Likewise.
18126         (armv5, armv5e): Delete arches.
18127         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18128         arm_arch5.
18129         (*call_reg_arm): Likewise.
18130         (*call_value_reg_armv5): Likewise.
18131         (*call_value_reg_arm): Likewise.
18132         (*call_symbol): Likewise.
18133         (*call_value_symbol): Likewise.
18134         (*sibcall_insn): Likewise.
18135         (*sibcall_value_insn): Likewise.
18136         (clzsi2): Likewise.
18137         (prefetch): Likewise.
18138         (define_split and define_peephole2 dependent on arm_arch5):
18139         Likewise.
18140         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18141         arm_arch5e.
18142         (TARGET_ARM_QBIT): Likewise.
18143         (TARGET_DSP_MULTIPLY): Likewise.
18144         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18145         (arm_arch5, arm_arch5e): Delete.
18146         (arm_arch5t, arm_arch5te): Declare.
18147         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18148         (arm_arch5t): Declare.
18149         (arm_option_reconfigure_globals): Update for the above.
18150         (arm_options_perform_arch_sanity_checks): Update comment, replace
18151         use of arm_arch5 with arm_arch5t.
18152         (use_return_insn): Likewise.
18153         (arm_emit_call_insn): Likewise.
18154         (output_return_instruction): Likewise.
18155         (arm_final_prescan_insn): Likewise.
18156         (arm_coproc_builtin_available): Likewise.
18157         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18158         arm_arch5e with arm_arch5t and arm_arch5te.
18159         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18160         (arm_arch5t, arm_arch5te): Declare.
18161         * config/arm/arm-tables.opt: Regenerate.
18162         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18163         * config/arm/t-multilib: Likewise.
18164         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18165         instead of arm_arch5.
18166         (*call_reg_thumb1): Likewise.
18167         (*call_value_reg_thumb1_v5): Likewise.
18168         (*call_value_reg_thumb1): Likewise.
18169         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18170         unreachable path.
18171         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18173 2018-05-18  Martin Liska  <mliska@suse.cz>
18175         PR gcov-profile/84846
18176         * doc/gcov.texi: Document -t option of gcov tool.
18178 2018-05-18  Martin Liska  <mliska@suse.cz>
18180         PR gcov-profile/84846
18181         * gcov.c (print_usage): Add new -t option.
18182         (process_args): Handle the option.
18183         (generate_results): Use stdout as output when requested by
18184         the option.
18186 2018-05-18  Martin Liska  <mliska@suse.cz>
18188         PR gcov-profile/84846
18189         * coverage.c (coverage_init): Write PWD to .gcno file.
18190         * doc/gcov.texi: Document how working directory is printed.
18191         * gcov-dump.c (dump_gcov_file): Print PWD.
18192         * gcov.c (output_intermediate_file): Likewise.
18193         (read_graph_file): Read PWD string.
18194         (output_lines): Print PWD.
18196 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18198         PR middle-end/85817
18199         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18200         for retval and return false if all args to phi are zero.
18202 2018-05-18  Richard Biener  <rguenther@suse.de>
18204         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18205         method.
18206         (evrp_dom_walker::before_dom_children): Call it.
18208 2018-05-18  Richard Biener  <rguenther@suse.de>
18210         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18211         results when processing array refs with variable index.
18213 2018-05-18  Toon Moene  <toon@moene.org>
18215         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18216         directly after that of -floop-interchange. Indicate that both
18217         options are enabled by default when specifying -O3.
18219 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18221         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18222         iterator.  Delete separate integer-mode vec_set<mode> expander.
18223         (aarch64_simd_vec_setv2di): Delete.
18224         (vec_setv2di): Delete.
18225         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18226         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
18227         the "w, r" alternative.
18229 2018-05-18  Martin Liska  <mliska@suse.cz>
18231         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18232         * tree-pass.h (make_pass_lower_switch_O0): New function.
18233         * tree-switch-conversion.c (node_has_low_bound): Remove.
18234         (node_has_high_bound): Likewise.
18235         (node_is_bounded): Likewise.
18236         (class pass_lower_switch): Make it a template type and create
18237         two instances.
18238         (pass_lower_switch::execute): Add template argument.
18239         (make_pass_lower_switch): New function.
18240         (make_pass_lower_switch_O0): New function.
18241         (do_jump_if_equal): Remove.
18242         (emit_case_nodes): Simplify to just handle all 3 cases and leave
18243         all the hard work to tree optimization passes.
18245 2018-05-18  Martin Liska  <mliska@suse.cz>
18247         * dbgcnt.c (limit_low): Renamed from limit.
18248         (limit_high): New variable.
18249         (dbg_cnt_is_enabled): Check for upper limit.
18250         (dbg_cnt): Adjust dumping.
18251         (dbg_cnt_set_limit_by_index): Add new argument for high
18252         value.
18253         (dbg_cnt_set_limit_by_name): Likewise.
18254         (dbg_cnt_process_single_pair): Parse new format.
18255         (dbg_cnt_process_opt): Use strtok.
18256         (dbg_cnt_list_all_counters): Remove 'value' and add
18257         'limit_high'.
18258         * doc/invoke.texi: Document changes.
18260 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
18262         * doc/sourcebuild.texi (scalar_all_fma): Document.
18263         * tree.def (FMA_EXPR): Delete.
18264         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18265         * internal-fn.c (ternary_direct): New macro.
18266         (expand_ternary_optab_fn): Likewise.
18267         (direct_ternary_optab_supported_p): Likewise.
18268         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18269         * builtins.c (fold_builtin_fma): Delete.
18270         (fold_builtin_3): Don't call it.
18271         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18272         * expr.c (expand_expr_real_2): Likewise.
18273         * fold-const.c (operand_equal_p): Likewise.
18274         (fold_ternary_loc): Likewise.
18275         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18276         * gimple.c (DEFTREECODE): Likewise.
18277         * gimplify.c (gimplify_expr): Likewise.
18278         * optabs-tree.c (optab_for_tree_code): Likewise.
18279         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18280         * tree-eh.c (operation_could_trap_p): Likewise.
18281         (stmt_could_throw_1_p): Likewise.
18282         * tree-inline.c (estimate_operator_cost): Likewise.
18283         * tree-pretty-print.c (dump_generic_node): Likewise.
18284         (op_code_prio): Likewise.
18285         * tree-ssa-loop-im.c (stmt_cost): Likewise.
18286         * tree-ssa-operands.c (get_expr_operands): Likewise.
18287         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18288         * fold-const-call.h (fold_fma): Delete.
18289         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18290         CFN_FNMA and CFN_FNMS.
18291         (fold_fma): Delete.
18292         * genmatch.c (combined_fn): New enum.
18293         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18294         (commutative_op): New function.
18295         (commutate): Use it.  Handle more than 2 operands.
18296         (dt_operand::gen_gimple_expr): Use commutative_op.
18297         (parser::parse_expr): Allow :c to be used with non-binary
18298         operators if the commutative operand is known.
18299         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18300         CFN_FMS, CFN_FNMA and CFN_FNMS.
18301         (backprop::process_assign_use): Remove FMA_EXPR handling.
18302         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18303         (gen_hsa_fma): New function.
18304         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18305         IFN_FNMA and IFN_FNMS.
18306         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18307         * gimple-fold.h (follow_all_ssa_edges): Declare.
18308         * gimple-fold.c (follow_all_ssa_edges): New function.
18309         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18310         gimple_build interface and use follow_all_ssa_edges to fold the result.
18311         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18312         instead of checking for optabs directly.
18313         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
18314         rather than FMA_EXPRs.
18315         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
18316         call to IFN_FMA instead of an FMA_EXPR.
18318 2018-05-17  Jim Wilson  <jimw@sifive.com>
18320         * expr.c (do_tablejump): When converting index to Pmode, if we have a
18321         sign extended promoted subreg, and the range does not have the sign bit
18322         set, then do a sign extend.
18324         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
18325         test, check for sign extended subreg and/or constant operands, and
18326         do a sign extend in that case.
18328 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
18330         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
18331         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
18332         Add untyped.
18333         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
18334         Change logics_shift_reg to logics_shift_imm.
18335         (thunderx2t99_fp_loadpair_basic): Delete.
18336         (thunderx2t99_fp_storepair_basic): Delete.
18337         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
18338         (thunderx2t99_asimd_polynomial): Delete.
18339         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
18340         and neon_fp_mul_d_scalar_q.
18341         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
18342         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
18343         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
18344         (thunderx2t99_asimd_lut): Add missing tbl types.
18345         (thunderx2t99_asimd_ext): Delete.
18346         (thunderx2t99_asimd_load1_1_mult): Delete.
18347         (thunderx2t99_asimd_load1_2_mult): Delete.
18348         (thunderx2t99_asimd_load1_ldp): New.
18349         (thunderx2t99_asimd_load1): New.
18350         (thunderx2t99_asimd_load2): Add missing *load2* types.
18351         (thunderx2t99_asimd_load3): New.
18352         (thunderx2t99_asimd_load4): New.
18353         (thunderx2t99_asimd_store1_1_mult): Delete.
18354         (thunderx2t99_asimd_store1_2_mult): Delete.
18355         (thunderx2t99_asimd_store2_mult): Delete.
18356         (thunderx2t99_asimd_store2_onelane): Delete.
18357         (thunderx2t99_asimd_store_stp): New.
18358         (thunderx2t99_asimd_store1): New.
18359         (thunderx2t99_asimd_store2): New.
18360         (thunderx2t99_asimd_store3): New.
18361         (thunderx2t99_asimd_store4): New.
18363 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
18365         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
18366         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
18368 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
18369             Segher Boessenkool  <segher@kernel.crashing.org>
18371         PR target/85698
18372         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
18373         operand.
18375 2018-05-17  Richard Biener  <rguenther@suse.de>
18377         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
18378         for pruning loop and prune defs feeding only already visited PHIs.
18380 2018-05-17  Richard Biener  <rguenther@suse.de>
18382         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
18384 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
18385             Richard Biener  <rguenther@suse.de>
18387         PR tree-optimization/85793
18388         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
18389         for VMAT_ELEMENTWISE.
18391 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18393         * internal-fn.h (lookup_internal_fn): Declare
18394         * internal-fn.c (lookup_internal_fn): New function.
18395         * gimple.c (gimple_build_call_from_tree): Handle calls to
18396         internal functions.
18397         * gimple-pretty-print.c (dump_gimple_call): Print "." before
18398         internal function names.
18399         * tree-pretty-print.c (dump_generic_node): Likewise.
18400         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
18402 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18404         * gimple-fold.h (gimple_build): Make the function forms take
18405         combined_fn rather than built_in_function.
18406         (gimple_simplify): Likewise.
18407         * gimple-match-head.c (gimple_simplify): Likewise.
18408         * gimple-fold.c (gimple_build): Likewise.
18409         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
18410         rather than gimple_build_call_internal.
18411         (get_initial_defs_for_reduction): Likewise.
18412         (vect_create_epilog_for_reduction): Likewise.
18413         (vectorizable_live_operation): Likewise.
18415 2018-05-17  Martin Liska  <mliska@suse.cz>
18417         * gimple-ssa-sprintf.c (format_directive): Do not use
18418         space in between 'G_' and '('.
18420 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
18422         PR target/85323
18423         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
18424         even if the mask is not all ones.
18426         PR target/85323
18427         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
18428         vector.
18429         (ix86_gimple_fold_builtin): Likewise.
18431         PR target/85323
18432         * config/i386/i386.c: Include tree-vector-builder.h.
18433         (ix86_vector_shift_count): New function.
18434         (ix86_fold_builtin): Fold shift builtins by scalar count.
18435         (ix86_gimple_fold_builtin): Likewise.
18437         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
18438         _mm512_setzero): New intrinsics.
18440 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
18441             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18443         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
18444         code generation for cases where splatting a value is not useful.
18445         * simplify-rtx.c (simplify_ternary_operation): Simplify
18446         vec_merge across a vec_duplicate and a paradoxical subreg forming
18447         a vector mode to a vec_concat.
18449 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
18451         * config.gcc: Support "goldmont-plus".
18452         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18453         "goldmont-plus".
18454         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18455         PROCESSOR_GOLDMONT_PLUS.
18456         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
18457         (processor_target_table): Add "goldmont-plus".
18458         (PTA_GOLDMONT_PLUS): Define.
18459         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
18460         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
18461         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
18462         (fold_builtin_cpu): Add "goldmont-plus".
18463         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
18464         (ix86_option_override_internal): Add "goldmont-plus".
18465         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
18466         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
18467         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
18468         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
18470 2018-05-17  Richard Biener  <rguenther@suse.de>
18472         PR tree-optimization/85757
18473         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
18474         remove defs that only feed that PHI from further processing.
18476 2018-05-16  Jim Wilson  <jimw@sifive.com>
18478         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
18479         asterisk to name.
18480         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
18482 2018-05-16  Mark Wielaard  <mark@klomp.org>
18484         * dwarf2out.c (count_index_strings): New function.
18485         (output_indirect_strings): Call count_index_strings and generate
18486         header for dwarf_version >= 5.
18488 2018-05-16  Mark Wielaard  <mark@klomp.org>
18490         * dwarf2out.c (dwarf_FORM): New function.
18491         (set_indirect_string): Use dwarf_FORM.
18492         (reset_indirect_string): Likewise.
18493         (size_of_die): Likewise.
18494         (value_format): Likewise.
18495         (output_die): Likewise.
18496         (add_skeleton_AT_string): Likewise.
18497         (output_macinfo_op): Likewise.
18498         (index_string): Likewise.
18499         (output_index_string_offset): Likewise.
18500         (output_index_string): Likewise.
18501         (count_index_strings): Likewise.
18503 2018-05-16  Carl Love  <cel@us.ibm.com>
18505         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18506         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
18508 2018-05-16  Martin Jambor  <mjambor@suse.cz>
18510         * ipa-prop.c (ipa_free_all_edge_args): Remove.
18511         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
18513 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
18515         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
18516         (fnma<mode>4): Likewise.
18517         (fms<mode>4): Likewise.
18518         (fnms<mode>4): Likewise.
18519         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
18520         (aarch64_fnma<mode>4): Likewise.
18521         (aarch64_fms<mode>4): Likewise.
18522         (aarch64_fnms<mode>4): Likewise.
18523         (aarch64_fnmadd<mode>4): Likewise.
18525 2018-05-16  Jason Merrill  <jason@redhat.com>
18527         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
18529 2018-05-16  Richard Biener  <rguenther@suse.de>
18531         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
18532         (dump_stmt_cost): Declare.
18533         (add_stmt_cost): Dump cost we add.
18534         (add_stmt_costs): New function.
18535         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
18536         No longer exported.
18537         (vect_analyze_stmt): Adjust prototype.
18538         (vectorizable_condition): Likewise.
18539         (vectorizable_live_operation): Likewise.
18540         (vectorizable_reduction): Likewise.
18541         (vectorizable_induction): Likewise.
18542         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
18543         cost vector to pass to vectorizable_ and record afterwards.
18544         (vect_model_reduction_cost): Take cost vector argument and adjust.
18545         (vect_model_induction_cost): Likewise.
18546         (vectorizable_reduction): Likewise.
18547         (vectorizable_induction): Likewise.
18548         (vectorizable_live_operation): Likewise.
18549         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
18550         SLP_TREE_NUMBER_OF_VEC_STMTS.
18551         (vect_analyze_slp_cost_1): Remove.
18552         (vect_analyze_slp_cost): Likewise.
18553         (vect_slp_analyze_node_operations): Take visited args and
18554         a target cost vector.  Avoid processing already visited stmt sets.
18555         (vect_slp_analyze_operations): Use a local cost vector to gather
18556         costs and register those of non-discarded instances.
18557         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
18558         (vect_schedule_slp_instance): Remove copying of
18559         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
18560         zero.
18561         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
18562         adding cost.  Record cost entry location.
18563         (vect_prologue_cost_for_slp_op): Function to compute cost of
18564         a constant or invariant generated for SLP vect in the prologue,
18565         split out from vect_analyze_slp_cost_1.
18566         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
18567         (vect_model_promotion_demotion_cost): Likewise.
18568         (vect_model_store_cost): Likewise, make static.
18569         (vect_model_load_cost): Likewise.
18570         (vectorizable_bswap): Add cost vector arg and adjust.
18571         (vectorizable_call): Likewise.
18572         (vectorizable_simd_clone_call): Likewise.
18573         (vectorizable_conversion): Likewise.
18574         (vectorizable_assignment): Likewise.
18575         (vectorizable_shift): Likewise.
18576         (vectorizable_operation): Likewise.
18577         (vectorizable_store): Likewise.
18578         (vectorizable_load): Likewise.
18579         (vectorizable_condition): Likewise.
18580         (vectorizable_comparison): Likewise.
18581         (can_vectorize_live_stmts): Likewise.
18582         (vect_analyze_stmt): Likewise.
18583         (vect_transform_stmt): Adjust calls to vectorizable_*.
18584         * tree-vectorizer.c: Include gimple-pretty-print.h.
18585         (dump_stmt_cost): New function.
18587 2018-05-16  Richard Biener  <rguenther@suse.de>
18589         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
18590         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
18591         * tree-ssa-dse.c: Include tree-ssa-loop.h.
18592         (check_name): New callback.
18593         (dse_classify_store): Track cycles via a visited bitmap of PHI
18594         defs and simplify handling of in-loop and across loop dead stores
18595         and properly fail for loop-variant refs.  Handle byte-tracking with
18596         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
18597         limiting the walk.
18599 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
18601         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
18602         (vect_get_mask_type_for_stmt): Likewise.
18603         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
18604         split out from...
18605         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
18606         to determine the statement's vector type and the vector type that
18607         should be used for calculating nunits.  Deal with cases in which
18608         the type has to be deferred.
18609         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
18610         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
18611         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
18612         (vect_determine_vf_for_stmt): New functions, split out from...
18613         (vect_determine_vectorization_factor): ...here.
18614         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
18615         (vect_get_mask_type_for_stmt): New functions, split out from
18616         vect_determine_vectorization_factor.
18618 2018-05-16  Richard Biener  <rguenther@suse.de>
18620         * tree-cfg.c (verify_gimple_assign_ternary): Properly
18621         verify the [VEC_]COND_EXPR embedded comparison.
18623 2018-05-15  Martin Sebor  <msebor@redhat.com>
18625         PR tree-optimization/85753
18626         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
18627         RECORD_TYPE in addition to ARRAY_TYPE.
18629 2018-05-15  Martin Sebor  <msebor@redhat.com>
18631         PR middle-end/85643
18632         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
18634 2018-05-15  Richard Biener  <rguenther@suse.de>
18636         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
18637         add by_clobber_p one.  Change algorithm to collect all defs
18638         representing uses we need to walk and try reducing them to
18639         a single one before failing.
18640         (dse_dom_walker::dse_optimize_stmt): Adjust.
18642 2018-05-13  Mark Wielaard  <mark@klomp.org>
18644         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
18645         (size_of_loc_descr): Likewise.
18646         (output_loc_operands): Likewise.
18647         (output_loc_operands_raw): Likewise.
18648         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
18649         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
18650         (hash_loc_operands): Likewise.
18651         (compare_loc_operands): Likewise.
18653 2018-05-14  Mark Wielaard  <mark@klomp.org>
18655         * dwarf2out.c (count_index_addrs): New function.
18656         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
18658 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18660         PR tree-optimization/83648
18661         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
18662         return value as malloc candidate.
18664 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18666         PR ipa/85734
18667         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
18668         param as true in call to suggest_attribute.
18670 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
18672         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
18673         -mreadonly-in-sdata.
18675 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18677         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
18678         New pattern.
18679         (aarch64_crypto_aesd_fused): Likewise.
18681 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
18683         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
18684         (movsi_aarch64): Likewise.
18685         (load_pairsi): Likewise.
18686         (load_pairdi): Likewise.
18687         (store_pairsi): Likewise.
18688         (store_pairdi): Likewise.
18689         (load_pairsf): Likewise.
18690         (load_pairdf): Likewise.
18691         (store_pairsf): Likewise.
18692         (store_pairdf): Likewise.
18693         (zero_extend): Likewise.
18694         (trunc): Swap alternatives.
18695         (fcvt_target): Add '?' to prefer w over r.
18697 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
18699         PR target/85756
18700         * config/i386/i386.md: Disallow non-commutative arithmetics in
18701         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
18702         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
18703         in the peephole2 before it.
18705 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
18707         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
18708         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
18709         (ix86_handle_option): Handle -mcldemote.
18710         * config.gcc: New header.
18711         * config/i386/cldemoteintrin.h: New file.
18712         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
18713         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18714         -mcldemote.
18715         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18716         OPTION_MASK_ISA_CLDEMOTE.
18717         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
18718         (ix86_valid_target_attribute_inner_p): Ditto.
18719         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
18720         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
18721         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
18722         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
18723         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
18724         (cldemote): New.
18725         * config/i386/i386.opt: Add -mcldemote.
18726         * config/i386/x86intrin.h: New header.
18727         * doc/invoke.texi: Add -mcldemote.
18729 2018-05-14  Richard Biener  <rguenther@suse.de>
18731         * doc/match-and-simplify.texi: Adjust :s documentation.
18733 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
18735         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
18736         intended memcpy size.
18737         (REORDER_45): Likewise.
18739 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
18741         * sort.cc: New file.
18742         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
18743         * vec.c (qsort_chk): Use gcc_qsort.
18744         * Makefile.in (OBJS-libcommon): Add sort.o.
18745         (build/sort.o): New target.  Use it...
18746         (BUILD_RTL): ... here, and...
18747         (build/gencfn-macros): ... here, and...
18748         (build/genmatch): ... here.
18750 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18751             Chung-Ju Wu  <jasonwucj@gmail.com>
18753         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
18754         * config/nds32/nds32-graywolf.md: New file.
18755         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
18756         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
18757         pipeline.
18758         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
18759         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
18760         * config/nds32/nds32.md (pipeline_model): Add graywolf.
18761         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
18762         * config/nds32/pipelines.md: Include n15 settings.
18764 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18765             Chung-Ju Wu  <jasonwucj@gmail.com>
18767         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
18768         * config/nds32/nds32-n13.md: New file.
18769         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
18770         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
18771         pipeline.
18772         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
18773         * config/nds32/nds32.md (pipeline_model): Add n13.
18774         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
18775         * config/nds32/pipelines.md: Include n13 settings.
18777 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18778             Chung-Ju Wu  <jasonwucj@gmail.com>
18780         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
18781         * config/nds32/nds32-n10.md: New file.
18782         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
18783         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
18784         pipeline.
18785         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
18786         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
18787         * config/nds32/nds32.md (pipeline_model): Add n10.
18788         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
18789         * config/nds32/pipelines.md: Include n10 settings.
18791 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
18792             Kito Cheng  <kito.cheng@gmail.com>
18793             Chung-Ju Wu  <jasonwucj@gmail.com>
18795         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
18796         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18797         Add enum values for DSP extension instructions.
18798         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
18799         New constraints.
18800         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
18801         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
18802         New code iterators.
18803         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
18804         * config/nds32/nds32-dspext.md: New file for DSP implementation.
18805         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
18806         * config/nds32/nds32-intrinsic.md: Likewise.
18807         * config/nds32/nds32_intrinsic.h: Likewise.
18808         * config/nds32/nds32-md-auxiliary.c: Likewise.
18809         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
18810         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
18811         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
18812         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
18813         * config/nds32/nds32-protos.h: New declarations for DSP extension.
18814         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
18815         TYPE_DMAC in switch statement.
18816         * config/nds32/nds32.c: New checking and implementation for DSP
18817         extension instructions.
18818         * config/nds32/nds32.h: Likewise.
18819         * config/nds32/nds32.md: Likewise.
18820         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
18821         * config/nds32/predicates.md: Implement new predicates for DSP
18822         extension.
18824 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
18826         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
18827         Reformat alternatives and attributes so it is easier to identify
18828         which constraints/attributes go with which instruction.
18829         (mov<mode>_hardfloat32, FMOVE64): Likewise.
18830         (mov<mode>_softfloat32, FMOVE64): Likewise.
18831         (mov<mode>_hardfloat64, FMOVE64): Likewise.
18832         (mov<mode>_softfloat64, FMOVE64): Likewise.
18834 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18836         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18837         subsection.
18838         (Basic PowerPC Built-in Functions): The new name of the
18839         subsection previously known as "PowerPC Built-in Functions".
18840         (Basic PowerPC Built-in Functions Available on all Configurations):
18841         New subsubsection.
18842         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
18843         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
18844         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
18845         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
18847 2018-05-11  Martin Jambor  <mjambor@suse.cz>
18849         PR ipa/85655
18850         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
18851         single const.
18853 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
18855         PR target/85733
18856         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
18858 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
18860         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
18861         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
18862         (ix86_handle_option): Handle -mwaitpkg.
18863         * config.gcc: New header.
18864         * config/i386/cpuid.h (bit_WAITPKG): New bit.
18865         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
18866         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
18867         function type.
18868         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18869         OPTION_MASK_ISA_WAITPKG.
18870         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
18871         (ix86_option_override_internal): Add PTA_WAITPKG.
18872         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
18873         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
18874         IX86_BUILTIN_TPAUSE.
18875         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
18876         __builtin_ia32_umwait and __builtin_ia32_tpause.
18877         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
18878         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
18879         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
18880         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
18881         UNSPECV_TPAUSE): New.
18882         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
18883         * config/i386/i386.opt: Add -mwaitpkg.
18884         * config/i386/waitpkgintrin.h: New file.
18885         * config/i386/x86intrin.h: New header.
18886         * doc/invoke.texi: Add -mwaitpkg.
18888 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
18890         PR target/85606
18891         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
18892         equivalent.
18893         (cortex-m0): Use armv6s-m isa.
18894         (cortex-m0plus): Likewise.
18895         (cortex-m1): Likewise.
18896         (cortex-m0.small-multiply): Likewise.
18897         (cortex-m0plus.small-multiply): Likewise.
18898         (cortex-m1.small-multiply): Likewise.
18900 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
18901             Jakub Jelinek  <jakub@redhat.com>
18903         PR tree-optimization/85692
18904         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
18905         source permute as well.
18907 2018-05-11  Martin Liska  <mliska@suse.cz>
18909         PR sanitizer/85556
18910         * doc/extend.texi: Document LLVM style format for no_sanitize
18911         attribute.
18913 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
18915         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
18916         mode_supports_vsx_dform_quad to mode_supports_dq_form.
18917         (mode_supports_vsx_dform_quad): Likewise.
18918         (mode_supports_vmx_dform): Move these functions to be next to the
18919         other mode_supports functions.
18920         (mode_supports_dq_form): Likewise.
18921         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
18922         mode_supports_dq_form.
18923         (reg_offset_addressing_ok_p): Likewise.
18924         (offsettable_ok_by_alignment): Likewise.
18925         (rs6000_legitimate_offset_address_p): Likewise.
18926         (legitimate_lo_sum_address_p): Likewise.
18927         (rs6000_legitimize_address): Likewise.
18928         (rs6000_legitimize_reload_address): Likewise.
18929         (rs6000_secondary_reload_inner): Likewise.
18930         (rs6000_preferred_reload_class): Likewise.
18931         (rs6000_output_move_128bit): Likewise.
18933 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
18935         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
18936         Generate SImode target register for null target.
18937         <case IX86_BUILTIN_XGETBV>: Ditto.
18938         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
18939         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
18941 2018-05-10  Carl Love  <cel@us.ibm.com>
18943         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18944         dcbtt and dcbtstt if operands[2] is 0.
18946 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
18948         PR target/85693
18949         * config/i386/sse.md (usadv64qi): New expander.
18951 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
18953         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
18954         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
18955         -maltivec=be support.
18956         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
18957         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
18958         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
18959         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
18960         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
18961         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
18962         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
18963         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
18964         altivec_vsumsws): Adjust.
18965         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
18966         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
18967         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
18968         support.
18969         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
18970         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
18971         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
18972         (altivec_lve<VI_char>x): Delete expand.
18973         (*altivec_lve<VI_char>x_internal): Rename to...
18974         (altivec_lve<VI_char>x): ... this.
18975         (altivec_lvxl_<mode>): Delete expand.
18976         (*altivec_lvxl_<mode>_internal): Rename to ...
18977         (altivec_lvxl_<mode>): ... this.
18978         (altivec_stvxl_<mode>): Delete expand.
18979         (*altivec_stvxl_<mode>_internal): Rename to ...
18980         (altivec_stvxl_<mode>): ... this.
18981         (altivec_stve<VI_char>x): Delete expand.
18982         (*altivec_stve<VI_char>x_internal): Rename to ...
18983         (altivec_stve<VI_char>x): ... this.
18984         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
18985         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
18986         reduc_plus_scal_<mode>): Adjust.
18987         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
18988         comment.
18989         (rs6000_cpu_cpp_builtins): Adjust.
18990         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
18991         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
18992         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
18993         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
18994         -maltivec=be support.
18995         (rs6000_split_vec_extract_var): Adjust.
18996         (rs6000_split_v4si_init): Adjust.
18997         (swap_selector_for_mode): Delete.
18998         (altivec_expand_lvx_be, altivec_expand_stvx_be,
18999         altivec_expand_stvex_be): Delete.
19000         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19001         -maltivec=be support.
19002         (rs6000_gimple_fold_builtin): Ditto.
19003         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19004         Adjust.
19005         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19006         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19007         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19008         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19009         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19010         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19011         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19012         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19013         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19014         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19015         anonymous split): Adjust.
19016         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19017         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19019 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19021         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19022         when --with-gxx-include-dir is also specified.
19023         * configure: Regenerate.
19025 2018-05-09  Jim Wilson  <jimw@sifive.com>
19027         PR target/84797
19028         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19029         * config/riscv/t-withmultilib: New.
19030         * config/riscv/withmultilib.h: New.
19031         * doc/install.texi: Document RISC-V --with-multilib-list support.
19033 2018-05-09  Richard Biener  <rguenther@suse.de>
19035         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19036         vector.
19037         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19038         actual scalar cost using the cost vector and the add_stmt_cost
19039         machinery.
19041 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19043         PR rtl-optimization/85645
19044         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19045         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19047 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19049         PR rtl-optimization/85645
19050         * shrink-wrap.c (spread_components): Return a boolean saying if
19051         anything was changed.
19052         (try_shrink_wrapping_separate): Iterate spread_components until
19053         nothing changes anymore.
19055 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19057         PR rtl-optimization/85645
19058         * regrename.c (build_def_use): Also kill the chains that include the
19059         destination of a REG_CFA_REGISTER note.
19061 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19063         PR rtl-optimization/85645
19064         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19065         insn that has a REG_CFA_REGISTER note.
19067 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
19069         * cfgexpand.c (expand_clobber): New function.
19070         (expand_gimple_stmt_1): Use it.
19071         * tree-vect-stmts.c (vect_clobber_variable): New function,
19072         split out from...
19073         (vectorizable_simd_clone_call): ...here.
19074         (vectorizable_store): Emit a clobber either side of an
19075         IFN_STORE_LANES sequence.
19076         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19078 2018-05-09  Tom de Vries  <tom@codesourcery.com>
19080         PR target/85626
19081         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19082         (define_insn "trap_if_false"): Add exit after trap.
19084 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
19086         PR rtl-optimization/85638
19087         * bb-reorder.c: Include common/common-target.h.
19088         (create_forwarder_block): New function extracted from...
19089         (fix_up_crossing_landing_pad): ...here.  Rename into...
19090         (dw2_fix_up_crossing_landing_pad): ...this.
19091         (sjlj_fix_up_crossing_landing_pad): New function.
19092         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19093         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19094         from both partitions and exit the loop after one iteration.
19096 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19098         Revert:
19099         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19100         subsection.
19101         (Basic PowerPC Built-in Functions): The new name of the
19102         subsection previously known as "PowerPC Built-in Functions".
19103         (Basic PowerPC Built-in Functions Available on all Configurations):
19104         New subsubsection.
19105         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19106         subsubsection.
19107         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19108         subsubsection.
19109         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19110         subsubsection.
19111         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19112         subsubsection.
19114 2018-05-08  Jim Wilson  <jimw@sifive.com>
19116         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19117         (LD_EMUL_SUFFIX): New.
19118         (LINK_SPEC): Use it.
19120 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19122         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19123         subsection.
19124         (Basic PowerPC Built-in Functions): The new name of the
19125         subsection previously known as "PowerPC Built-in Functions".
19126         (Basic PowerPC Built-in Functions Available on all Configurations):
19127         New subsubsection.
19128         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19129         subsubsection.
19130         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19131         subsubsection.
19132         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19133         subsubsection.
19134         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19135         subsubsection.
19137 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19139         PR target/85683
19140         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19141         after cmpelim optimization.
19143 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
19145         * config.gcc: Support "goldmont".
19146         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19147         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19148         PROCESSOR_GOLDMONT.
19149         * config/i386/i386.c (m_GOLDMONT): Define.
19150         (processor_target_table): Add "goldmont".
19151         (PTA_GOLDMONT): Define.
19152         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19153         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19154         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19155         (fold_builtin_cpu): Add "goldmont".
19156         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19157         (ix86_option_override_internal): Add "goldmont".
19158         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19159         (processor_type): Add PROCESSOR_GOLDMONT.
19160         * config/i386/i386.md: Add CPU "glm".
19161         * config/i386/glm.md: New file.
19162         * config/i386/x86-tune.def: Add m_GOLDMONT.
19163         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19165 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19167         PR target/85572
19168         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19169         E_V4DImode.
19170         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19171         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
19172         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
19174         PR target/85317
19175         * config/i386/i386.c (ix86_fold_builtin): Handle
19176         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19178         PR target/85480
19179         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19180         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19182 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
19184         PR target/85658
19185         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19186         (check_arch): Likewise.
19187         (check_fpu): Return the result rather than printing it.
19188         (end arch): Fix operator precedence.
19189         (end cpu): Likewise.
19190         (END): Print the result from check_fpu.
19192 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19193             Alan Hayward  <alan.hayward@arm.com>
19194             David Sherwood  <david.sherwood@arm.com>
19196         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19197         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19198         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19199         (*fcmuo<mode>_and): New patterns.
19201 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19203         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19204         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19205         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19206         (cmp_op, sve_imm_con): New code attributes.
19207         (SVE_COND_INT_CMP, imm_con): Delete.
19208         (cmp_op): Remove above unspecs from int attribute.
19209         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19210         to...
19211         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
19212         comparison-specific unspecs.
19213         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19214         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19215         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19216         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19217         (*vec_fcm<cmp_op><mode>): Rename to...
19218         (*fcm<cmp_op><mode>): ...this and adjust likewise.
19219         (*vec_fcmuo<mode>): Rename to...
19220         (*fcmuo<mode>): ...this and adjust likewise.
19221         (*pred_fcm<cmp_op><mode>): New pattern.
19222         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19223         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19224         functions.
19225         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19226         and UNORDERED.
19227         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19228         (aarch64_emit_sve_predicated_cond): New function.
19229         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19230         (aarch64_emit_unspec_cond_or): Replace with...
19231         (aarch64_emit_sve_or_conds): ...this new function.  Use
19232         aarch64_emit_sve_ptrue_op for the individual comparisons and
19233         aarch64_emit_binop to OR them together.
19234         (aarch64_emit_inverted_unspec_cond): Replace with...
19235         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
19236         aarch64_emit_sve_ptrue_op for the comparison and
19237         aarch64_emit_unop to invert the result.
19238         (aarch64_expand_sve_vec_cmp_float): Update after the above
19239         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
19241 2018-05-07  Nathan Sidwell  <nathan@acm.org>
19243         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19244         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19245         (Backwards Compatibility): Likewise.
19247 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19249         PR bootstrap/85681
19250         Revert:
19251         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19253         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19254         <prefetch_dynamic_strides>: New const bool field.
19255         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19256         prefetch_dynamic_strides.
19257         (exynosm1_prefetch_tune): Likewise.
19258         (thunderxt88_prefetch_tune): Likewise.
19259         (thunderx_prefetch_tune): Likewise.
19260         (thunderx2t99_prefetch_tune): Likewise.
19261         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19262         to false.
19263         (aarch64_override_options_internal): Update to set
19264         PARAM_PREFETCH_DYNAMIC_STRIDES.
19265         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19266         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19267         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19268         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19269         prefetch-dynamic-strides setting.
19271         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19273         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19274         <minimum_stride>: New const int field.
19275         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19276         minimum_stride field.
19277         (exynosm1_prefetch_tune): Likewise.
19278         (thunderxt88_prefetch_tune): Likewise.
19279         (thunderx_prefetch_tune): Likewise.
19280         (thunderx2t99_prefetch_tune): Likewise.
19281         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19282         (aarch64_override_options_internal): Update to set
19283         PARAM_PREFETCH_MINIMUM_STRIDE.
19284         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19285         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19286         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19287         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19288         stride is constant and is below the minimum stride threshold.
19290 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19292         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19293         to 512.
19295 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19297         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19298         <prefetch_dynamic_strides>: New const bool field.
19299         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19300         prefetch_dynamic_strides.
19301         (exynosm1_prefetch_tune): Likewise.
19302         (thunderxt88_prefetch_tune): Likewise.
19303         (thunderx_prefetch_tune): Likewise.
19304         (thunderx2t99_prefetch_tune): Likewise.
19305         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19306         to false.
19307         (aarch64_override_options_internal): Update to set
19308         PARAM_PREFETCH_DYNAMIC_STRIDES.
19309         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19310         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19311         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19312         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19313         prefetch-dynamic-strides setting.
19315 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19317         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19318         <minimum_stride>: New const int field.
19319         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19320         minimum_stride field.
19321         (exynosm1_prefetch_tune): Likewise.
19322         (thunderxt88_prefetch_tune): Likewise.
19323         (thunderx_prefetch_tune): Likewise.
19324         (thunderx2t99_prefetch_tune): Likewise.
19325         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19326         (aarch64_override_options_internal): Update to set
19327         PARAM_PREFETCH_MINIMUM_STRIDE.
19328         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19329         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19330         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19331         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19332         stride is constant and is below the minimum stride threshold.
19334 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
19336         PR c++/85659
19337         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
19338         the type is addressable.  Don't force op into register if it has
19339         BLKmode.
19341 2018-05-05  Roland McGrath  <mcgrathr@google.com>
19343         PR other/77609
19344         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
19345         any section for which we don't know a specific type it should have,
19346         regardless of name.  Previously this was done only for the exact
19347         names ".init_array", ".fini_array", and ".preinit_array".
19348         (default_elf_asm_named_section): Add comment about
19349         relationship with default_section_type_flags and SECTION_NOTYPE.
19350         (get_section): Don't consider it a type conflict if one side has
19351         SECTION_NOTYPE and the other doesn't, as long as neither has the
19352         SECTION_BSS et al used in the default_section_type_flags logic.
19354 2018-05-05  Tom de Vries  <tom@codesourcery.com>
19356         PR target/85653
19357         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
19358         (workaround_barsyncs): New function.
19359         (nvptx_reorg): Use workaround_barsyncs.
19360         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19361         (define_expand "nvptx_membar_cta"): New define_expand.
19362         (define_insn "*nvptx_membar_cta"): New insn.
19364 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
19366         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
19367         To improve optimization opportunities.
19368         * builtin-types.def: The new needed builtin types for the above.
19370 2018-05-04  Richard Biener  <rguenther@suse.de>
19372         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
19373         * gimple-ssa-store-merging.c
19374         (imm_store_chain_info::output_merged_store): Remove redundant create,
19375         release split_store vector contents on failure.
19376         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
19377         scalar stmt vector on cache hit.
19379 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
19381         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19382         Xilinx FP support.
19383         * config.gcc (powerpc-xilinx-eabi*): Remove.
19384         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
19385         support.
19386         (fusion_addis_mem_combo_load): Ditto.
19387         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
19388         FP support.
19389         (rs6000_cpu_cpp_builtins): Ditto.
19390         * config/rs6000/rs6000-linux.c
19391         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
19392         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
19393         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
19394         support.
19395         (rs6000_setup_reg_addr_masks): Ditto.
19396         (rs6000_init_hard_regno_mode_ok): Ditto.
19397         (rs6000_option_override_internal): Ditto.
19398         (legitimate_lo_sum_address_p): Ditto.
19399         (rs6000_legitimize_address): Ditto.
19400         (rs6000_legitimize_reload_address): Ditto.
19401         (rs6000_legitimate_address_p): Ditto.
19402         (abi_v4_pass_in_fpr): Ditto.
19403         (setup_incoming_varargs): Ditto.
19404         (rs6000_gimplify_va_arg): Ditto.
19405         (rs6000_split_multireg_move): Ditto.
19406         (rs6000_savres_strategy): Ditto.
19407         (rs6000_emit_prologue_components): Ditto.
19408         (rs6000_emit_epilogue_components): Ditto.
19409         (rs6000_emit_prologue): Ditto.
19410         (rs6000_emit_epilogue): Ditto.
19411         (rs6000_elf_file_end): Ditto.
19412         (rs6000_function_value): Ditto.
19413         (rs6000_libcall_value): Ditto.
19414         * config/rs6000/rs6000.h: Ditto.
19415         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
19416         (TARGET_MINMAX): ... this.  New.
19417         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
19418         * config/rs6000/rs6000.md: Remove Xilinx FP support.
19419         (*movsi_internal1_single): Delete.
19420         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
19421         mfpu=, mxilinx-fpu): Delete.
19422         * config/rs6000/singlefp.h: Delete.
19423         * config/rs6000/sysv4.h: Remove Xilinx FP support.
19424         * config/rs6000/t-rs6000: Ditto.
19425         * config/rs6000/t-xilinx: Delete.
19426         * config/rs6000/titan.md: Adjust for fp_type removal.
19427         * config/rs6000/vsx.md: Remove Xilinx FP support.
19428         (VStype_simple): Delete.
19429         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
19430         * config/rs6000/xfpu.h: Delete.
19431         * config/rs6000/xfpu.md: Delete.
19432         * config/rs6000/xilinx.h: Delete.
19433         * config/rs6000/xilinx.opt: Delete.
19434         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
19435         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
19437 2018-05-04  Tom de Vries  <tom@codesourcery.com>
19439         PR libgomp/85639
19440         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
19441         if ignore == 0.
19443 2018-05-04  Richard Biener  <rguenther@suse.de>
19445         PR middle-end/85627
19446         * tree-complex.c (update_complex_assignment): We are always in SSA form.
19447         (expand_complex_div_wide): Likewise.
19448         (expand_complex_operations_1): Likewise.
19449         (expand_complex_libcall): Preserve EH info of the original stmt.
19450         (tree_lower_complex): Handle removed blocks.
19451         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
19452         on complex multiplication and division libcall builtins.
19454 2018-05-04  Richard Biener  <rguenther@suse.de>
19456         PR middle-end/85574
19457         * fold-const.c (negate_expr_p): Restrict negation of operand
19458         zero of a division to when we know that can happen without
19459         overflow.
19460         (fold_negate_expr_1): Likewise.
19462 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
19464         PR libstdc++/85466
19465         * real.h (real_nextafter): Declare.
19466         * real.c (real_nextafter): New function.
19467         * fold-const-call.c (fold_const_nextafter): New function.
19468         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
19469         CASE_CFN_NEXTTOWARD.
19470         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
19471         even when arg1_mode is different from arg0_mode.
19473 2018-05-03  Nathan Sidwell  <nathan@acm.org>
19475         * doc/extend.texi (Deprecated Features): Remove
19476         -ffriend-injection.
19477         (Backwards Compatibility): Likewise.
19478         * doc/invoke.texi (C++ Language Options): Likewise.
19479         (C++ Dialect Options): Likewise.
19481 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
19483         PR target/85530
19484         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
19485         _mm512_mask_mullox_epi64): New intrinsics.
19487 2018-05-03  Tom de Vries  <tom@codesourcery.com>
19489         PR testsuite/85106
19490         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19491         dump files): Add offload-tree.
19493 2018-05-03  Richard Biener  <rguenther@suse.de>
19495         PR tree-optimization/85615
19496         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
19497         to loops not nested in BBs loop father to avoid creating multi-entry
19498         loops.
19500 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19502         PR tree-optimization/70291
19503         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
19504         arguments.  Change return type to tree.  Emit libcall as a new
19505         statement rather than replacing existing one when inplace_p is true.
19506         (expand_complex_multiplication_components): New function.
19507         (expand_complex_multiplication): Expand floating-point complex
19508         multiplication using the above.
19509         (expand_complex_division): Rename inner_type parameter to type.
19510         Update expand_complex_libcall call-site.
19511         (expand_complex_operations_1): Update expand_complex_multiplication
19512         and expand_complex_division call-sites.
19514 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19516         PR target/85582
19517         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19518         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19519         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
19520         the highest significant bit of the shift count mask is clear.  In
19521         check whether and[sq]i3 is needed verify that all significant bits
19522         of the shift count other than the highest are set.
19524 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19526         PR libgomp/82428
19527         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
19528         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
19529         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
19530         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
19531         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
19532         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
19533         __builtin_goacc_parlevel_size.
19535 2018-05-02  Richard Biener  <rguenther@suse.de>
19537         PR tree-optimization/85597
19538         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
19539         do not use split vect_get_vec_defs call but call vect_get_slp_defs
19540         directly.
19542 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19544         PR testsuite/85106
19545         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19546         dump files): Add ltrans-tree.
19548 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19550         PR testsuite/85106
19551         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19552         dump files): Add wpa-ipa.
19554 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
19556         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
19557         powerpc*-*-linux*paired* target.
19558         * config/rs6000/750cl.h: Delete.
19559         * config/rs6000/paired.h: Delete.
19560         * config/rs6000/paired.md: Delete.
19561         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
19562         float support.
19563         * config/rs6000/rs6000-builtin.def: Remove paired float support.
19564         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
19565         comment.  Remove paired float support.
19566         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
19567         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
19568         VECTOR_PAIRED.
19569         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
19570         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
19571         declarations.
19572         * config/rs6000/rs6000.c: Remove paired float support.
19573         (paired_expand_vector_init, paired_expand_vector_move,
19574         paired_emit_vector_compare, paired_emit_vector_cond_expr,
19575         (paired_expand_lv_builtin, paired_expand_stv_builtin,
19576         paired_expand_builtin, paired_expand_predicate_builtin,
19577         paired_init_builtins): Delete.
19578         * config/rs6000/rs6000.h: Remove paired float support.
19579         * config/rs6000/rs6000.md: Remove paired float support.
19580         (move_from_CR_ov_bit): Delete.
19581         * config/rs6000/rs6000.opt (mpaired): Delete.
19582         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
19583         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
19585 2018-05-02  Richard Biener  <rguenther@suse.de>
19587         PR middle-end/85567
19588         * gimplify.c (gimplify_save_expr): When in SSA form allow
19589         SAVE_EXPRs to compute to SSA vars.
19591 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19593         PR target/85582
19594         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19595         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19596         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
19597         clobber operands[2], instead use a new pseudo.  Formatting fixes.
19599 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
19601         PR tree-optimization/85586
19602         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
19603         exit early for statements in the same group if the accesses are
19604         not strided.
19606 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19608         PR lto/85451
19609         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
19610         error message.
19612 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
19614         PR tree-optimization/85143
19615         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
19617 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19619         PR lto/85451
19620         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
19621         not found" error message.
19623 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19625         PR other/83786
19626         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
19627         * vec.c (test_ordered_remove_if): New function.
19628         (vec_c_tests): Call test_ordered_remove_if.
19629         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
19630         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
19631         * tree-vect-patterns.c (vect_pattern_recog_1): Use
19632         VEC_ORDERED_REMOVE_IF.
19634 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19636         PR tree-optimization/82665
19637         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
19638         pointer subtraction where arguments come from a memchr call.
19640 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
19642         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
19643         --push-state --as-needed and --pop-state instead of --as-needed and
19644         --no-as-needed if ld supports it.
19645         * configure: Regenerated.
19647         PR web/85578
19648         * doc/install.texi2html: Replace _002d with - and _002a with * in
19649         generated html files using sed.
19651 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19653         PR c++/85523
19654         * gcc-rich-location.c (blank_line_before_p): New function.
19655         (use_new_line): New function.
19656         (gcc_rich_location::add_fixit_insert_formatted): New function.
19657         * gcc-rich-location.h
19658         (gcc_rich_location::add_fixit_insert_formatted): New function.
19660 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19662         * selftest.c (assert_streq): Rename "expected" and "actual" to
19663         "val1" and "val2".  Extend NULL-handling to cover both inputs
19664         symmetrically, while still requiring both to be non-NULL for a pass.
19665         * selftest.h (assert_streq): Rename "expected" and "actual" to
19666         "val1" and "val2".
19667         (ASSERT_EQ): Likewise.
19668         (ASSERT_EQ_AT): Likewise.
19669         (ASSERT_KNOWN_EQ): Likewise.
19670         (ASSERT_KNOWN_EQ_AT): Likewise.
19671         (ASSERT_NE): Likewise.
19672         (ASSERT_MAYBE_NE): Likewise.
19673         (ASSERT_MAYBE_NE_AT): Likewise.
19674         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
19675         the assertion to pass.
19676         (ASSERT_STREQ_AT): Likewise.
19678 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
19680         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
19681         interaction with -pie.
19683 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19685         * selftest.h: Fix alphabetization of per-source-file selftest
19686         declarations.
19688 2018-04-30  Jason Merrill  <jason@redhat.com>
19690         PR c++/61982 - dead stores to destroyed objects.
19691         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
19692         of clobber.
19694 2018-04-30  Jason Merrill  <jason@redhat.com>
19696         * tree.c (build_clobber): New.
19697         * tree.h: Declare it.
19698         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
19700 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19702         * diagnostic-show-locus.c (layout::layout): Update for
19703         location_get_source_line returning a char_span.
19704         (struct char_span): Move to input.h.
19705         (struct correction): Update for fields in char_span becoming
19706         private.
19707         (struct source_line): Update for location_get_source_line
19708         returning a char_span.
19709         (layout::print_line): Likewise.
19710         * edit-context.c (edited_file::print_content): Likewise.
19711         (edited_file::print_diff_hunk): Likewise.
19712         (edited_file::print_run_of_changed_lines): Likewise.
19713         (edited_file::get_num_lines): Likewise.
19714         (edited_line::edited_line): Likewise.
19715         * final.c (asm_show_source): Likewise.
19716         * input.c (location_get_source_line): Convert return type
19717         from const char * to char_span, losing the final "line_len"
19718         param.
19719         (dump_location_info): Update for the above.
19720         (get_substring_ranges_for_loc): Likewise.  Use a char_span
19721         when handling the literal within the line.
19722         (test_reading_source_line): Update for location_get_source_line
19723         returning a char_span.
19724         * input.h (class char_span): Move here from
19725         diagnostic-show-locus.c, converting from a struct to a class.
19726         Make data members private.
19727         (char_span::operator bool): New.
19728         (char_span::length): New.
19729         (char_span::get_buffer): New.
19730         (char_span::operator[]): New.
19731         (char_span::subspan): Make const.
19732         (char_span::xstrdup): New.
19733         (location_get_source_line): Convert return type from const char *
19734         to char_span, losing the final "line_size" param.
19736 2018-04-30  Jan Hubicka  <jh@suse.cz>
19738         * lto-wrapper.c (ltrans_priorities): New static var.
19739         (cmp_priority): New.
19740         (run_gcc): Read priorities and if doing parallel build order
19741         the Makefile by them.
19743 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19745         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
19747 2018-04-30  Richard Biener  <rguenther@suse.de>
19749         * tree-cfg.c (verify_address): Remove base argument, add
19750         flag whether to check TREE_ADDRESSABLE and do that.
19751         (verify_expr): Remove.
19752         (verify_types_in_gimple_reference): Add pieces from verify_expr.
19753         (verify_gimple_assign_single): Likewise.
19754         (verify_gimple_switch): Likewise.
19755         (verify_expr_location_1): Dereference tp once.  Add (disabled)
19756         piece from verify_expr.
19757         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
19759 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19761         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
19763 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19765         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
19766         (small_data_pattern): Likewise.
19767         (arc_rewrite_small_data): Likewise.
19768         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
19769         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
19770         (get_symbol_alignment): New function.
19771         (legitimate_small_data_address_p): Likewise.
19772         (legitimate_scaled_address): Update, call
19773         legitimate_small_data_address_p.
19774         (output_sdata): New static variable.
19775         (arc_print_operand): Update how we handle small data operands.
19776         (arc_print_operand_address): Likewise.
19777         (arc_legitimate_address_p): Update, use
19778         legitimate_small_data_address_p.
19779         (arc_rewrite_small_data_p): Remove.
19780         (arc_rewrite_small_data_1): Likewise.
19781         (arc_rewrite_small_data): Likewise.
19782         (small_data_pattern): Likewise.
19783         (compact_sda_memory_operand): Update to use
19784         legitimate_small_data_address_p and get_symbol_alignment.
19785         (prepare_move_operands): Don't rewite sdata pattern.
19786         (prepare_extend_operands): Remove.
19787         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
19788         pattern.
19789         (zero_extendqisi2): Likewise.
19790         (zero_extendhisi2): Likewise.
19791         (extendqihi2): Likewise.
19792         (extendqisi2): Likewise.
19793         (extendhisi2): Likewise.
19794         (addsi3): Likewise.
19795         (subsi3): Likewise.
19796         (andsi3): Likewise.
19797         * config/arc/constraints.md (Usd): Change it to memory constraint.
19799 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19801         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
19802         as source of std instructions.
19803         * config/arc/arc.md (movsi_insn): Update pattern predicate to
19804         allow 6-bit constants as source for store instructions.
19805         (movdi_insn): Update instruction pattern to allow 6-bit constants
19806         as source for store instructions.
19808 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
19810         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
19812 2018-04-30  Nathan Sidwell  <nathan@acm.org>
19813             Sandra Loosemore <sandra@codesourcery.com>
19815         * dumpfile.c (dump_open): Allow '-' for stdout.
19816         * doc/invoke.texi (Developer Options): Document dump filename
19817         determination early.  Document stdin/stdout selection.
19819 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
19821         Microblaze Target: PIC data text relative
19823         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
19824         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
19825         Add declaration.
19826         * config/microblaze/microblaze.h (microblaze_constant_address_p):
19827         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
19828         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
19829         New addressing mode for data-text relative position indepenedent code.
19830         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
19831         'ADDRESS_SYMBOLIC_TXT_REL'.
19832         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
19833         (microblaze_legitimate_pic_operand): Exclude function calls from
19834         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
19835         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
19836         addresses cases.
19837         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19838         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19839         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
19840         for 'address + offset'.
19841         (microblaze_expand_prologue): Add new function prologue call for
19842         'r20' assignation.
19843         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
19844         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
19845         table in case of TARGET_PIC_DATA_TEXT_REL.
19846         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
19847         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
19848         Add new macros 'UNSPEC_TEXT',
19849         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
19850         + exclude function calls from 'UNSPEC_PLT' in case of data text
19851         relative mode.
19852         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19853         new target hook for generating address diff vector tables in case of
19854         flag_pic.
19855         * doc/tm.texi : Regenerate.
19856         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
19857         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
19858         of addr diff vector generation.
19859         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19860         target hook definition.
19861         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
19862         Add default function for generate_pic_addr_diff_vec -> flag_pic.
19863         * doc/invoke.texi (Add new pic option): Add new microblaze pic
19864         option for data text relative.
19866 2018-04-30  Richard Biener  <rguenther@suse.de>
19868         * tree-chrec.h (evolution_function_is_constant_p): Remove
19869         redundant check.
19870         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
19872 2018-04-30  Richard Biener  <rguenther@suse.de>
19874         PR bootstrap/85571
19875         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
19877 2018-04-30  Richard Biener  <rguenther@suse.de>
19879         PR tree-optimization/28364
19880         PR tree-optimization/85275
19881         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
19882         copying first exit test.
19884 2018-04-28  Mark Wielaard  <mark@klomp.org>
19886         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
19887         dwarf_version >= 5.
19888         (dwarf_AT): Handle DW_AT_addr_base.
19889         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
19891 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
19893         PR target/84431
19894         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
19895         (*ashl<dwi>3_doubleword_mask_1): Ditto.
19896         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
19897         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
19899 2018-04-28  Richard Biener  <rguenther@suse.de>
19901         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
19902         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
19903         to reflect use.  Only add interesting stmts.
19905 2018-04-27  Martin Jambor  <mjambor@suse.cz>
19907         PR ipa/85549
19908         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
19909         the jump function allows for passing through aggregate values.
19911 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
19913         * input.h (in_system_header_at): Convert from macro to inline
19914         function.
19915         (from_macro_expansion_at): Likewise.
19916         (from_macro_definition_at): Likewise.
19918 2018-04-27  Jeff Law  <law@redhat.com>
19920         * config.gcc: Mark tile* targets as deprecated/obsolete.
19922 2018-04-27  Richard Biener  <rguenther@suse.de>
19924         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
19925         fix for ILP32.
19927 2018-04-27  Richard Biener  <rguenther@suse.de>
19929         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
19931 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
19933         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
19934         with Yd constraint. Set "preferred_for_speed" attribute from
19935         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
19936         with Yd constraint.
19937         (*movdi_internal): Ditto.
19938         (movti_interunit splitters): Remove
19939         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
19940         (movdi_interunit splitters): Ditto.
19941         * config/i386/constraints.md (Ye): Remove.
19942         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
19944 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19946         PR target/85512
19947         * config/aarch64/constraints.md (Usg): Limit to 31.
19948         (Usj): Limit to 63.
19950 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
19952         PR tree-optimization/85529
19953         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
19954         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
19955         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
19956         zero extension or masking of the MSB bit.
19957         (optimize_range_tests): Add FIRST_BB argument, pass it through
19958         to optimize_range_tests_var_bound.
19959         (maybe_optimize_range_tests, reassociate_bb): Adjust
19960         optimize_range_tests callers.
19962 2018-04-26  Richard Biener  <rguenther@suse.de>
19963             Jakub Jelinek  <jakub@redhat.com>
19965         * cgraph.h (symbol_table): Just declare debug method here.
19966         * symtab.c (symbol_table::debug): Define.
19968 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
19970         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
19972 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
19974         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
19975         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
19976         (*movdi_internal): Substitute Yi and Yj constraint with x
19977         and Ym and Yn constraint with y constraint.  Update "isa"
19978         attribute and set "preferred_for_speed" attribute from
19979         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
19980         (*movsi_internal): Ditto.
19981         (*movdf_internal): Ditto.
19982         (*movsf_internal): Ditto.
19983         (*zero_extendsidi2): Ditto.
19984         * config/i386/sse.md (vec_set<mode>_0): Ditto.
19985         (sse2_loadld): Ditto.
19986         (*vec_extract<ssevecmodelower>_0): Ditto.
19987         (*vec_extractv4si_0_zext_sse4): Ditto.
19988         (vec_concatv2di): Ditto.
19989         (*vec_dup<mode>): Ditto.
19990         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
19991         * config/i386/constraints.md (Yi): Remove.
19992         (Yj): Remove.
19993         (Ym): Remove.
19994         (Yn): Remove.
19996 2018-04-26  Nathan Sidwell  <nathan@acm.org>
19998         * dumpfile.c (dump_open): New.
19999         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20000         (dump_finish): Detect stdio/stderr by value not name.
20002 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20004         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20006 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20008         PR target/84952
20009         * config/nvptx/nvptx.c (verify_neutering_jumps)
20010         (verify_neutering_labels): New function
20011         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20013 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20015         PR target/84025
20016         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20017         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20018         need neutering.
20020 2018-04-26  Richard Biener <rguenther@suse.de>
20021             Tom de Vries  <tom@codesourcery.com>
20023         PR lto/85422
20024         * lto-streamer-out.c (output_function): Fixup loops if required to match
20025         discovery done in the reader.
20027 2018-04-26  Richard Biener  <rguenther@suse.de>
20029         PR tree-optimization/85116
20030         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20031         have a loop exit from the single latch predecessor.  Remove
20032         case of header with just condition.
20033         (ch_base::copy_headers): Exclude infinite loops from any
20034         processing.
20035         (pass_ch::execute): Record exits.
20037 2018-04-26  Richard Biener  <rguenther@suse.de>
20039         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20040         prologue cost vector and pass it to vect_get_load_cost.
20041         (vect_get_peeling_costs_all_drs): Likewise.
20042         (vect_peeling_hash_get_lowest_cost): Likewise.
20043         (vect_enhance_data_refs_alignment): Likewise.
20045 2018-04-26  Richard Biener  <rguenther@suse.de>
20047         PR middle-end/85450
20048         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20049         checking of integer<->pointer conversions.
20050         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20051         sign-/zero-extending pointer types.
20052         (expand_omp_for_static_chunk): Likewise.
20054 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
20055             Jean Lee  <xiaoyur347@gmail.com>
20057         * config/mips/mips.c (mips_asan_shadow_offset): New function.
20058         (TARGET_ASAN_SHADOW_OFFSET): Define.
20059         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20060         true for -fsanitize=address.
20062 2018-04-25  Mark Wielaard  <mark@klomp.org>
20064         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20065         shorter ones.
20067 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20069         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20070         than "alu", remove explicit "memory" and "imm_disp" attributes.
20071         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20073         PR middle-end/85414
20074         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20075         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20076         gen_lowpart_no_emit.
20078 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
20080         PR target/85473
20081         * config/i386/i386.c (ix86_expand_builtin): Change memory
20082         operand to XI, extend p0 to Pmode.
20083         * config/i386/i386.md: Change unspec volatile and operand
20084         1 mode to XI, change operand 0 mode to P.
20086 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20088         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20089         GET_MODE_MASK before any checking.
20090         (nds32_can_use_bset_p): Likewise.
20091         (nds32_can_use_btgl_p): Likewise.
20093 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20095         * config/nds32/nds32-doubleword.md: New define_split pattern for
20096         illegal register number.
20098 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20100         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20102 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
20104         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20106 2018-04-25  Richard Biener  <rguenther@suse.de>
20108         * lto-streamer.h (LTO_major_version): Bump to 8.
20110 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
20112         * BASE-VER: Set to 9.0.0.
20114 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
20116         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20117         in __abskf2 and __powikf2.
20119 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20121         PR target/85512
20122         * config/aarch64/constraints.md (Usg, Usj): New constraints.
20123         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20124         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20125         Use the above on operand 2.  Reindent.
20126         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20128 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20130         PR target/85485
20131         * common/config/i386/i386-common.c (ix86_handle_option): Don't
20132         handle OPT_mcet.
20133         * config/i386/i386.opt (mcet): Removed.
20134         * doc/install.texi: Remove -mcet documentation.
20135         * doc/invoke.texi: Likewise.
20137 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
20139         PR target/85485
20140         * doc/install.texi: Remove -mcet from bootstrap-cet.
20142 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20144         PR target/85511
20145         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20146         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20147         if TARGET_64BIT.
20149         PR target/85503
20150         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20151         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20152         containing a CONST_VECTOR.
20154 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20156         * doc/install.texi: Update newlib dependency for nvptx.
20158 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20160         PR target/85508
20161         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20162         instead of INTVAL when shifting x left.
20164 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20166         PR tree-optimization/85478
20167         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20168         vect_grouped_store_supported for single element vectors.
20170 2018-04-24  Richard Biener  <rguenther@suse.de>
20172         PR target/85491
20173         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20174         load cost increase to the case of non-constant step.
20176 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20178         PR target/84828
20179         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20180         destination if any_malformed_asm.
20182 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
20184         PR middle-end/85496
20185         * expr.c (store_field): In the bitfield case, if the value comes from
20186         a function call and is returned in registers by means of a PARALLEL,
20187         do not change the mode of the temporary unless BLKmode and VOIDmode.
20189 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
20191         PR rtl-optimization/85423
20192         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20193         dependencies to debug insns when the previous insn is non-debug.
20195 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20197         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20198         enums into a single definition.
20199         (fls): Fix predicates and printing.
20200         (seti): Likewise.
20202 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20204         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20205         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20206         and short u6 immediate.
20207         (check_if_valid_sleep_operand): Remove.
20208         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20210 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20212         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20213         flag_always_save_lp condition.
20214         * config/nds32/nds32.opt (malways-save-lp): New option.
20216 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20218         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20219         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20220         * config/nds32/nds32.h
20221         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20222         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20224 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20226         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20227         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20229 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20230             Chung-Ju Wu  <jasonwucj@gmail.com>
20232         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20233         Declare.
20234         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20235         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20237 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20239         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20241 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20243         * config/nds32/nds32-protos.h (nds32_data_alignment,
20244         nds32_local_alignment): Declare.
20245         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20246         nds32_local_alignment): New functions.
20247         (TARGET_CONSTANT_ALIGNMENT): Define.
20248         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20250 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20252         * config/nds32/nds32.c
20253         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20254         (TARGET_MODES_TIEABLE_P): Likewise.
20256 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20258         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20259         level Ofast and Og.
20261 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
20262             Chung-Ju Wu  <jasonwucj@gmail.com>
20264         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20265         for unaligned access.
20266         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20267         unaligned access.
20268         * config/nds32/nds32-intrinsic.md: Likewise.
20269         * config/nds32/nds32_intrinsic.h: Likewise.
20270         * config/nds32/nds32.h (nds32_builtins): Likewise.
20271         * config/nds32/nds32.opt (munaligned-access): New option.
20272         * config/nds32/nds32.c (nds32_asm_file_start): Display
20273         flag_unaligned_access status.
20275 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20277         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20278         -mno-relax is present.
20279         * config/riscv/linux.h (LINK_SPEC): Ditto.
20281 2018-04-20  Martin Sebor  <msebor@redhat.com>
20283         PR c/85365
20284         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20285         for null pointers.
20286         (gimple_fold_builtin_stxcpy_chk): Same.
20287         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20289 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
20291         PR target/85456
20292         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20293         __powikf2 when long double is IEEE 128-bit.
20295 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20297         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20298         step to make sure stack always aligned.
20300 2018-04-20  Carl Love  <cel@us.ibm.com>
20302         PR target/83402
20303         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20304         size check for arg0.
20306 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
20307             Tom de Vries  <tom@codesourcery.com>
20309         PR target/85445
20310         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20311         Emit insns for calls too.
20312         (nvptx_find_par): Always look for worker-level predecessor insn.
20313         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
20314         calls.
20315         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
20316         (nvptx_process_pars): Propagate frames for calls.
20318 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
20320         PR target/85469
20321         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
20322         Removed.
20323         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
20324         (ix86_handle_option): Don't handle OPT_mibt.
20325         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
20326         __SHSTK__.
20327         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
20328         has_ibt and ibt.
20329         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
20330         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
20331         (ix86_target_macros): Define __CET__ with flag_cf_protection
20332         for -fcf-protection.
20333         * config/i386/i386.c (isa2_opts): Remove -mibt.
20334         * config/i386/i386.h (TARGET_IBT): Removed.
20335         (TARGET_IBT_P): Likewise.
20336         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
20337         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
20338         * config/i386/i386.opt (mcet): Update help message.
20339         (mshstk): Likewise.
20340         (mibt): Removed.
20341         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
20342         -mcet as an alias for -mshstk.
20344 2018-04-20  Richard Biener <rguenther@suse.de>
20346         PR middle-end/85475
20347         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
20348         complexity by forcing a single use of the multiply operand.
20350 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20352         ipa/85449
20353         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
20354         recursion dependency to only apply to non-clones.
20356 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20358         ipa/85447
20359         * ipa-cp.c (create_specialized_node): Check that clones of
20360         self-recursive edges exist during IPA-CP.
20362 2018-04-19  Toon Moene  <toon@moene.org>
20364         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
20365         by -O3.
20367 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20369         PR tree-optimization/85467
20370         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
20371         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
20372         VECTOR_CST element to type.
20374 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20376         PR target/85397
20377         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
20378         * config/i386/i386.md (builtin_setjmp_setup): Removed.
20379         (builtin_longjmp): Likewise.
20380         (save_stack_nonlocal): New pattern.
20381         (restore_stack_nonlocal): Likewise.
20383 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20385         PR target/85404
20386         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20387         Replace ASM_OUTPUT_LABEL with fprintf.
20389 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20391         PR target/85417
20392         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20393         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
20394         * config/i386/i386-c.c (ix86_target_macros_internal): Also
20395         define __IBT__ and __SHSTK__ for -fcf-protection.
20396         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
20397         TARGET_IBT.
20398         (ix86_trampoline_init): Likewise.
20399         (x86_output_mi_thunk): Likewise.
20400         (ix86_notrack_prefixed_insn_p): Likewise.
20401         (ix86_option_override_internal): Don't disallow -fcf-protection.
20402         * config/i386/i386.md (rdssp<mode>): Also enable for
20403         -fcf-protection.
20404         (incssp<mode>): Likewise.
20405         (nop_endbr): Likewise.
20406         * config/i386/i386.opt (mcet): Change help message to built-in
20407         functions only.
20408         (mibt): Likewise.
20409         (mshstk): Likewise.
20410         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
20411         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
20412         enable CET built-in functions.
20414 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
20416         * common/config/i386/i386-common.c
20417         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
20418         OPTION_MASK_ISA_MOVDIRI_UNSET,
20419         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
20420         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
20421         * config.gcc (movdirintrin.h): New header.
20422         * config/i386/cpuid.h (bit_MOVDIRI,
20423         bit_MOVDIR64B): New bits.
20424         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
20425         and -mmvodir64b.
20426         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
20427         (VOID, PVOID, PCVOID)): New function types.
20428         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
20429         __builtin_ia32_directstoreu_u64,
20430         __builtin_ia32_movdir64b): New builtins.
20431         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
20432         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
20433         and -mmovdiri.
20434         (ix86_valid_target_attribute_inner_p): Ditto.
20435         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
20436         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
20437         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
20438         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
20439         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
20440         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
20441         (movdiri<mode>, movdir64b_<mode>): New.
20442         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
20443         * config/i386/immintrin.h: Include movdirintrin.h.
20444         * config/i386/movdirintrin.h: New file.
20445         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
20447 2018-04-19  Richard Biener  <rguenther@suse.de>
20449         PR middle-end/85455
20450         * cfg.c (clear_bb_flags): When loop state says we have
20451         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
20453 2018-04-19  Richard Biener  <rguenther@suse.de>
20455         PR tree-optimization/84737
20456         * tree-vect-data-refs.c (vect_copy_ref_info): New function
20457         copying restrict info.
20458         (vect_setup_realignment): Use it.
20459         * tree-vectorizer.h (vect_copy_ref_info): Declare.
20460         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
20461         the first DR to all generated stores.
20462         (vectorizable_load): Likewise for loads.
20464 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20466         PR tree-optimization/85446
20467         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
20468         the integral and pointer types to have the same precision.
20470         * doc/install.texi: Document --disable-cet being the default and
20471         --enable-cet=auto.
20473 2018-04-18  Martin Liska  <mliska@suse.cz>
20475         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
20476         style.
20478 2018-04-18  Martin Liska  <mliska@suse.cz>
20480         Revert
20481         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20483         PR ipa/83983
20484         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20485         arguments if they are comparable.
20487 2018-04-18  Martin Liska  <mliska@suse.cz>
20489         Revert
20490         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20492         PR lto/84805
20493         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20494         incomplete types.
20496 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
20498         PR target/85388
20499         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
20500         ENDBR after calling __morestack.
20502 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
20504         PR jit/85384
20505         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
20506         by using gcc_base_ver to generate a gcc_driver_version, and use
20507         it when generating GCC_DRIVER_NAME.
20508         * configure: Regenerate.
20510 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
20512         PR target/81084
20513         * config.gcc: Obsolete powerpc*-*-*spe*.
20515 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20517         PR debug/84637
20518         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
20519         (stabstr_D): Change type of unum from unsigned int to
20520         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
20521         type.
20523 2018-04-17  Jim Wilson  <jimw@sifive.com>
20525         PR 84856
20526         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
20527         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
20528         Set arg_pointer_offset after using pretend_args_size.
20530 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20532         PR rtl-optimization/85431
20533         * dse.c (record_store): Ignore zero width stores.
20535         PR sanitizer/85230
20536         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
20537         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
20538         __builtin_stack_restore rather than after it.
20539         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
20540         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
20541         argument instead of virtual_dynamic_stack_rtx.
20543 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20545         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
20546         New prototype.
20547         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20548         Add note to error message to explain internal mapping of overloaded
20549         built-in function name to non-overloaded built-in function name.
20550         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
20551         function.
20553 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
20555         PR target/85424
20556         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
20557         where the inputs overlap with the output.
20559 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20561         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
20562         (=v, v) alternative and explicit "memory" attribute.
20563         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
20564         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20565         attributes.
20566         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
20567         "sselog1" type instead of "sselog".
20568         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
20569         "sselog".  Remove explicit "memory" attribute.
20570         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
20571         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20572         attributes.
20573         (vec_extract_hi_v32hi): Merge all alternatives into one, use
20574         "sselog1" type instead of "sselog".  Remove explicit "memory"
20575         attribute.
20576         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
20577         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20578         attribute.
20579         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
20580         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20581         attributes.
20582         (vec_extract_hi_v64qi): Merge all alternatives into one, use
20583         "sselog1" type instead of "sselog".  Remove explicit "memory"
20584         attribute.
20585         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
20586         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20587         attribute.
20589         PR target/85430
20590         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
20592         PR middle-end/85414
20593         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
20594         on a SUBREG.
20596 2018-04-17  Martin Jambor  <mjambor@suse.cz>
20598         PR ipa/85421
20599         * ipa-cp.c (create_specialized_node): Call
20600         expand_all_artificial_thunks if necessary.
20602 2018-04-17  Martin Liska  <mliska@suse.cz>
20604         PR lto/85405
20605         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
20606         in message, remote space in between '_G' and '('.
20608 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20610         PR target/85281
20611         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
20612         avx512f_vmcmp<mode>3<round_saeonly_name>,
20613         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
20614         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
20615         avx512f_rndscale<mode><round_saeonly_name>,
20616         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
20617         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
20618         Use %<iptr>2 instead of %2 for -masm=intel.
20619         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
20620         avx512f_vcvttss2usi<round_saeonly_name>,
20621         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
20622         -masm=intel.
20623         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
20624         avx512f_vcvttsd2usi<round_saeonly_name>,
20625         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
20626         Use %q1 instead of %1 for -masm=intel.
20627         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
20628         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
20629         of %3 for -masm=intel.
20630         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
20631         -masm=intel.
20632         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
20633         -masm=intel.
20634         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
20635         -masm=intel.
20636         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
20637         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
20638         %g1.
20639         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
20640         -masm=intel.
20641         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
20642         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
20643         %g1 and one with %0 and %1.
20644         (avx512er_vmrcp28<mode><round_saeonly_name>,
20645         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
20646         %1 for -masm=intel.
20647         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
20648         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
20649         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
20650         of %0 and %{%4%} for -masm=intel.
20651         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
20652         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
20653         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
20654         order of %0 and %{%5%}%{z%} for -masm=intel.
20656 2018-04-17  Jan Hubicka  <jh@suse.cz>
20658         PR lto/85405
20659         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
20661 2018-04-17  Martin Liska  <mliska@suse.cz>
20663         PR ipa/85329
20664         * multiple_target.c (create_dispatcher_calls): Set apostrophes
20665         for target_clone error message.  Make default implementation
20666         clone to be a local declaration.
20667         (separate_attrs): Add new argument and check for an empty
20668         string.
20669         (expand_target_clones): Handle it.
20670         (ipa_target_clone): Make redirection just for target_clones
20671         functions.
20673 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
20674             Tom de Vries  <tom@codesourcery.com>
20676         PR middle-end/84955
20677         * omp-expand.c (expand_oacc_for): Add dummy false branch for
20678         tiled basic blocks without omp continue statements.
20680 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
20682         PR target/83660
20683         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
20684         vec_extract expression as having side effects to make sure it gets
20685         a cleanup point.
20687 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
20689         PR target/85403
20690         * config/i386/i386.c (get_builtin_code_for_version): Check
20691         error_mark_node.
20693 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
20695         PR target/84331
20696         * config.gcc: Support "skylake".
20697         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20698         PROCESSOR_SKYLAKE.
20699         * config/i386/i386.c (m_SKYLAKE): Define.
20700         (processor_target_table): Add "skylake".
20701         (ix86_option_override_internal): Add "skylake".
20702         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
20703         PROCESSOR_CANNONLAKE.
20704         (get_builtin_code_for_version): Fix priority for
20705         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
20706         PROCESSOR_SKYLAKE-AVX512.
20707         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
20708         (processor_type): Add PROCESSOR_SKYLAKE.
20710 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
20711             Jason Merrill  <jason@redhat.com>
20713         PR c++/85112
20714         * convert.c (convert_to_integer_1): Use direct recursion for
20715         enumeral types and types with a precision less than the number
20716         of bits in their mode.
20718 2018-04-16  Julia Koval  <julia.koval@intel.com>
20720         PR target/84413
20721         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
20722         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
20724 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
20726         PR target/85293
20727         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
20728         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
20729         and -mno-direct-move.
20731 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
20733         PR target/83402
20734         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
20735         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
20736         Ensure negative shifts result in {0}.
20738 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
20740         PR rtl-optimization/79916
20741         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
20742         regs (if any) to define how to gnerate SD moves when LRA is in
20743         progress.
20745 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
20747         PR rtl-optimization/85393
20748         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
20749         * except.c (expand_dw2_landing_pad_for_region): Make static.
20750         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
20751         a label and unconditional jump to old_bb, rather than
20752         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
20753         basic block.
20755         PR rtl-optimization/85376
20756         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
20757         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
20758         instead of a specific value.
20760 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20761             Bin Cheng  <bin.cheng@arm.com>
20763         PR tree-optimization/82965
20764         PR tree-optimization/83991
20765         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
20766         by_profile_only parameter.
20767         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
20768         information if the loop was predicted to iterate too many times.
20769         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
20771 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20773         PR lto/71991
20774         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
20775         always inline.
20777 2018-04-13  Martin Liska  <mliska@suse.cz>
20778             Jakub Jelinek  <jakub@redhat.com>
20780         PR middle-end/81657
20781         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
20782         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
20783         * builtins.c (expand_builtin_memory_copy_args): Use
20784         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
20785         handle dest_addr == pc_rtx.
20787 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
20789         PR target/85291
20790         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
20791         asked to not generate direct moves.
20792         (fix_trunc<mode>si2_stfiwx): Similar.
20793         (fix_trunc<mode>si2_internal): Similar.
20795 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20797         PR debug/83157
20798         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
20799         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
20800         lookup if dest in some wider mode is known to be const0_rtx and
20801         if so, record permanent equivalence for it to be ZERO_EXTEND of
20802         the narrower mode destination.
20804 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
20806         * lto-streamer-out.c (output_function): Revert 259346.
20807         * omp-expand.c (expand_oacc_for): Likewise.
20809 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
20811         PR rtl-optimization/85354
20812         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
20813         * sel-sched.c (sel_global_init): ... here.
20815 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
20817         PR target/85238
20818         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
20819         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
20820         mode for PE-COFF targets.
20821         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
20822         (i386_pe_asm_lto_end): Likewise.
20823         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
20824         (TARGET_ASM_LTO_END): Likewise.
20825         * config/i386/winnt.c (saved_debug_info_level): New static variable.
20826         (i386_pe_asm_lto_start): New function.
20827         (i386_pe_asm_lto_end): Likewise.
20829 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
20830             Richard Biener  <rguenther@suse.de>
20832         PR middle-end/84955
20833         * lto-streamer-out.c (output_function): Fix CFG loop state before
20834         streaming out.
20835         * omp-expand.c (expand_oacc_for): Handle calls to internal
20836         functions like regular functions.
20838 2018-04-12  Richard Biener  <rguenther@suse.de>
20840         PR lto/85371
20841         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
20842         for the early LTO debug to properly generate references to it
20843         during DIE emission.  Do not re-use that for the skeleton for
20844         split-dwarf.
20845         (dwarf2out_early_finish): Likewise.
20847 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20849         PR target/85328
20850         * config/i386/sse.md
20851         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
20852         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
20853         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
20854         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
20855         and output is a reg, avoid creating invalid lowpart subreg, but
20856         instead split into a 512-bit move.  Don't split if not AVX512VL,
20857         input is xmm16+ reg and output is a mem.
20858         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
20859         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
20860         xmm16+ reg and output is a mem.
20862 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20864         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
20865         also for flag_dwarf2_cfi_asm.
20867 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20869         PR rtl-optimization/85342
20870         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
20871         a bool scalar var inside of the loop instead.  Don't try to update
20872         recog_data.operand after failed apply_change_group.
20874 2018-04-12  Tom de Vries  <tom@codesourcery.com>
20876         PR target/85296
20877         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
20878         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
20879         array with flexible array member as array without given dimension.
20880         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
20881         argument for undefined param to true.
20883 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
20885         PR target/85321
20886         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20887         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
20888         from PowerPC section.
20889         * config/rs6000/sysv4.opt (mcall-): Improve help text.
20890         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
20891         help text that is too long.
20892         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
20893         help text that is too long.
20894         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
20895         help text that is too long.
20897 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
20899         * config/alpha/alpha.md (stack_probe_internal): Rename
20900         from "probe_stack".  Update all callers.
20902 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20904         PR rtl-optimization/84566
20905         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
20906         sched_macro_fuse_insns.
20908 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20910         PR target/84301
20911         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
20912         (compute_block_dependences): ... from here.
20914 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20916         PR tree-optimization/85331
20917         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
20918         from int to HOST_WIDE_INT.
20920 2018-04-11  Martin Jambor  <mjambor@suse.cz>
20922         PR ipa/84149
20923         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
20924         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
20925         not the same as the source val.
20926         (cgraph_edge_brings_value_p): New parameter.
20927         (gather_edges_for_value): Pass destination value to
20928         cgraph_edge_brings_value_p.
20929         (perhaps_add_new_callers): Likewise.
20930         (get_info_about_necessary_edges): Likewise and exclude values brought
20931         only by self-recursive edges.
20932         (create_specialized_node): Redirect only clones of self-calling edges.
20933         (+self_recursive_pass_through_p): New function.
20934         (find_more_scalar_values_for_callers_subset): Use it.
20935         (find_aggregate_values_for_callers_subset): Likewise.
20936         (known_aggs_to_agg_replacement_list): Removed.
20937         (decide_whether_version_node): Re-calculate known constants for all
20938         remaining context clones.
20940 2018-04-11  Richard Biener  <rguenther@suse.de>
20942         PR lto/85339
20943         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
20944         from early DWARF output.
20945         (dwarf2out_early_finish): Output line info unconditionally into
20946         early DWARF and add reference to it.
20948 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20950         PR target/85281
20951         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
20952         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
20953         other than V2DFmode using iptr mode attribute.
20954         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
20956 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20958         PR rtl-optimization/84659
20959         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
20961 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20963         PR debug/85302
20964         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
20965         SIZEP is NULL.
20966         (output_loc_list): Pass address of a dummy size variable even in the
20967         locview handling loop.
20968         (index_location_lists): Add comment on why skip_loc_list_entry can't
20969         call size_of_locs.
20971 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20973         PR target/85261
20974         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
20975         into register.
20977 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
20979         PR target/85321
20980         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20981         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
20982         and -mstring-compare-inline-limit.
20984 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20986         PR target/85287
20987         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
20988         for stack clash protection in a register whenever we need it to be in
20989         a register.
20991 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20993         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
20994         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
20996 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20998         PR target/85321
20999         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21000         the help text.
21001         (mlong-double-): Ditto.
21002         * config/rs6000/sysv4.opt (msdata=): Ditto.
21003         (mtls-size=): Ditto.
21005 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21007         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21008         erroneous entries for
21009         "vector int vec_ldl (int, long int *)", and
21010         "vector unsigned int vec_ldl (int, unsigned long int *)".
21011         Add comments and entries for
21012         "vector bool char vec_ldl (int, bool char *)",
21013         "vector bool short vec_ldl (int, bool short *)",
21014         "vector bool int vec_ldl (int, bool int *)",
21015         "vector bool long long vec_ldl (int, bool long long *)",
21016         "vector pixel vec_ldl (int, pixel *)",
21017         "vector long long vec_ldl (int, long long *)",
21018         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21019         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21020         type tree bool_long_long_type_node and correct definition of
21021         bool_V2DI_type_node to make reference to this new type tree.
21022         (rs6000_mangle_type): Replace erroneous reference to
21023         bool_long_type_node with bool_long_long_type_node.
21024         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21025         comments to emphasize sign distinctions for char and int types and
21026         replace RS6000_BTI_bool_long constant with
21027         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21028         use of RS6000_BTI_pixel.
21029         (bool_long_type_node): Remove this macro definition.
21030         (bool_long_long_type_node): New macro definition
21032 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21034         PR rtl-optimization/85300
21035         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21036         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21037         simplify_unary_operation fails.
21039 2018-04-10  Martin Liska  <mliska@suse.cz>
21041         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21042         cgraph_edge and ipa_ref.
21044 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21046         PR target/85177
21047         PR target/85255
21048         * config/i386/sse.md
21049         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21050         computation of the VEC_MERGE selector from mask.
21051         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21052         Fix decoding of the VEC_MERGE selector into mask.
21054 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21056         PR tree-optimization/85286
21057         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21059 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
21061         * final.c (final_1): Set insn_last_address as well as
21062         insn_current_address.
21064 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21066         PR target/85173
21067         * explow.c (emit_stack_probe): Call validize_mem on memory location
21068         before passing it to gen_probe_stack.  Create address operand and
21069         legitimize it for the probe_stack_address case.
21071 2018-04-09  Jan Hubicka  <jh@suse.cz>
21073         PR lto/85078
21074         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21075         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21076         * tree.c (free_lang_data_in_type): Fix handling of binfos;
21077         walk basetypes.
21078         (free_lang_data): Rebuild type inheritance graph.
21080 2018-04-09  Martin Sebor  <msebor@redhat.com>
21082         * invoke.texi (-finline-small-functions): Mention other optimization
21083         options.
21084         (-findirect-inlining, -fpartial-inlining): Same.
21085         (-finline-functions-called-once): Same.
21086         (-freorder-blocks-and-partition): Same.
21088 2018-04-09  Jan Hubicka  <jh@suse.cz>
21090         PR rtl/84058
21091         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21092         jumps; choose last target that matches the criteria (i.e.
21093         no partition changes for non-crossing jumps).
21094         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21095         support for redirecting crossing jumps to non-crossing.
21097 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
21099         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21100         also for naked functions.
21102 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
21104         * config/arc/arc.md (add_shift): New pattern.
21105         (add_shift2): Likewise.
21106         (sub_shift): Likewise.
21107         (sub_shift_cmp0_noout): Likewise.
21108         (compare_si_ashiftsi): Likewise.
21109         (xbfu_cmp0_noout): New combine pattern.
21110         (xbfu_cmp0"): Likewise.
21111         (movsi_set_cc_insn): Place the predicable variant first.
21112         (commutative_binary_cmp0_noout): Remove clobber.
21113         (commutative_binary_cmp0): New pattern.
21114         (noncommutative_binary_cmp0): Likewise.
21115         (noncommutative_binary_cmp0_noout): Likewise.
21116         (noncommutative_binary_comparison_result_used): Removed.
21117         (rsub_cmp0): New pattern.
21118         (rsub_cmp0_noout): Likewise.
21119         (extzvsi): Changed, keep only meaningful variants.
21120         (SQH, SEZ): New iterators.
21121         (SQH_postfix): New mode attribute.
21122         (SEZ_prefix): New code attribute.
21123         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21124         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21125         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21126         of numerical value.
21127         (noncommutative_operator): Check the availability of barrel
21128         shifter option.
21130 2018-04-09  Richard Biener  <rguenther@suse.de>
21132         PR tree-optimization/85284
21133         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21134         Only use the niter constraining form of simple_iv when the exit
21135         is always executed.
21137 2018-04-09  Tom de Vries  <tom@codesourcery.com>
21139         PR target/84041
21140         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21141         (define_expand "*memory_barrier"): New define_expand.
21142         (define_insn "memory_barrier"): New insn.
21144 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21146         PR rtl-optimization/80463
21147         PR rtl-optimization/83972
21148         PR rtl-optimization/83480
21150         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21151         correct producer for the insn.
21152         (tidy_control_flow): Fixup seqnos in case of debug insns.
21154 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21156         PR rtl-optimization/83913
21158         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21159         different sched-times when merging exprs.
21161 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21163         PR rtl-optimization/83962
21165         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21166         tidy_fallthru_edge and tidy_control_flow.
21168 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21170         PR rtl-optimization/83530
21172         * sel-sched.c (force_next_insn): New global variable.
21173         (remove_insn_for_debug): When force_next_insn is true, also leave only
21174         next insn in the ready list.
21175         (sel_sched_region): When the region wasn't scheduled, make another pass
21176         over it with force_next_insn set to 1.
21178 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21180         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21181         into tm_file.
21182         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21183         for interrupt control.
21184         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21185         functions for interrupt control.
21186         * config/nds32/nds32-intrinsic.md: Likewise.
21187         * config/nds32/nds32_intrinsic.h: Likewise.
21188         * config/nds32/nds32.h (nds32_builtins): Likewise.
21190 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21192         * config/nds32/nds32.c (nds32_init_machine_status,
21193         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21194         strict_aligned_p field.
21195         (nds32_expand_to_rtl_hook): New function.
21196         (TARGET_EXPAND_TO_RTL_HOOK): Define.
21197         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21199 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21200             Chung-Ju Wu  <jasonwucj@gmail.com>
21202         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21203         * config/nds32/nds32-n7.md: New file.
21204         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21205         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21206         pipeline.
21207         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21208         * config/nds32/nds32.md (pipeline_model): Add n7.
21209         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21210         * config/nds32/pipelines.md: Include n7 settings.
21212 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21213             Chung-Ju Wu  <jasonwucj@gmail.com>
21215         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21216         * config/nds32/nds32-e8.md: New file.
21217         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21218         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21219         pipeline.
21220         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21221         * config/nds32/nds32.md (pipeline_model): Add e8.
21222         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21223         * config/nds32/pipelines.md: Include e8 settings.
21225 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21226             Chung-Ju Wu  <jasonwucj@gmail.com>
21228         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21229         * config/nds32/nds32-n8.md: New file.
21230         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21231         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21232         pipeline.
21233         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21234         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21235         * config/nds32/nds32.md (pipeline_model): Add n8.
21236         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21237         * config/nds32/pipelines.md: Include n8 settings.
21239 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21240             Chung-Ju Wu  <jasonwucj@gmail.com>
21242         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21243         * config/nds32/nds32-n9-2r1w.md: New file.
21244         * config/nds32/nds32-n9-3r2w.md: New file.
21245         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21246         nds32_register_ports): New or modify for cpu n9.
21247         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21248         pipeline.
21249         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21250         * config/nds32/nds32-utils.c: New file.
21251         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21252         TARGET_MUL_SLOW): Define.
21253         * config/nds32/nds32.md (pipeline_model): New attribute.
21254         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21255         New options that support cpu n9.
21256         * config/nds32/pipelines.md: Include n9 settings.
21257         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21259 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21261         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21262         information if necessary.
21263         (output_cond_branch_compare_zero): Likewise.
21264         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21265         (nds32_target_alignment): Refine for alignment.
21266         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21267         (FUNCTION_BOUNDARY): Modify.
21268         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21269         align case.
21270         * config/nds32/nds32.opt (malways-align, malign-functions): New.
21272 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21274         * config/nds32/constants.md (unspec_volatile_element): Add values for
21275         TLB operation and data prefetch.
21276         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21277         functions for TLB operation and data prefetch.
21278         * config/nds32/nds32-intrinsic.md: Likewise.
21279         * config/nds32/nds32_intrinsic.h: Likewise.
21280         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21281         (nds32_print_operand): Likewise.
21282         * config/nds32/nds32.h (nds32_builtins): Likewise.
21284 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21285         Andrew Pinski <pinsika@gcc.gnu.org>
21287         PR middle-end/82976
21288         * match.pd: Use constant_boolean_node of correct type instead of
21289         boolean_true_node or boolean_false_node for simplifying
21290         pointer comparisons to zero.
21292 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21294         PR tree-optimization/80021
21295         * tree.c (verify_type_variant): Make error call in verify_variant_match
21296         translatable and remove final full stop.
21298 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21300         * config/nds32/constants.md (unspec_volatile_element): Add
21301         UNSPEC_VOLATILE_EH_RETURN.
21302         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21303         nds32_output_stack_pop): Support dwarf exception handling process.
21304         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21305         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21306         exception handling process.
21307         (nds32_compute_stack_frame): Likewise.
21308         (nds32_return_addr_rtx): Likewise.
21309         (nds32_initial_elimination_offset): Likewise.
21310         (nds32_expand_prologue): Likewise.
21311         (nds32_expand_epilogue): Likewise.
21312         (nds32_dynamic_chain_address): New function.
21313         * config/nds32/nds32.h (machine_function): Add fields for dwarf
21314         exception handling.
21315         (DYNAMIC_CHAIN_ADDRESS): Define.
21316         (EH_RETURN_DATA_REGNO): Define.
21317         (EH_RETURN_STACKADJ_RTX): Define.
21318         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
21319         patterns for dwarf exception handling.
21321 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21323         * config/nds32/nds32.h: Clean up obsolete macros.
21325 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21327         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21328         Add enum values for particular instructions.
21329         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21330         particular intrinsic functions.
21331         * config/nds32/nds32-intrinsic.md: Likewise.
21332         * config/nds32/nds32_intrinsic.h: Likewise.
21333         * config/nds32/nds32.h (nds32_builtins): Likewise.
21334         * config/nds32/nds32.md (type): Add pbsad and pbsada.
21335         (btst, ave): New patterns for particular instructions.
21337 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21339         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21340         Add enum values for atomic load/store and memory sync.
21341         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
21342         and memory sync.
21343         * config/nds32/nds32-intrinsic.md: Likewise.
21344         * config/nds32/nds32_intrinsic.h: Likewise.
21345         * config/nds32/nds32.h (nds32_builtins): Likewise.
21347 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21349         PR tree-optimization/85257
21350         * fold-const.c (native_encode_vector): If not all elts could fit
21351         and off is -1, return 0 rather than offset.
21352         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
21353         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
21354         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
21355         adjust buffer in native_interpret_expr call.
21357 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21359         * config/nds32/constants.md (unspec_volatile_element): Add cache
21360         control enum values.
21361         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
21362         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
21363         * config/nds32/nds32.c (nds32_cctl_names): New.
21364         (nds32_print_operand): Handle cache control register names.
21365         * config/nds32/nds32.h (nds32_builtins): New enum values.
21366         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
21367         macros.
21368         * config/nds32/nds32.md (type): Add mmu.
21369         * config/nds32/pipelines.md (simple_insn): Add mmu.
21371 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21373         * config/nds32/nds32.md (type): Remove call.
21374         * config/nds32/pipelines.md (simple_insn): Likewise.
21376 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21378         * config/nds32/constants.md (unspec_volatile_element): Add
21379         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
21380         UNSPEC_VOLATILE_FMFCFG.
21381         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
21382         description for fmfcfg and fmfcsr.
21383         (bdesc_1arg): Add fmtcsr.
21384         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
21385         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
21386         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
21387         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
21388         unspec_fmfcfg): New patterns.
21389         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
21390         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
21391         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
21392         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
21393         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
21394         __nds32__fmfcfg): Define.
21396 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21398         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
21399         intrinsic register names.
21400         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
21401         intrinsic register enum values and macros.
21403 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21405         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
21406         for load/store addressing form.
21407         (nds32_print_operand_address): Likewise.
21409 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
21411         PR target/85196
21412         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
21413         based on LABEL_REF.  Remove useless assertion.
21414         (pic_address_needs_scratch): Fix formatting.
21415         (sparc_legitimize_pic_address): Minor tweaks.
21416         (sparc_delegitimize_address): Adjust assertion accordingly.
21417         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
21418         into symbolic_operand.
21419         (movsi_high_pic_label_ref): Likewise.
21420         (movsi_lo_sum_pic_label_ref): Likewise.
21421         (movdi_pic_label_ref): Likewise.
21422         (movdi_high_pic_label_ref): Likewise.
21423         (movdi_lo_sum_pic_label_ref): Likewise.
21425 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
21427         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
21428         custom LIB_SPEC setup.
21430 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
21431             Kito Cheng  <kito.cheng@gmail.com>
21433         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
21434         * config/riscv/freebsd.h: New.
21436 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21438         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
21439         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
21440         file.
21442 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21443             Kito Cheng  <kito.cheng@gmail.com>
21445         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
21446         nds32_output_call, nds32_symbol_binds_local_p): New functions.
21447         * config/nds32/nds32-protos.h (nds32_output_call,
21448         nds32_output_return): Declare.
21449         * config/nds32/nds32.md: Refine all the call and return patterns.
21451 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21453         PR debug/85252
21454         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
21455         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
21457         PR rtl-optimization/84872
21458         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
21459         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
21460         EDGE_CROSSING edge.
21462 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
21464         * expr.c (copy_blkmode_to_reg): Revert 254862.
21465         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
21467 2018-04-06  Richard Biener  <rguenther@suse.de>
21469         PR middle-end/85244
21470         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
21471         after seeing a component reference with an adjacent field.  Treat
21472         refs to arrays at struct end of external decls similar to
21473         refs to unconstrained commons.
21475 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21477         PR sanitizer/85213
21478         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
21479         look through SAVE_EXPRs with non-side-effects argument.  Adjust
21480         recursive calls.
21481         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
21482         save_p here.
21484 2018-04-06  Richard Biener  <rguenther@suse.de>
21486         PR middle-end/85180
21487         * alias.c (find_base_term): New wrapper around find_base_term
21488         unwinding CSELIB_VAL_PTR changes.
21489         (find_base_term): Do not restore CSELIB_VAL_PTR during the
21490         recursion.
21492 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21494         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
21495         instructions.
21496         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
21497         constant definitions.
21498         ("nop"): lr 0,0 -> nopr r0
21499         ("nop_lr0", "nop_lr1"): New insn definitions.
21501 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21503         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
21504         NDS32_V3PUSH_AVAILABLE_P macro.
21506 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
21507             Chung-Ju Wu  <jasonwucj@gmail.com>
21509         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
21510         (nds32*-*-*): Add float and fpu_config into supported_defaults.
21511         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
21512         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
21513         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
21514         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
21515         * config/nds32/constraints.md: New constraints and checking for hard
21516         float configuration.
21517         * config/nds32/iterators.md: New mode iterator and attribute for hard
21518         float configuration.
21519         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
21520         patterns.
21521         * config/nds32/nds32-fpu.md: New file.
21522         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
21523         deal with hard float code generation.
21524         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
21525         ARCH_V3S.
21526         (abi_type, float_reg_number): New enum type.
21527         * config/nds32/nds32-predicates.c: New predicates for hard float.
21528         * config/nds32/nds32-protos.h: Declare functions for hard float.
21529         * config/nds32/nds32.c: Implementation for hard float configuration.
21530         * config/nds32/nds32.h: Definitions for hard float configuration.
21531         * config/nds32/nds32.md: Include hard float machine description and
21532         modify patterns for hard float configuration.
21533         * config/nds32/nds32.opt: New options for hard float configuration.
21534         * config/nds32/predicates.md: New predicates for hard float
21535         configuration.
21537 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
21539         * common/config/nds32/nds32-common.c
21540         (nds32_option_optimization_table): Enable -mreleax-hint by default.
21542 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
21544         PR middle-end/85195
21545         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
21546         CONSTRUCTOR_ELT (ctor, ...)->value.
21548 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
21550         PR target/85193
21551         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
21553 2018-04-05  Tom de Vries  <tom@codesourcery.com>
21555         PR target/85204
21556         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
21557         cond jump.
21559 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21560             Kito Cheng  <kito.cheng@gmail.com>
21562         * config/nds32/constraints.md (U33): Fine-tune checking condition.
21563         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
21564         * config/nds32/nds32.h (nds32_16bit_address_type): Add
21565         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
21567 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21568             Kito Cheng  <kito.cheng@gmail.com>
21570         * config/nds32/constraints.md (Ufe): New memory constraint.
21571         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
21572         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
21573         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
21574         operands.
21575         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
21576         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
21578 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21580         * config/nds32/nds32.md: Use optimize_size in the condition for
21581         alu-shift instructions.
21583 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21585         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
21587 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21589         * config/nds32/nds32.md (negsi2): Refine pattern.
21591 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21592             Chung-Ju Wu  <jasonwucj@gmail.com>
21594         * config/nds32/iterators.md (shift_rotate): New code iterator.
21595         (shift): New code attribute.
21596         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
21597         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
21598         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
21599         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
21600         bit-wise operations.
21601         (andsi3, *andsi3): Ditto.
21602         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
21603         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
21604         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
21605         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
21606         nds32_ior_operand, nds32_xor_operand): New predicates.
21608 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21610         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
21611         (addsi3, subsi3): ... this.
21613 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21615         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
21617 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21619         * config/nds32/nds32.md: Adjust indention.
21621 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21623         * config/nds32/nds32.md (feature): New attribute.
21625 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21627         * config/nds32/nds32.md (subtype): New attribute.
21629 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21631         PR target/85203
21632         * config/arm/arm-builtins.c (arm_expand_builtin): Change
21633         expansion to perform a bitwise AND of the argument followed by a
21634         boolean negation of the result.
21636 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
21638         PR rtl-optimization/84878
21639         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
21640         the basic block.  Assert the use reference is not artificial and that
21641         it has an associated insn.
21643 2018-04-04  Michael Matz  <matz@suse.de>
21645         * builtins.c (compute_objsize): Pass correct operand
21646         to array_at_struct_end_p.
21648 2018-04-04  Richard Biener  <rguenther@suse.de>
21650         PR lto/85176
21651         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
21652         from contexts for DINFO_LEVEL_TERSE and below.
21654 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21656         * config/nds32/nds32-doubleword.md (move_<mode>): Require
21657         resiter_operand condition.
21658         * config/nds32/nds32.md (*move<mode>): Ditto.
21660 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21661             Monk Chiang  <sh.chiang04@gmail.com>
21663         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
21665 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21667         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
21669 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21670             Kito Cheng  <kito.cheng@gmail.com>
21672         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
21673         nds32_cond_code_str, output_cond_branch,
21674         output_cond_branch_compare_zero, nds32_expand_cbranch,
21675         nds32_expand_cstore, nds32_expand_movcc,
21676         nds32_output_cbranchsi4_equality_zero,
21677         nds32_output_cbranchsi4_equality_reg,
21678         nds32_output_cbranchsi4_equality_reg_or_const_int,
21679         nds32_output_cbranchsi4_greater_less_zero: New functions.
21680         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
21681         nds32_expand_cstore, nds32_expand_movcc,
21682         nds32_output_cbranchsi4_equality_zero,
21683         nds32_output_cbranchsi4_equality_reg,
21684         nds32_output_cbranchsi4_equality_reg_or_const_int,
21685         nds32_output_cbranchsi4_greater_less_zero): Declare.
21686         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
21687         nds32_rimm11s_operand): New predicates.
21688         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
21689         * config/nds32/nds32.md: Rewrite all the branch and conditional move
21690         patterns.
21692 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21694         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
21695         * config/nds32/nds32.md: Ditto.
21696         * config/nds32/pipelines.md: Ditto.
21698 2018-04-04  Richard Biener  <rguenther@suse.de>
21700         PR tree-optimization/85168
21701         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
21702         propagating abnormals.
21704 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21706         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
21708 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21709             Kito Cheng  <kito.cheng@gmail.com>
21711         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
21712         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
21713         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
21714         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
21715         * config/nds32/nds32.md (sibcall_internal): New.
21716         (sibcall_register): Remove.
21717         (sibcall_immediate): Remove.
21718         (sibcall_value_internal): New.
21719         (sibcall_value_register): Remove.
21720         (sibcall_value_immediate): Remove.
21721         * config/nds32/predicates.md (nds32_general_register_operand): New.
21722         (nds32_call_address_operand): New.
21724 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21726         PR rtl-optimization/85167
21727         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
21728         bb_defs if *split_p, instead preinitialize it to NULL.
21730         PR tree-optimization/85156
21731         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
21732         evaluating the argument multiple times.
21734 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
21736         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
21737         than vector.
21738         (_mm_cvtpd_ps): Likewise.
21739         (_mm_cvttpd_epi32): Likewise.
21740         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
21741         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
21742         vector, pixel, and bool following altivec.h include.
21744 2018-04-03  Martin Sebor  <msebor@redhat.com>
21746         * doc/extend.texi (Common Function Attributes): Clarify.
21747         (const attribute): Likewise.
21748         (pure attribute): Likewise.
21750 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21752         PR target/85169
21753         * config/i386/i386.c (ix86_expand_vector_set): Use
21754         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
21756 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
21758         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
21759         instructions when changing rounding bits to preserve precision bits
21760         in the x87 control word.
21762 2018-04-03  Martin Liska  <mliska@suse.cz>
21764         PR tree-optimization/82491
21765         * rtl.h (strip_offset_and_add): Replace += suboffset with
21766         poly_uint64 () + suboffset.
21768 2018-03-29  Martin Liska  <mliska@suse.cz>
21769             Martin Jambor  <mjambor@suse.cz>
21771         PR ipa/84947
21772         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
21773         param_type is not an integral or pointer type.
21775 2018-04-03  Richard Biener  <rguenther@suse.de>
21777         * sese.h (recompute_all_dominators): Remove.
21779 2018-04-02  Martin Sebor  <msebor@redhat.com>
21781         * doc/invoke.texi (-Wrestrict): Fix typos.
21783 2018-04-02  Jim Wilson  <jimw@sifive.com>
21785         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
21786         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
21787         (<optab>di3, <optab>si3_extend): Likewise.
21788         (<optab>si3_mask, <optab>si3_mask_1): New.
21789         (<optab>di3_mask, <optab>di3_mask_1): New.
21790         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
21791         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
21792         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
21794 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
21796         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
21797         example.
21799 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
21801         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
21802         (nds32_canonicalize_comparison): New function.
21804 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21805             Kito Cheng  <kito.cheng@gmail.com>
21806             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
21808         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
21809         * config/nds32/constants.md (unspec_volatile_element): Add
21810         UNSPEC_VOLATILE_RELAX_GROUP.
21811         * config/nds32/nds32-relax-opt.c: New file.
21812         * config/nds32/nds32-predicates.c
21813         (nds32_symbol_load_store_p): New function.
21814         * config/nds32/nds32-protos.h
21815         (nds32_symbol_load_store_p): Declare function.
21816         (make_pass_nds32_relax_opt): Declare new rtl pass function.
21817         * config/nds32/nds32.c
21818         (nds32_register_pass): New function to register pass.
21819         (nds32_register_passes): New function to register passes.
21820         * config/nds32/nds32.md (relax_group): New pattern.
21821         * config/nds32/nds32.opt (mrelax-hint): New option.
21822         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
21824 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
21826         * config/nds32/t-nds32: Modify files dependency.
21828 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21830         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
21831         (PROFILE_HOOK): Define its implementation.
21833 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21835         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
21836         type and 32-bit size.
21838 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
21840         PR middle-end/85090
21841         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
21842         (V_128_256): New mode iterator.
21843         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
21844         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
21845         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
21846         of V.
21847         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
21848         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
21850 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
21852         PR target/83315
21853         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
21854         NaN inputs correctly.
21856 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
21858         PR target/80546
21859         * config/rs6000/vsx.md (??r): New mode attribute.
21860         (*vsx_mov<mode>_64bit): Use it.
21861         (*vsx_mov<mode>_32bit): Likewise.
21863 2018-03-30  Martin Sebor  <msebor@redhat.com>
21865         PR tree-optimization/84818
21866         * builtins.c (check_access): Use warning_n.
21868 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21870         PR target/83822
21871         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
21872         condition.
21873         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
21874         condition.
21876 2018-03-30  Julia Koval  <julia.koval@intel.com>
21878         PR target/84413
21879         * x86-tune.def (movx, partial_reg_dependency): Enable for
21880         m_SKYLAKE_AVX512.
21882 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
21884         PR inline-asm/84985
21885         * lra-constraints.c (process_alt_operands): Move setting
21886         this_alternative_matches below.
21888 2018-03-29  Martin Liska  <mliska@suse.cz>
21890         PR lto/84995.
21891         * doc/invoke.texi: Document how LTO works with debug info.
21892         Describe auto-load support of binutils.  Mention 'x86-64'
21893         as valid option value of -march option.
21895 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
21897         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
21899         PR c/85094
21900         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
21901         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
21902         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
21903         checking.
21905 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
21907         PR target/84912
21908         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
21909         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
21910         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
21911         for RS6000_BTM_POWERPC64.
21912         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
21913         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
21914         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
21915         definition.
21916         (DIVDE): Use it.
21917         (DIVDEU): Likewise.
21919 2018-03-28  Carl Love  <cel@us.ibm.com>
21921         Revert
21922         2017-09-27  Carl Love  <cel@us.ibm.com>
21924         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
21925         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
21926         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
21927         fctiw instruction.
21929 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21931         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
21932         instead of __vector bool.
21933         (_mm_max_pu8): Likewise.
21934         (_mm_min_pi16): Likewise.
21936 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
21938         PR target/84912
21939         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
21940         (DIVWEUO): Likewise.
21941         (DIVDEO): Likewise.
21942         (DIVDEUO): Likewise.
21943         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
21944         DIVWEUO and DIVDEUO.
21945         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
21946         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
21947         (div_extend): Likewise.
21948         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
21949         builtin function.
21950         (__builtin_divweuo): Likewise.
21951         (__builtin_divdeo): Likewise.
21952         (__builtin_divdeuo): Likewise.
21954 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
21956         PR target/85095
21957         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
21958         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
21960         PR tree-optimization/82004
21961         * gimple-match-head.c (optimize_pow_to_exp): New function.
21962         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
21963         Don't fold to exp if optimize_pow_to_exp is false.
21965 2018-03-28  Martin Liska  <mliska@suse.cz>
21967         PR other/84819
21968         * calls.c (initialize_argument_information): Fix trailing space.
21969         * common.opt: Fix typo and provide better explanation for
21970         -fsanitize-coverage option.
21971         * config/i386/i386.opt: Fix typo.
21973 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
21974             Martin Liska  <mliska@suse.cz>
21976         PR sanitizer/85081
21977         * gimplify.c (asan_poison_variable): Don't do the check for
21978         gimplify_omp_ctxp here.
21979         (gimplify_decl_expr): Do it here.
21980         (gimplify_target_expr): Likewise.
21982 2018-03-28  Martin Liska  <mliska@suse.cz>
21984         PR target/84988
21985         * config/i386/i386.c (ix86_function_arg_advance): Do not call
21986         chkp_type_bounds_count if MPX is not enabled.
21988 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21990         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
21992 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
21994         PR target/84914
21995         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
21996         function to create the function decl for complex long double
21997         multiply and divide for -mabi=ieeelongdouble.
21998         (init_float128_ieee): Call it.
22000 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22002         PR target/85044
22003         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22004         -fcf-protection=branch -mibt.
22005         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22007 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22009         PR target/81863
22010         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22011         arm_word_relocations.
22013 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22015         PR target/85056
22016         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22017         extern array declarations.
22019 2018-03-27  Richard Biener  <rguenther@suse.de>
22021         PR middle-end/84067
22022         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22023         explicit single_use checks.
22025 2018-03-27  Richard Biener  <rguenther@suse.de>
22027         PR tree-optimization/85082
22028         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22029         Valueize the VUSE.
22031 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22033         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22034         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22035         Turn on fasynchronous-unwind-tables and funwind-tables.
22037 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22039         PR target/85073
22040         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22041         (*bmi_blsr_<mode>_ccz): Ditto.
22043 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22045         PR tree-optimization/85063
22046         * omp-general.c (offloading_function_p): New function.  Factor out
22047         of ...
22048         * omp-offload.c (pass_omp_target_link::gate): ... here.
22049         * omp-general.h (offloading_function_p): Declare.
22050         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22051         with attribute omp declare target for offloading functions.
22053 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
22055         PR tree-optimization/84005
22056         * tree-data-ref.h (get_base_for_alignment): Declare.
22057         * tree-data-ref.c (get_base_for_alignment_1): New function.
22058         (get_base_for_alignment): Likewise.
22059         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22060         get_base_for_alignment to find a suitable base object, instead
22061         of always using drb->base_address.
22063 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22065         PR inline-asm/85022
22066         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22067         known size by default.
22069 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
22071         PR inline-asm/85030
22072         * lra-constraints.c (process_alt_operands): Don't match BLKmode
22073         and non BLKmode operands.
22075 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22077         PR target/85026
22078         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22079         Clean up attributes.
22081 2018-03-23  Richard Biener  <rguenther@suse.de>
22083         PR debug/85020
22084         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22085         we are going to emit early debug for LTO.
22087 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
22089         PR inline-asm/85034
22090         * function.c (match_asm_constraints_1): Don't optimize if input
22091         doesn't satisfy general_operand predicate for output's mode.
22093         PR inline-asm/85022
22094         * alias.c (write_dependence_p): Don't require for x_canonicalized
22095         non-VOIDmode if x has VOIDmode.
22097         PR sanitizer/85029
22098         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22099         just don't try to optimize it rather than assert it never happens.
22101 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22103         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22104         macro expansions for definition of ST_INTERNAL_<mode> and
22105         LD_INTERNAL_<mode> builtins.
22106         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22107         Remove prototype.
22108         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22109         function.
22110         (altivec_expand_st_builtin): Likewise.
22111         (altivec_expand_builtin): Remove calls to deleted functions.
22112         (rs6000_address_for_altivec): Delete this function.
22113         * config/rs6000/vector.md: Remove expands for
22114         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22116 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
22118         PR target/84826
22119         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22120         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22121         re-computing once computed.
22122         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22123         (arm_init_machine_status): Initialize
22124         machine->static_chain_stack_bytes.
22126 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22128         PR target/84760
22129         * doc/extend.texi: Add four new prototypes for vec_ld.
22130         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22131         definitions for more logical presentation.
22132         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22133         entries for V1TI variants of __builtin_altivec_ld builtin.
22134         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22135         handling of V1TI variant of LVX icode pattern.
22136         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22137         (rs6000_gimple_fold_builtin): Likewise.
22138         (altivec_init_builtins): Add code to define
22139         __builtin_altivec_lvx_v1ti function.
22141 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22143         PR inline-asm/84941
22144         * function.c (match_asm_constraints_1): Don't do the optimization
22145         if input isn't a REG, SUBREG, MEM or constant.
22147 2018-03-22  Tom de Vries  <tom@codesourcery.com>
22149         PR tree-optimization/84956
22150         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22151         bb_has_abnormal_pred.
22153 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
22155         PR sanitizer/85018
22156         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22157         DECL_INITIAL (decl) to decl at the end.
22158         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22159         adjust the comment.
22161 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22163         * doc/extend.texi (__builtin_tgmath): Document when complex
22164         integer types are treated as _Complex _Float64.
22166 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22168         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22170 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22172         PR tree-optimization/84960
22173         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22174         if it is ENTRY block, move them into single succ of ENTRY in that case.
22176 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
22178         PR tree-optimization/84811
22179         * poly-int.h (poly_span_traits): Remove the T3 parameter and
22180         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22181         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22182         (known_subrange_p): Update accordingly.  Cast each value involved
22183         in the size comparison, rather than casting the result of the
22184         subtraction.
22186 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22188         PR tree-optimization/84982
22189         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22190         by flipping the least significant bit rather than all bits from
22191         bitpos to bitpos + bitsize - 1.
22193 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22195         * doc/extend.texi (Deprecated Features): Remove mention of
22196         long-deleted deprecations.
22198 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22200         PR jit/84288
22201         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22202         * configure: Regenerate.
22204 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22206         PR tree-optimization/83126
22207         * tree-parloops.c (num_phis): New function.
22208         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22210 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22212         * doc/extend.texi (Deprecated Features): Update deprecated flags,
22213         mention anon-struct/union members and trailing attributes.
22215 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
22217         PR tree-optimization/84969
22218         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22219         builtin memset partitions if they set different rhs values.
22221 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22223         PR rtl-optimization/84989
22224         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22225         VEC_DUPLICATE with scalar result mode.
22227 2018-03-21  Martin Liska  <mliska@suse.cz>
22229         PR ipa/84963
22230         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22231         not intended return statement.
22233 2018-03-21  Martin Liska  <mliska@suse.cz>
22235         PR target/84988
22236         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22237         (chkp_find_bound_slots_1): Limit number of iterations.
22239 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
22241         PR target/84838
22242         * Minor grammar fixes for x86 options.
22244 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22246         PR debug/84875
22247         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22248         holding REG_CFA_RESTORE notes, instead turn them into a USE.
22250 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
22252         PR target/83789
22253         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22254         (altivec_lvx_<mode>_1op): Likewise.
22255         (altivec_stvx_<mode>_2op): Likewise.
22256         (altivec_stvx_<mode>_1op): Likewise.
22257         (altivec_lvx_<VM2:mode>): New define_expand.
22258         (altivec_stvx_<VM2:mode>): Likewise.
22259         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22260         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22261         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22262         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22263         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22264         (rs6000_gen_lvx): Likewise.
22265         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22266         (altivec_expand_stv_builtin): Likewise.
22267         (altivec_expand_builtin): Likewise.
22268         * config/rs6000/vector.md: Likewise.
22270 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22272         PR target/82518
22273         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22274         BYTES_BIG_ENDIAN.
22276 2018-03-20  Richard Biener  <rguenther@suse.de>
22278         PR target/84986
22279         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22280         sign-conversions as zero, fall back to standard scalar_stmt
22281         cost for the rest.
22283 2018-03-20  Martin Liska  <mliska@suse.cz>
22285         PR ipa/84825
22286         * predict.c (rebuild_frequencies): Handle case when we have
22287         PROFILE_ABSENT, but flag_guess_branch_prob is false.
22289 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22291         PR target/84990
22292         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22293         flag_section_anchors.
22294         * varasm.c (use_blocks_for_decl_p): Remove hack for
22295         dw2_force_const_mem.
22297         PR target/84845
22298         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22299         to ...
22300         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
22301         be created, use lowpart_subreg of operands[0] rather than operands[0]
22302         itself.
22303         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22304         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22305         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22306         and n constraint instead of aarch64_shift_imm_di and Usd.
22307         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22308         (*aarch64_<optab>_reg_minus<mode>3): ... this.
22310 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
22312         PR target/82989
22313         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
22314         to favor GPR over NEON registers.
22315         (<shift>di3_neon): Likewise.
22317 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22319         PR target/84952
22320         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
22321         (nvptx_process_pars): Emit bar.sync asap and alap.
22323 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22325         PR target/84954
22326         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
22327         seen_label if seen_label is already set.
22329 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22331         PR target/84945
22332         * config/i386/i386.c (fold_builtin_cpu): For features above 31
22333         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
22334         Use 1U instead of 1.  Formatting fixes.
22336         PR c/84953
22337         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
22338         instead of TREE_TYPE (s1) for the return value.
22340 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
22342         PR tree-optimization/84946
22343         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
22344         bitsize + bitsize in poly_uint64 rather than poly_int64.
22346         PR sanitizer/78651
22347         * dwarf2asm.c: Include fold-const.c.
22348         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
22349         of decl rather than decl itself.
22351         PR rtl-optimization/84643
22352         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
22354 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
22356         PR sanitizer/78651
22357         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
22358         calling assemble_variable.
22360 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
22362         PR target/81647
22363         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
22364         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
22366 2018-03-19  Jim Wilson  <jimw@sifive.com>
22368         PR bootstrap/84856
22369         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
22370         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
22371         (riscv_first_stack_step): Likewise.
22372         (riscv_option_override): Use STACK_BOUNDARY instead of
22373         MIN_STACK_BOUNDARY.
22374         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
22375         MIN_STACK_BOUNDARY.
22376         (BIGGEST_ALIGNMENT): Set to 128.
22377         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
22378         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22379         STACK_BOUNDARY.
22381 2018-03-19  Richard Biener  <rguenther@suse.de>
22383         PR tree-optimization/84933
22384         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
22385         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
22387 2018-03-19  Richard Biener  <rguenther@suse.de>
22389         PR tree-optimization/84859
22390         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
22391         (cond_if_else_store_replacement): Perform sinking operation on
22392         single-store BBs regardless of MAX_STORES_TO_SINK setting.
22393         Generalize what a BB with a single eligible store is.
22395 2018-03-19  Richard Biener  <rguenther@suse.de>
22397         PR tree-optimization/84929
22398         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
22399         chrec_is_positive against non-chrec arg.
22401 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
22403         PR target/84711
22404         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
22406 2018-03-18  Martin Liska  <mliska@suse.cz>
22408         PR rtl-optimization/84635
22409         * regrename.c (build_def_use): Use matches_mode only when
22410         matches >= 0.
22412 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
22414         PR tree-optimization/84913
22415         * tree-vect-loop.c (vectorizable_reduction): Don't try to
22416         vectorize chains of COND_EXPRs.
22418 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22420         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
22422 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22424         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
22426 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22428         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
22430 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22431             Kito Cheng  <kito.cheng@gmail.com>
22433         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
22434         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
22435         (nds32_adjust_reg_alloc_order): New function.
22436         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
22438 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
22440         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
22441         nds32_print_operand, nds32_print_operand_address): Use
22442         HOST_WIDE_INT_PRINT_DEC instead.
22444 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22446         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
22448 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
22450         PR target/84902
22451         * config/i386/i386.c (initial_ix86_tune_features,
22452         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
22453         unsigned long long.
22454         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
22455         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
22456         rather than 1u << ix86_tune.  Formatting fix.
22457         (ix86_option_override_internal): Change ix86_arch_mask from
22458         unsigned int to unsigned HOST_WIDE_INT, initialize to
22459         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
22460         (ix86_function_specific_restore): Likewise.
22462 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22464         PR target/84899
22465         * postreload.c (reload_combine_recognize_pattern): Perform
22466         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
22467         truncate_int_for_mode the result for the destination's mode.
22469         PR c/84909
22470         * hsa-gen.c (mem_type_for_type): Fix comment typo.
22471         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
22472         Likewise.
22473         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22474         Likewise.
22476 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
22478         PR target/84876
22479         * lra-assigns.c (lra_split_hard_reg_for): Don't use
22480         regno_allocno_class_array and sorted_pseudos.
22481         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
22482         insns where regno is used.
22484 2018-03-16  Martin Liska  <mliska@suse.cz>
22486         PR ipa/84833
22487         * multiple_target.c (create_dispatcher_calls): Redirect
22488         reference in the symbol table.
22490 2018-03-16  Martin Liska  <mliska@suse.cz>
22492         PR ipa/84722
22493         * multiple_target.c (create_dispatcher_calls): Redirect also
22494         an alias.
22496 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22498         PR c++/79937
22499         PR c++/82410
22500         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
22501         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
22502         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
22504 2018-03-16  Julia Koval  <julia.koval@intel.com>
22506         * doc/invoke.texi (Skylake Server): Add CLWB.
22507         Cannonlake): Remove CLWB.
22509 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22511         PR tree-optimization/84841
22512         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
22513         1 << 3.
22514         (FLOAT_ONE_CONST_TYPE): Define.
22515         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
22516         (sort_by_operand_rank): Put entries with higher constant_type last
22517         rather than first to match comments.
22519 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
22521         * config/nios2/nios2.md (movsi_internal): Fix thinko in
22522         split predicate.
22524 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22526         PR c++/79085
22527         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
22528         check and use address of target always.
22530 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
22532         PR target/84574
22533         * config/i386/i386.c (indirect_thunk_needed): Update comments.
22534         (indirect_thunk_bnd_needed): Likewise.
22535         (indirect_thunks_used): Likewise.
22536         (indirect_thunks_bnd_used): Likewise.
22537         (indirect_return_needed): New.
22538         (indirect_return_bnd_needed): Likewise.
22539         (output_indirect_thunk_function): Add a bool argument for
22540         function return.
22541         (output_indirect_thunk_function): Don't generate alias for
22542         function return thunk.
22543         (ix86_code_end): Call output_indirect_thunk_function to generate
22544         function return thunks.
22545         (ix86_output_function_return): Set indirect_return_bnd_needed
22546         and indirect_return_needed instead of indirect_thunk_bnd_needed
22547         and indirect_thunk_needed.
22549 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
22551         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
22552         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
22553         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
22555 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
22556             Paul Hua <paul.hua.gm@gmail.com>
22558         PR c/84852
22559         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
22561 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
22563         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
22564         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
22565         resp. SFmode cases.
22567 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
22569         PR target/84711
22570         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
22571         instead of GET_MODE_SIZE when comparing Units.
22573 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
22575         PR target/68256
22576         * varasm.c (hash_section): Return an unchangeble hash value
22577         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
22578         Return !aarch64_can_use_per_function_literal_pools_p ().
22580 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22582         PR target/84860
22583         * optabs.c (emit_conditional_move): Pass address of cmode's copy
22584         rather than address of cmode as last argument to prepare_cmp_insn.
22586 2018-03-15  Julia Koval  <julia.koval@intel.com>
22588         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
22589         F_AVX512VNNI, F_AVX512BITALG): New.
22591 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
22593         PR target/83451
22594         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
22595         insn for floating-point loads and stores.
22597 2018-03-14  Carl Love  <cel@us.ibm.com>
22599         * config/rs6000/rs6000-c.c: Add macro definitions for
22600         ALTIVEC_BUILTIN_VEC_PERMXOR.
22601         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
22602         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
22603         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
22604         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
22605         UNSPEC_VPERMXOR.
22606         * config/doc/extend.texi: Add prototypes for vec_permxor.
22608 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
22610         PR c/84852
22611         * diagnostic-show-locus.c (class layout_point): Convert m_line
22612         from int to linenum_type.
22613         (line_span::comparator): Use linenum "compare" function when
22614         comparing line numbers.
22615         (test_line_span): New function.
22616         (layout_range::contains_point): Convert param "row" from int to
22617         linenum_type.
22618         (layout_range::intersects_line_p): Likewise.
22619         (layout::will_show_line_p): Likewise.
22620         (layout::print_source_line): Likewise.
22621         (layout::should_print_annotation_line_p): Likewise.
22622         (layout::print_annotation_line): Likewise.
22623         (layout::print_leading_fixits): Likewise.
22624         (layout::annotation_line_showed_range_p): Likewise.
22625         (struct line_corrections): Likewise for field m_row.
22626         (line_corrections::line_corrections): Likewise for param "row".
22627         (layout::print_trailing_fixits): Likewise.
22628         (layout::get_state_at_point): Likewise.
22629         (layout::get_x_bound_for_row): Likewise.
22630         (layout::print_line): Likewise.
22631         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
22632         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
22633         * input.c (selftest::test_linenum_comparisons): New function.
22634         (selftest::input_c_tests): Call it.
22635         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
22636         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
22637         * selftest.h (ASSERT_GT): New macro.
22638         (ASSERT_GT_AT): New macro.
22639         (ASSERT_LT): New macro.
22640         (ASSERT_LT_AT): New macro.
22642 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
22644         PR rtl-optimization/84780
22645         * combine.c (distribute_links): Don't make a link based on pc_rtx.
22647 2018-03-14  Martin Liska  <mliska@suse.cz>
22649         * tree.c (record_node_allocation_statistics): Use
22650         get_stats_node_kind.
22651         (get_stats_node_kind): New function extracted from
22652         record_node_allocation_statistics.
22653         (free_node): Use get_stats_node_kind.
22655 2018-03-14  Richard Biener  <rguenther@suse.de>
22657         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
22658         that the value-set of ANTIC_IN doesn't grow.
22660         Revert
22661         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22662         member.
22663         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22664         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22666 2018-03-14  Julia Koval  <julia.koval@intel.com>
22668         * config.gcc (icelake-client, icelake-server): New.
22669         (icelake): Remove.
22670         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
22671         (initial_ix86_arch_features): Ditto.
22672         (PTA_SKYLAKE): Add SGX.
22673         (PTA_ICELAKE): Remove.
22674         (PTA_ICELAKE_CLIENT): New.
22675         (PTA_ICELAKE_SERVER): New.
22676         (ix86_option_override_internal): Split up icelake on icelake client and
22677         icelake server.
22678         (get_builtin_code_for_version): Ditto.
22679         (fold_builtin_cpu): Ditto.
22680         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
22681         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
22682         * config/i386/i386.h (processor_type): Ditto.
22683         * doc/invoke.texi: Ditto.
22685 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
22687         PR sanitizer/83392
22688         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
22689         INTEGER_CST offset, add it together with bitpos / 8 and
22690         sign extend based on POINTER_SIZE.
22692         PR target/84844
22693         Revert
22694         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
22696         PR target/78090
22697         * config/i386/constraints.md (Yc): New register constraint.
22698         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
22699         Use Yc constraint for alternative 2 of operand 0.  Remove
22700         preferred_for_speed attribute.
22702 2018-03-14  Richard Biener  <rguenther@suse.de>
22704         PR tree-optimization/84830
22705         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
22706         with the old one to avoid oscillations.
22708 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
22710         PR target/83712
22711         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
22712         pseudos.
22713         (assign_by_spills): Return a flag of reload assignment failure.
22714         Do not process the reload assignment failures.  Do not spill other
22715         reload pseudos if they has the same reg class.  Update n if
22716         necessary.
22717         (lra_assign): Add a return arg.  Set up from the result of
22718         assign_by_spills call.
22719         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22720         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22721         usage_insns if it is not NULL.
22722         (spill_hard_reg_in_range): New function.
22723         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22724         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22725         function prototypes.
22726         (lra_assign): Change prototype.
22727         * lra.c (lra): Add code to deal with fails by splitting hard reg
22728         live ranges.
22730 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
22732         * config/riscv/riscv.opt (mrelax): New option.
22733         * config/riscv/riscv.c (riscv_file_start): Emit ".option
22734         "norelax" when riscv_mrelax is disabled.
22735         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
22737 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22739         PR target/84743
22740         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
22741         reassociation for int modes.
22743 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22745         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
22746         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
22747         for big-endian.
22748         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
22749         * config/aarch64/aarch64-sve.md
22750         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
22751         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
22752         (*extend<mode><Vwide>2): Rename to...
22753         (aarch64_sve_extend<mode><Vwide>2): ...this.
22754         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
22755         renaming the old pattern to...
22756         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
22757         unsigned packs.
22758         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
22759         define_expand, renaming the old pattern to...
22760         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
22761         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
22762         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
22763         account when deciding which SVE instruction the optab should use.
22764         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
22766 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22768         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
22769         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
22770         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
22771         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
22772         (tlsdesc_small_<mode>): Turn a define_expand and use
22773         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
22774         (tlsdesc_small_advsimd_<mode>): ...this.
22775         (tlsdesc_small_sve_<mode>): New pattern.
22777 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22779         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
22780         (UNSPEC_UMUL_HIGHPART): New constants.
22781         (MUL_HIGHPART): New int iteraor.
22782         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
22783         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
22784         define_expand.
22785         (*<su>mul<mode>3_highpart): New define_insn.
22787 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
22789         PR lto/84805
22790         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
22791         incomplete types.
22793 2018-03-13  Martin Liska  <mliska@suse.cz>
22795         PR ipa/84658.
22796         * (sem_item_optimizer::sem_item_optimizer): Initialize new
22797         vector.
22798         (sem_item_optimizer::~sem_item_optimizer): Release it.
22799         (sem_item_optimizer::merge_classes): Register variable aliases.
22800         (sem_item_optimizer::fixup_pt_set): New function.
22801         (sem_item_optimizer::fixup_points_to_sets): Likewise.
22802         * ipa-icf.h: Declare new variables and functions.
22804 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
22806         PR middle-end/84834
22807         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
22808         integer_pow2p@2 and test integer_pow2p in condition.
22809         (A < 0 ? C : 0): Similarly for @1.
22811         PR middle-end/84831
22812         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
22813         characters starting at p contain '\0' character, don't look beyond
22814         that.
22816         PR target/84827
22817         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
22818         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
22820         PR target/84828
22821         * reg-stack.c (change_stack): Change update_end var from int to
22822         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
22823         also call set_block_for_insn on the newly added insns and rescan.
22825         PR target/84786
22826         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
22827         on the last operand.
22829         PR c++/84704
22830         * tree.c (stabilize_reference_1): Return save_expr (e) for
22831         STATEMENT_LIST even if it doesn't have side-effects.
22833 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
22835         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
22837 2018-03-12  Renlin Li  <renlin.li@arm.com>
22839         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
22840         aarch64_output_scalar_simd_mov_immediate.
22842 2018-03-12  Martin Sebor  <msebor@redhat.com>
22844         PR tree-optimization/83456
22845         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
22846         for perfectly overlapping calls to memcpy.
22847         (gimple_fold_builtin_memory_chk): Same.
22848         (gimple_fold_builtin_strcpy): Handle no-warning.
22849         (gimple_fold_builtin_stxcpy_chk): Same.
22850         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
22852 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
22854         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
22855         parameter.  Use it for SFmode.
22856         (rs6000_function_arg_advance_1): Adjust.
22857         (rs6000_function_arg): Adjust.
22858         (rs6000_gimplify_va_arg): Pass false for that new parameter.
22860 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
22862         PR rtl-optimization/84169
22863         PR rtl-optimization/84780
22864         * combine.c (can_combine_p): Check for a 2-insn combination whether
22865         the destination register is used between the two insns, too.
22867 2018-03-12  Richard Biener  <rguenther@suse.de>
22869         PR tree-optimization/84803
22870         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
22871         for refs DR analysis didn't process.
22873 2018-03-12  Richard Biener  <rguenther@suse.de>
22875         PR tree-optimization/84777
22876         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
22877         force-vectorize loops ignore whether we are optimizing for size.
22879 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
22881         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
22882         (TARGET_MD_ASM_ADJUST): Define.
22884 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
22885             Kito Cheng  <kito.cheng@gmail.com>
22886             Chung-Ju Wu  <jasonwucj@gmail.com>
22888         * config/nds32/nds32.c (nds32_compute_stack_frame,
22889         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
22890         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
22891         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
22892         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
22893         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
22894         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
22895         * config/nds32/nds32.md (prologue, epilogue): Use macro
22896         NDS32_V3PUSH_AVAILABLE_P to do checking.
22898 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
22900         PR debug/58150
22901         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
22902         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
22903         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
22904         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
22905         addition of most attributes on !orig_type_die or the attribute not
22906         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
22908 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22909             Chung-Ju Wu  <jasonwucj@gmail.com>
22911         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
22912         __NDS32_VH__ macro.
22913         * config/nds32/nds32.opt (mvh): New option.
22915 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22916             Chung-Ju Wu  <jasonwucj@gmail.com>
22918         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
22919         function.
22920         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
22921         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
22922         definition.
22924 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22925             Chung-Ju Wu  <jasonwucj@gmail.com>
22927         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
22928         function.
22929         * config/nds32/nds32-multiple.md (strlensi): New pattern.
22930         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
22932 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
22933             Kito Cheng  <kito.cheng@gmail.com>
22934             Chung-Ju Wu  <jasonwucj@gmail.com>
22936         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
22937         UNSPEC_FFMISM and UNSPEC_FLMISM.
22938         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
22939         for ffb, ffmism and flmism.
22940         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
22941         (unspec_ffmism): Ditto.
22942         (unspec_flmism): Ditto.
22943         (nds32_expand_builtin_impl): Check if string extension is available.
22944         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
22945         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
22947 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
22949         Reverting patch:
22950         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
22952         PR target/83712
22953         * lra-assigns.c (assign_by_spills): Return a flag of reload
22954         assignment failure.  Do not process the reload assignment
22955         failures.  Do not spill other reload pseudos if they has the same
22956         reg class.
22957         (lra_assign): Add a return arg.  Set up from the result of
22958         assign_by_spills call.
22959         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22960         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22961         usage_insns if it is not NULL.
22962         (spill_hard_reg_in_range): New function.
22963         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22964         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22965         function prototypes.
22966         (lra_assign): Change prototype.
22967         * lra.c (lra): Add code to deal with fails by splitting hard reg
22968         live ranges.
22970 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
22972         PR target/84807
22973         * config/i386/i386.opt: Replace Enforcment with Enforcement.
22975 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
22977         PR debug/84620
22978         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
22979         (dw_val_node): Add val_symbolic_view.
22980         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
22981         (symview_upper_bound): New.
22982         (new_line_info_table): Initialize symviews_since_reset.
22983         (dwarf2out_source_line): Count symviews_since_reset and set
22984         symview_upper_bound.
22985         (dw_val_equal_p): Handle symview.
22986         (add_AT_symview): New.
22987         (print_dw_val): Handle symview.
22988         (attr_checksum, attr_checksum_ordered): Likewise.
22989         (same_dw_val_p, size_of_die): Likewise.
22990         (value_format, output_die): Likewise.
22991         (add_high_low_attributes): Use add_AT_symview for entry_view.
22992         (dwarf2out_finish): Reset symview_upper_bound, clear
22993         zero_view_p.
22995 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
22997         PR target/83969
22998         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
22999         Add strict argument and use it.
23000         (rs6000_split_multireg_move): Update for new strict argument.
23001         (mem_operand_gpr): Disallow all non-offsettable addresses.
23002         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23004 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23006         PR target/84772
23007         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23008         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23009         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23011         PR c++/84767
23012         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23013         decl, use remap_type if we want to use the type.
23015 2018-03-09  Martin Sebor  <msebor@redhat.com>
23017         PR tree-optimization/84526
23018         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23019         Remove dead code.
23020         (builtin_access::generic_overlap): Be prepared to handle non-array
23021         base objects.
23023 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23025         PR rtl-optimization/84682
23026         * lra-constraints.c (process_address_1): Check is_address flag
23027         for address constraints.
23028         (process_alt_operands): Likewise.
23029         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23030         preprocess_constraints.
23031         * recog.h (preprocess_constraints): Add oploc parameter.
23032         Adjust callers.
23033         * recog.c (preprocess_constraints): Test address_operand for
23034         CT_ADDRESS constraints.
23036 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23038         PR target/83712
23039         * lra-assigns.c (assign_by_spills): Return a flag of reload
23040         assignment failure.  Do not process the reload assignment
23041         failures.  Do not spill other reload pseudos if they has the same
23042         reg class.
23043         (lra_assign): Add a return arg.  Set up from the result of
23044         assign_by_spills call.
23045         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23046         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23047         usage_insns if it is not NULL.
23048         (spill_hard_reg_in_range): New function.
23049         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23050         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23051         function prototypes.
23052         (lra_assign): Change prototype.
23053         * lra.c (lra): Add code to deal with fails by splitting hard reg
23054         live ranges.
23056 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23058         PR target/83193
23059         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23060         Accept complain bool parameter.  Only emit errors if it is true.
23061         (arm_parse_cpu_option_name): Likewise.
23062         (arm_target_thumb_only): Adjust callers of the above.
23063         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23064         prototype to take a default true bool parameter.
23065         (arm_parse_arch_option_name): Likewise.
23067 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
23068             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
23070         PR jit/64089
23071         PR jit/84288
23072         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23073         * configure: Regenerate.
23074         * configure.ac ("linker --version-script option"): New.
23075         ("linker soname option"): New.
23077 2018-03-09  Richard Biener  <rguenther@suse.de>
23079         PR tree-optimization/84775
23080         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23081         immediate uses of predicate stmts and mark them modified.
23083         Revert
23084         PR tree-optimization/84178
23085         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23086         to caller.
23087         (version_loop_for_if_conversion): Delay update_ssa call.
23088         (tree_if_conversion): Delay update_ssa until after predicate
23089         insertion.
23091 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
23093         PR target/84763
23094         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23095         when the function accesses prior frames.
23097 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23099         PR debug/84456
23100         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23101         gen_llsym, otherwise call maybe_gen_llsym.
23103         PR inline-asm/84742
23104         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23105         has ',' character inside of it.
23107 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23109         PR target/84748
23110         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23111         as clobbering CC_REGNUM.
23113 2018-03-08  Richard Biener  <rguenther@suse.de>
23115         PR middle-end/84552
23116         * tree-scalar-evolution.c: Include tree-into-ssa.h.
23117         (follow_copies_to_constant): Do not follow SSA names registered
23118         for update.
23120 2018-03-08  Richard Biener  <rguenther@suse.de>
23122         PR tree-optimization/84178
23123         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23124         to caller.
23125         (version_loop_for_if_conversion): Delay update_ssa call.
23126         (tree_if_conversion): Delay update_ssa until after predicate
23127         insertion.
23129 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
23131         PR tree-optimization/84178
23132         * tree-if-conv.c (release_bb_predicate): Remove the
23133         the assertion that the stmts have NULL use_ops.
23134         Discard the statements, asserting that they haven't
23135         yet been added to a BB.
23137 2018-03-08  Richard Biener  <rguenther@suse.de>
23139         PR tree-optimization/84746
23140         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23141         (phi_translate): Pass in destination ANTIC_OUT set.
23142         (phi_translate_1): Likewise.  For a simplified result lookup
23143         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23144         (phi_translate_set): Adjust.
23145         (do_pre_regular_insertion): Likewise.
23146         (do_pre_partial_partial_insertion): Likewise.
23148 2018-03-08  Martin Liska  <mliska@suse.cz>
23150         PR gcov-profile/84735
23151         * doc/gcov.texi: Document usage of profile files.
23152         * gcov-io.h: Document changes in the format.
23154 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23156         PR debug/84404
23157         PR debug/84408
23158         * dwarf2out.c (struct dw_line_info_table): Update comments for
23159         view == -1.
23160         (FORCE_RESET_NEXT_VIEW): New.
23161         (FORCE_RESETTING_VIEW_P): New.
23162         (RESETTING_VIEW_P): Check for -1 too.
23163         (ZERO_VIEW_P): Likewise.
23164         (new_line_info_table): Force-reset next view.
23165         (dwarf2out_begin_function): Likewise.
23166         (dwarf2out_source_line): Simplify zero_view_p initialization.
23167         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23168         view directly.  Omit view when omitting .loc at line 0.
23170 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23172         PR tree-optimization/84740
23173         * tree-switch-conversion.c (process_switch): Call build_constructors
23174         only if info.phi_count is non-zero.
23176         PR tree-optimization/84739
23177         * tree-tailcall.c (find_tail_calls): Check call arguments against
23178         DECL_ARGUMENTS (current_function_decl) rather than
23179         DECL_ARGUMENTS (func) when checking for tail recursion.
23181 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23183         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23184         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23185         Volker Reichelt's entry and add entries for people that perform
23186         GCC fuzzy testing and report numerous bugs.
23188 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
23190         PR target/82411
23191         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23192         readonly data in sdata, if that is disabled.
23193         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23194         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23195         -mreadonly-in-sdata option.
23197 2018-03-07  Martin Sebor  <msebor@redhat.com>
23199         PR tree-optimization/84468
23200         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23201         basic block when looking for nul assignment.
23203 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
23205         PR target/84277
23206         * except.h (output_function_exception_table): Adjust prototype.
23207         * except.c (output_function_exception_table): Remove FNNAME parameter
23208         and add SECTION parameter.  Ouput one part of the table at a time.
23209         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23210         the first part of the exception table and emit unwind directives.
23211         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23212         (i386_pe_seh_cold_init): Likewise.
23213         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23214         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23215         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23216         (ix86_output_call_insn): Emit a nop in one more case for SEH.
23217         * config/i386/winnt.c: Include except.h.
23218         (struct seh_frame_state): Add reg_offset, after_prologue and
23219         in_cold_section fields.
23220         (i386_pe_seh_end_prologue): Set seh->after_prologue.
23221         (i386_pe_seh_cold_init): New function.
23222         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23223         to seh->in_cold_section.
23224         (seh_emit_push): Record the offset of the push.
23225         (seh_emit_save): Record the offet of the save.
23226         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23227         Test seh->after_prologue to disregard the epilogue.
23228         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23229         (i386_pe_end_cold_function): New function.
23231 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23233         PR fortran/84565
23234         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23235         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23237         PR c++/84704
23238         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23239         on tmp_var.
23240         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23241         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23243         PR middle-end/84723
23244         * multiple_target.c: Include tree-inline.h and intl.h.
23245         (expand_target_clones): Diagnose and fail if node->definition and
23246         !tree_versionable_function_p (node->decl).
23248 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
23250         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23251         sprint_ul.
23252         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23253         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23254         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23256 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23258         PR target/84710
23259         * combine.c (try_combine): Use reg_or_subregno instead of handling
23260         just paradoxical SUBREGs and REGs.
23262 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
23264         * config/arc/arc.c (arc_finalize_pic): Remove function.
23265         (arc_must_save_register): We use single base PIC register, remove
23266         checks to save/restore the PIC register.
23267         (arc_expand_prologue): Likewise.
23268         * config/arc/arc-protos.h (arc_set_default_type_attributes):
23269         Remove.
23270         (arc_verify_short): Likewise.
23271         (arc_attr_type): Likewise.
23272         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23273         (walk_stores): Likewise.
23274         (arc_address_cost): Make it static.
23275         (arc_verify_short): Likewise.
23276         (branch_dest): Likewise.
23277         (arc_attr_type): Likewise.
23278         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23279         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23280         (arc_final_prescan_insn): Remove inserting the nops due to
23281         hardware hazards.  It is done in reorg step.
23282         (insn_length_variant_t): Remove.
23283         (insn_length_parameters_t): Likewise.
23284         (arc_insn_length_parameters): Likewise.
23285         (arc_get_insn_variants): Likewise.
23286         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23288 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23290         PR inline-asm/84683
23291         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23292         assertion failure.
23294         PR tree-optimization/84687
23295         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23296         on new_node->decl.
23297         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23299 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23301         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23302         Rename to ppc_speculation_barrier.
23303         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23304         __builtin_ppc_speculation_barrier.
23306 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23308         PR target/84700
23309         * combine.c (combine_simplify_rtx): Don't try to simplify if
23310         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23311         are equal to x.
23313 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
23315         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
23316         to 32 bytes when compiling for POWER9.
23318 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23320         PR target/84564
23321         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
23322         regparm >= 3 with no arg reg available also for calls with
23323         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
23325         PR target/84524
23326         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
23327         orig,vex.
23328         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
23330 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
23332         PR target/84264
23333         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
23335 2018-03-05  Richard Biener  <rguenther@suse.de>
23337         PR tree-optimization/84486
23338         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
23339         When inserting a __builtin_assume_aligned call set the LHS
23340         SSA name alignment info accordingly.
23342 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
23344         PR tree-optimization/84114
23345         * config/aarch64/aarch64.c (aarch64_reassociation_width)
23346         Avoid reassociation of FLOAT_MODE addition.
23348 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
23350         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
23351         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
23352         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
23353         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
23354         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
23355         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
23356         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
23357         and -mwbnoinvd.
23358         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
23359         __builtin_ia32_wbinvd): New builtins.
23360         (SPECIAL_ARGS2): New.
23361         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
23362         (SPECIAL_ARGS2): New.
23363         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
23364         (ix86_valid_target_attribute_inner_p): Ditto.
23365         (ix86_init_mmx_sse_builtins): Add special_args2.
23366         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
23367         TARGET_WBNOINVD_P): New.
23368         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
23369         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
23370         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
23371         * config/i386/immintrin.h (_wbinvd): New intrinsic.
23372         * config/i386/pconfigintrin.h: New file.
23373         * config/i386/wbnoinvdintrin.h: Ditto.
23374         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
23375         wbnoinvdintrin.h.
23376         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
23378 2018-03-05  Richard Biener  <rguenther@suse.de>
23380         PR tree-optimization/84670
23381         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23382         member.
23383         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23384         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23385         (compute_antic_aux): Only assert the number of values in ANTIC_IN
23386         doesn't grow if all successors (recursively) were visited at least
23387         once.
23389 2018-03-05  Richard Biener  <rguenther@suse.de>
23391         PR tree-optimization/84650
23392         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
23393         if executed in the loop pipeline.
23395 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
23397         * doc/configfiles.texi (Configuration Files): Move info about
23398         conditionalizing $target-protos.h to...
23399         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
23400         differs from $target-protos.h.
23402 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
23403             Chung-Ju Wu  <jasonwucj@gmail.com>
23405         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
23406         * config/nds32/nds32-multiple.md (setmemsi): Define.
23407         * config/nds32/nds32-memory-manipulation.c
23408         (nds32_gen_dup_4_byte_to_word_value): New.
23409         (emit_setmem_word_loop): New.
23410         (emit_setmem_byte_loop): New.
23411         (nds32_expand_setmem_loop): New.
23412         (nds32_expand_setmem_loop_v3m): New.
23413         (nds32_expand_setmem_unroll): New.
23414         (nds32_expand_setmem): New.
23416 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23417             Chung-Ju Wu  <jasonwucj@gmail.com>
23419         * config/nds32/nds32-memory-manipulation.c
23420         (nds32_emit_load_store): New.
23421         (nds32_emit_post_inc_load_store): New.
23422         (nds32_emit_mem_move): New.
23423         (nds32_emit_mem_move_block): New.
23424         (nds32_expand_movmemsi_loop_unknown_size): New.
23425         (nds32_expand_movmemsi_loop_known_size): New.
23426         (nds32_expand_movmemsi_loop): New.
23427         (nds32_expand_movmemsi_unroll): New.
23428         (nds32_expand_movmemqi): Rename ...
23429         (nds32_expand_movmemsi): ... to this.
23430         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
23431         (movmemsi): ... to this.
23432         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
23433         (nds32_expand_movmemsi): ... to this.
23435 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23436             Monk Chiang  <sh.chiang04@gmail.com>
23437             Chung-Ju Wu  <jasonwucj@gmail.com>
23439         * config/nds32/nds32-protos.h
23440         (nds32_expand_load_multiple): New arguments.
23441         (nds32_expand_store_multiple): Ditto.
23442         (nds32_valid_multiple_load_store): Rename ...
23443         (nds32_valid_multiple_load_store_p): ... to this.
23444         * config/nds32/nds32-memory-manipulation.c
23445         (nds32_expand_load_multiple): Refine implementation.
23446         (nds32_expand_store_multiple): Ditto.
23447         * config/nds32/nds32-multiple.md
23448         (load_multiple): Update nds32_expand_load_multiple interface.
23449         (store_multiple): Update nds32_expand_store_multiple interface.
23450         * config/nds32/nds32-predicates.c
23451         (nds32_valid_multiple_load_store): Rename ...
23452         (nds32_valid_multiple_load_store_p): ... to this and refine
23453         implementation.
23454         * config/nds32/predicates.md
23455         (nds32_load_multiple_and_update_address_operation): New predicate.
23456         (nds32_store_multiple_and_update_address_operation): New predicate.
23458 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23459             Chung-Ju Wu  <jasonwucj@gmail.com>
23461         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
23462         (combo): New attribute.
23463         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
23465 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
23467         * config/nds32/nds32.opt: Change -mcmodel= default value.
23469 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
23470             Monk Chiang  <sh.chiang04@gmail.com>
23471             Chung-Ju Wu  <jasonwucj@gmail.com>
23473         * config/nds32/constants.md (unspec_element): New enum.
23474         * config/nds32/constraints.md (Umw): New constraint.
23475         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
23476         * config/nds32/nds32-intrinsic.md: Likewise.
23477         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
23478         (nds32_valid_smw_lwm_base_p): New.
23479         (nds32_output_smw_single_word): New.
23480         (nds32_output_lmw_single_word): New.
23481         (nds32_expand_unaligned_load): New.
23482         (nds32_expand_unaligned_store): New.
23483         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
23484         (nds32_output_smw_single_word): Declare.
23485         (nds32_output_lmw_single_word): Declare.
23486         (nds32_expand_unaligned_load): Declare.
23487         (nds32_expand_unaligned_store): Declare.
23488         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
23489         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
23490         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
23491         NDS32_BUILTIN_UASTORE_DW.
23492         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
23493         predicate.
23495 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
23496             Kito Cheng  <kito.cheng@gmail.com>
23497             Chung-Ju Wu  <jasonwucj@gmail.com>
23499         * config/nds32/nds32-intrinsic.c
23500         (nds32_expand_builtin_null_ftype_reg): Delete.
23501         (nds32_expand_builtin_reg_ftype_imm): Ditto.
23502         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
23503         (nds32_read_argument): New.
23504         (nds32_legitimize_target): Ditto.
23505         (nds32_legitimize_argument): Ditto.
23506         (nds32_check_constant_argument): Ditto.
23507         (nds32_expand_unop_builtin): Ditto.
23508         (nds32_expand_unopimm_builtin): Ditto.
23509         (nds32_expand_binop_builtin): Ditto.
23510         (nds32_builtin_decl_impl): Ditto.
23511         (builtin_description): Ditto.
23512         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
23513         (nds32_init_builtins_impl): Ditto.
23514         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
23515         (nds32_builtin_decl): New.
23516         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
23517         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
23519 2018-03-02  Jeff Law  <law@redhat.com>
23521         * reorg.c (stop_search_p): Handle DEBUG_INSN.
23522         (redundant_insn, fill_simple_delay_slots): Likewise.
23523         (fill_slots_from_thread): Likewise.
23524         * resource.c (mark_referenced_resources): Likewise.
23525         (mark_set_resources, find_dead_or_set_registers): Likewise.
23527 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23529         * substring-locations.h (format_warning_va): Formatting fix for
23530         ATTRIBUTE_GCC_DIAG.
23531         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
23532         argument.
23533         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
23534         * substring-locations.c: Include intl.h.
23535         (format_warning_va): Turned into small wrapper around
23536         format_warning_n_va, renamed to ...
23537         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
23538         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
23539         use ngettext.
23540         (format_warning_at_substring_n): New function.
23541         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
23542         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
23543         format_warning_at_substring with just a shorter name instead of
23544         const function pointer.
23545         (fmtwarn_n): New function.
23546         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
23547         appropriate, get rid of all the fmtstr temporaries, move conditionals
23548         with G_() wrapped string literals directly into fmtwarn arguments,
23549         cast dir.len to (int), formatting fixes.
23551 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
23553         * doc/invoke.texi: Remove "Cilk Plus" references.
23555 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23556             Richard Biener  <rguenther@suse.de>
23558         PR ipa/84628
23559         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
23560         for error or warning attributes if CALL_FROM_THUNK_P is set.
23561         Formatting fixes.
23563 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23565         PR target/56540
23566         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
23567         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
23569         PR target/56540
23570         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
23571         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
23573         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
23574         instead of -1U in last predictors element's probability member.
23576 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
23578         PR ipa/83983
23579         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23580         arguments if they are comparable.
23582 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23584         PR tree-optimization/84634
23585         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
23586         masks and masked_loop_p with a single loop_masks, making sure it's
23587         null for bb vectorization.
23589 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23591         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
23592         (vect_analyze_data_ref_access): Use loop->safe_len rather than
23593         loop->force_vectorize to check whether there is no alias.
23595 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23597         PR target/84614
23598         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
23599         prototypes.
23600         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
23601         comments.
23602         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
23603         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
23604         instead of a loop around prev_real_insn.
23605         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
23606         prev_real_insn.
23608         PR inline-asm/84625
23609         * config/i386/i386.c (ix86_print_operand): Use conditional
23610         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
23611         zero vector.
23613 2018-03-02  Richard Biener  <rguenther@suse.de>
23615         PR tree-optimization/84427
23616         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
23617         (bitmap_set_subtract_values): Rewrite to handle multiple
23618         exprs per value.
23619         (clean): Likewise.
23620         (prune_clobbered_mems): Likewise.
23621         (phi_translate): Take edge instead of pred/phiblock.
23622         (phi_translate_1): Likewise.
23623         (phi_translate_set): Likewise.  Insert all translated
23624         exprs for a value into the set, keeping possibly multiple
23625         expressions per value.
23626         (compute_antic_aux): Adjust for phi_translate changes.
23627         When intersecting union the expressions and prune those
23628         not in the final value set, keeping possibly multiple
23629         expressions per value.  Do not use value-insertion
23630         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
23631         all expressions.  Add verification that the value-sets
23632         only shrink during iteration.
23633         (compute_partial_antic_aux): Adjust for the phi_translate changes.
23634         (do_pre_regular_insertion): Likewise.
23635         (do_pre_partial_partial_insertion): Likewise.
23637 2018-03-02  Richard Biener  <rguenther@suse.de>
23639         PR target/82005
23640         * config/darwin.c (saved_debug_info_level): New static global.
23641         (darwin_asm_lto_start): Disable debug info generation for LTO out.
23642         (darwin_asm_lto_end): Restore debug info generation settings.
23644 2018-03-01  Martin Liska  <mliska@suse.cz>
23646         PR sanitizer/82484
23647         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
23648         volatile arguments.
23650 2018-03-01  Richard Biener  <rguenther@suse.de>
23652         PR debug/84645
23653         * dwarf2out.c (gen_variable_die): Properly handle late VLA
23654         type annotation with LTO when debug was disabled at compile-time.
23656 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
23658         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
23659         XINT with INTVAL.
23660         (mips_final_postscan_insn): Likewise.
23662 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
23664         PR rtl-optimization/84528
23665         * alias.c (init_alias_target): Add commentary.
23666         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
23667         a unique base value if the frame pointer is not eliminated
23668         to the stack pointer.
23670 2018-03-01  Tom de Vries  <tom@codesourcery.com>
23672         PR rtl-optimization/83327
23673         * lra-int.h (hard_regs_spilled_into): Declare.
23674         * lra.c (hard_regs_spilled_into): Define.
23675         (init_reg_info): Init hard_regs_spilled_into.
23676         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
23677         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
23678         (process_bb_lives): Handle hard_regs_spilled_into.
23679         (lra_create_live_ranges_1): Before doing liveness propagation, clear
23680         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
23682 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
23684         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
23685         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
23686         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
23687         * config/rs6000/aix72.h: New file.
23689 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
23691         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
23692         instead of warning_at with conditional singular and plural messages
23693         where possible.
23695         PR target/52991
23696         * stor-layout.c (update_alignment_for_field): For
23697         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
23698         && !DECL_PACKED (field), do the alignment update, just use
23699         only desired_align instead of MAX (type_align, desired_align)
23700         as the alignment.
23701         (place_field): Don't do known_align < desired_align handling
23702         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
23703         is non-NULL, instead do it after rli->prev_field handling and
23704         only if not within a bitfield word.  For DECL_PACKED (field)
23705         use type_align of BITS_PER_UNIT.
23707 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
23709         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
23710         superfluous parentheses and trailing spaces.
23712 2018-02-28  Richard Biener  <rguenther@suse.de>
23714         PR tree-optimization/84584
23715         * graphite-scop-detection.c (scop_detection::add_scop): Discard
23716         SCoPs with fake exit edge.
23718 2018-02-28  Martin Liska  <mliska@suse.cz>
23720         PR testsuite/84597
23721         * timevar.c (timer::print): Fix format to properly print 100%
23722         values.
23724 2018-02-28  Richard Biener  <rguenther@suse.de>
23726         PR middle-end/84607
23727         * genmatch.c (capture_info::walk_match): Do not mark
23728         captured expressions without operands as expr_p given
23729         they act more like predicates and should be subject to
23730         "lost tail" side-effect preserving.
23732 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
23734         PR rtl-optimization/81611
23735         * auto-inc-dec.c (attempt_change): Move dead note from
23736         mem_insn if it's the next use of regno
23737         (find_address): Take address use of reg holding
23738         non-incremented value.  Add parm to limit search to the named
23739         reg only.
23740         (merge_in_block): Attempt to use a mem insn that is the next
23741         use of the original regno.
23743 2018-02-27  Martin Sebor  <msebor@redhat.com>
23745         PR c++/83871
23746         * doc/invoke.texi (-Wmissing-attributes): New option.
23747         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
23749 2018-02-27  Martin Sebor  <msebor@redhat.com>
23751         PR translation/84207
23752         * diagnostic-core.h (warning_n, error_n, inform_n): Change
23753         n argument to unsigned HOST_WIDE_INT.
23754         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
23755         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
23756         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
23757         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
23759 2018-02-27  Richard Biener  <rguenther@suse.de>
23761         PR tree-optimization/84512
23762         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
23763         Do not use the estimate returned from record_stmt_cost for
23764         the scalar iteration cost but sum properly using add_stmt_cost.
23766 2018-02-27  Richard Biener  <rguenther@suse.de>
23768         PR tree-optimization/84466
23769         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23770         Adjust last change to less strictly validate use operands.
23772 2018-02-27  Martin Liska  <mliska@suse.cz>
23774         PR gcov-profile/84548
23775         * gcov.c (process_file): Allow partial overlap and consider it
23776         also as group functions.
23777         (output_lines): Properly calculate range of lines for a group.
23779 2018-02-27  Martin Liska  <mliska@suse.cz>
23781         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
23782         'ggc' suffixes.  Change first column width.
23783         (timer::print): Fix formatting of the column.
23785 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
23787         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
23788         preserve inline entry blocks for the sake of debug inline
23789         entry point markers alone.
23790         (remove_unused_locals): Suggest in comments a better place to
23791         force the preservation of inline entry blocks that are
23792         otherwise unused, but do not preserve them.
23794 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23796         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
23798 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23800         PR target/84039
23801         * config/i386/constraints.md (Bs): Replace
23802         ix86_indirect_branch_register with
23803         TARGET_INDIRECT_BRANCH_REGISTER.
23804         (Bw): Likewise.
23805         * config/i386/i386.md (indirect_jump): Likewise.
23806         (tablejump): Likewise.
23807         (*sibcall_memory): Likewise.
23808         (*sibcall_value_memory): Likewise.
23809         Peepholes of indirect call and jump via memory: Likewise.
23810         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
23811         (*sibcall_value_GOT_32): Likewise.
23812         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23813         (GOT_memory_operand): Likewise.
23814         (call_insn_operand): Likewise.
23815         (sibcall_insn_operand): Likewise.
23816         (GOT32_symbol_operand): Likewise.
23817         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
23819 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
23821         PR rtl-optimization/83496
23822         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
23823         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
23824         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
23825         redundant insn, if any.
23826         (relax_delay_slots): Likewise.
23827         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
23829 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
23831         PR tree-optimization/83965
23832         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
23833         that grouped statements are part of a reduction chain.  Return
23834         true if the statement is not marked as a reduction itself but
23835         is part of a group.
23836         (vect_recog_dot_prod_pattern): Don't check whether the statement
23837         is part of a group here.
23838         (vect_recog_sad_pattern): Likewise.
23839         (vect_recog_widen_sum_pattern): Likewise.
23841 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
23843         PR debug/84545
23844         * final.c (rest_of_clean_state): Also look for calls inside sequences.
23846 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23848         PR target/84530
23849         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
23850         the bool argument.
23851         (ix86_output_indirect_function_return): New prototype.
23852         (ix86_split_simple_return_pop_internal): Likewise.
23853         * config/i386/i386.c (indirect_return_via_cx): New.
23854         (indirect_return_via_cx_bnd): Likewise.
23855         (indirect_thunk_name): Handle return va CX_REG.
23856         (output_indirect_thunk_function): Create alias for
23857         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
23858         (ix86_output_indirect_jmp): Remove the bool argument.
23859         (ix86_output_indirect_function_return): New function.
23860         (ix86_split_simple_return_pop_internal): Likewise.
23861         * config/i386/i386.md (*indirect_jump): Don't pass false
23862         to ix86_output_indirect_jmp.
23863         (*tablejump_1): Likewise.
23864         (simple_return_pop_internal): Change it to define_insn_and_split.
23865         Call ix86_split_simple_return_pop_internal to split it for
23866         -mfunction-return=.
23867         (simple_return_indirect_internal): Call
23868         ix86_output_indirect_function_return instead of
23869         ix86_output_indirect_jmp.
23871 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
23873         PR bootstrap/84405
23874         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
23875         memset and value initialization afterwards.
23877 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
23879         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
23881 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23883         PR target/84521
23884         * common/config/aarch64/aarch64-common.c
23885         (aarch_option_optimization_table[]): Switch
23886         off fomit-frame-pointer
23888 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
23889             Chung-Ju Wu  <jasonwucj@gmail.com>
23891         * config/nds32/nds32-multiple.md (load_multiple): Disallow
23892         volatile memory.
23893         (store_multiple): Ditto.
23895 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
23897         * config.gcc: Add --with-cpu support for nds32 target.
23898         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
23899         * config/nds32/nds32.opt: Add -mcpu= option.
23901 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
23903         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
23904         isel=yes): Warn for these deprecated options.
23906 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
23908         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
23909         ISA_2_5_MASKS_EMBEDDED.
23911 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
23913         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
23914         p->max as pointers rather than using iterative_hash_expr.
23916 2018-02-23  Carl Love  <cel@us.ibm.com>
23918         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
23919         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
23920         BU_P8V_OVERLOAD_2.
23921         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
23922         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
23923         P8V_BUILTIN_VEC_VUNSIGNED2.
23925 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
23927         PR target/81572
23928         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
23929         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
23930         LRA_UNKNOWN_ALT.
23931         * lra-constraints.c (curr_insn_transform): Set up
23932         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
23933         LRA_UNKNOWN_ALT.
23934         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
23935         * lra-eliminations.c (spill_pseudos): Ditto.
23936         (process_insn_for_elimination): Ditto.
23937         * lra-lives.c (reg_early_clobber_p): Use the new macros.
23938         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
23939         LRA_NON_CLOBBERED_ALT.
23941 2018-02-22  Martin Sebor  <msebor@redhat.com>
23943         PR tree-optimization/84480
23944         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
23945         to maybe_diag_stxncpy_trunc.  Call it.
23946         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
23947         from gimple_fold_builtin_strcpy.  Print inlining stack.
23948         (handle_builtin_stxncpy): Print inlining stack.
23949         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
23951 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
23953         PR target/84176
23954         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
23955         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
23956         and -fcheck-pointer-bounds are used together.
23957         (indirect_thunk_prefix): New enum.
23958         (indirect_thunk_need_prefix): New function.
23959         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
23960         "_nt" instead of "_bnd" for NOTRACK prefix.
23961         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
23962         (output_indirect_thunk_function): Likewise.
23963         (): Likewise.
23964         (ix86_code_end): Update output_indirect_thunk_function calls.
23965         (ix86_output_indirect_branch_via_reg): Replace
23966         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
23967         (ix86_output_indirect_branch_via_push): Likewise.
23968         (ix86_output_function_return): Likewise.
23969         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
23970         incompatible with -fcf-protection=branch and
23971         -fcheck-pointer-bounds.
23973 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
23975         PR target/83335
23976         * config/aarch64/aarch64.c (aarch64_print_address_internal):
23977         Change gcc_assert call to output_operand_lossage.
23979 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
23981         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
23983 2018-02-22  DJ Delorie  <dj@redhat.com>
23984             Sebastian Perta  <sebastian.perta@renesas.com>
23985             Oleg Endo  <olegendo@gcc.gnu.org>
23987         * config/rx/rx.c (rx_rtx_costs): New function.
23988         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
23990 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
23992         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
23994 2018-02-22  Martin Liska  <mliska@suse.cz>
23996         PR driver/83193
23997         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
23998         Add "native" as a possible value.
24000 2018-02-22  Martin Liska  <mliska@suse.cz>
24002         PR driver/83193
24003         * config/i386/i386.c (ix86_option_override_internal):
24004         Add "native" as a possible value for -march and -mtune.
24006 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24008         PR target/84502
24009         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24010         to all type variants.
24012         PR tree-optimization/84503
24013         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24014         width as info->bitpos + info->bitsize - start.
24015         (merged_store_group::merge_overlapping): Simplify width computation.
24016         (check_no_overlap): New function.
24017         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24018         start + width and last_order of the group, fail if check_no_overlap
24019         fails.
24020         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24021         to group if check_no_overlap fails.
24023 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24025         * config/rs6000/altivec.md: Delete contraint arguments to
24026         define_expand, define_split, and define_peephole2, and in
24027         define_insn_and_split if always unused.
24028         * config/rs6000/darwin.md: Ditto.
24029         * config/rs6000/dfp.md: Ditto.
24030         * config/rs6000/rs6000.md: Ditto.
24031         * config/rs6000/sync.md: Ditto.
24032         * config/rs6000/vector.md: Ditto.
24033         * config/rs6000/vsx.md: Ditto.
24035 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24037         * config/rs6000/altivec.md: Write output control strings as braced
24038         blocks instead of double-quoted strings.
24039         * config/rs6000/darwin.md: Ditto.
24040         * config/rs6000/rs6000.md: Ditto.
24041         * config/rs6000/vector.md: Ditto.
24042         * config/rs6000/vsx.md: Ditto.
24044 2018-02-21  Jason Merrill  <jason@redhat.com>
24046         PR c++/84314 - ICE with templates and fastcall attribute.
24047         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24049 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24051         * ipa-cp.c (determine_versionability): Fix comment typos.
24053 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24055         PR c/84229
24056         * ipa-cp.c (determine_versionability): Do not version functions caling
24057         va_arg_pack.
24059 2018-02-21  Martin Liska  <mliska@suse.cz>
24061         PR driver/83193
24062         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24063         Add "native" as a possible value.
24064         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
24065         the macro when native cpu detection is available.
24067 2018-02-21  Martin Liska  <mliska@suse.cz>
24069         PR driver/83193
24070         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24071         Add "native" as a possible value.
24072         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24073         when native cpu detection is available.
24075 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
24076             Martin Sebor  <msebor@redhat.com>
24078         PR tree-optimization/84478
24079         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24080         false.
24081         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24082         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24083         support which is conservatively correct, for 2 only stay conservative
24084         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
24085         argument to the 2 argument get_range_strlen, adjust 6 arg
24086         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24087         false.
24088         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24089         (gimple_fold_builtin_strlen): Pass true as last argument to
24090         get_range_strlen.
24092 2018-02-20  Martin Sebor  <msebor@redhat.com>
24094         PR middle-end/84095
24095         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24096         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
24097         (builtin_memref::builtin_memref): Factor out parts into
24098         set_base_and_offset and call it.
24100 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24102         PR middle-end/84406
24103         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24104         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24105         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
24106         search at the associated MODE_INT.
24108 2018-02-20  Jeff Law  <law@redhat.com>
24110         PR middle-end/82123
24111         PR tree-optimization/81592
24112         PR middle-end/79257
24113         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24114         for range data rather than using global data.
24115         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24116         range data rather than using global data.
24117         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24118         pass it to children as needed.
24119         (struct directive::fmtresult): Similarly.
24120         (struct directive::set_width): Similarly.
24121         (struct directive::set_precision): Similarly.
24122         (format_integer, format_directive, parse_directive): Similarly.
24123         (format_none): Accept unnamed vr_values parameter.
24124         (format_percent, format_floating, format_character): Similarly.
24125         (format_string, format_plain): Similarly.
24126         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24127         the EVRP range analyzer for range data rather than using global data.
24128         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24129         gimple-ssa-evrp-analyze.h
24130         (class sprintf_dom_walker): Add after_dom_children member function.
24131         Add evrp_range_analyzer member.
24132         (sprintf_dom_walker::before_dom_children): Call into the EVRP
24133         range analyzer as needed.
24134         (sprintf_dom_walker::after_dom_children): New member function.
24135         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24136         if not optimizing.
24137         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24138         (evrp_range_analyzer::pop_to_marker): Likewise.
24140 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
24142         PR tree-optimization/84419
24143         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24144         with the required type if its current type is compatible but
24145         different.
24147 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
24149         PR middle-end/82004
24150         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24151         after vectorization.
24153 2018-02-20  Martin Liska  <mliska@suse.cz>
24155         PR driver/83193
24156         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24157         possible values if we don't have a hint.
24159 2018-02-20  Martin Liska  <mliska@suse.cz>
24161         PR c/84310
24162         PR target/79747
24163         * final.c (shorten_branches): Build align_tab array with one
24164         more element.
24165         * opts.c (finish_options): Add alignment option limit check.
24166         (MAX_CODE_ALIGN): Likewise.
24167         (MAX_CODE_ALIGN_VALUE): Likewise.
24168         * doc/invoke.texi: Document maximum allowed option value for
24169         all -falign-* options.
24171 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
24173         PR target/84146
24174         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24175         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24176         * var-tracking.c (emit_note_insn_var_location): Remove all references
24177         to NOTE_INSN_CALL_ARG_LOCATION.
24178         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24179         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24180         Use copy_rtx_if_shared.
24181         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24182         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24183         (dwarf2out_var_location): Remove handling of
24184         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24185         on call_insn.
24186         * final.c (final_scan_insn): Remove all references to
24187         NOTE_INSN_CALL_ARG_LOCATION.
24188         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
24189         before dumping final insns.
24190         * except.c (emit_note_eh_region_end): Remove all references to
24191         NOTE_INSN_CALL_ARG_LOCATION.
24192         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24193         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24194         * config/arc/arc.c (hwloop_optimize): Likewise.
24195         * config/arm/arm.c (create_fix_barrier): Likewise.
24196         * config/s390/s390.c (s390_chunkify_start): Likewise.
24197         * config/sh/sh.c (find_barrier): Likewise.
24198         * config/i386/i386.c (rest_of_insert_endbranch,
24199         ix86_seh_fixup_eh_fallthru): Likewise.
24200         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24201         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24202         * config/frv/frv.c (frv_function_prologue): Likewise.
24203         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
24204         reg note.
24205         (note_outside_basic_block_p): Remove all references to
24206         NOTE_INSN_CALL_ARG_LOCATION.
24207         * gengtype.c (adjust_field_rtx_def): Likewise.
24208         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24209         Likewise.
24210         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24211         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24213         PR c++/84444
24214         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24215         is ADDR_EXPR.
24217         PR tree-optimization/84452
24218         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24219         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24220         is NULL.
24222 2018-02-19  Martin Liska  <mliska@suse.cz>
24224         PR sanitizer/82183
24225         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24227 2018-02-19  Martin Liska  <mliska@suse.cz>
24228             Richard Sandiford  <richard.sandiford@linaro.org>
24230         PR tree-optimization/82491
24231         * gimple-fold.c (get_base_constructor): Make earlier bail out
24232         to prevent ubsan.
24234 2018-02-19  Carl Love  <cel@us.ibm.com>
24236         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24237         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24238         BU_P8V_OVERLOAD_1.
24239         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24240         P8V_BUILTIN_VEC_NEG.
24242 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
24244         * config/rl78/rl78.md (movdf): New define expand.
24246 2018-02-19  Martin Liska  <mliska@suse.cz>
24248         PR other/80589
24249         * doc/invoke.texi: Fix typo.
24250         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24252 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
24254         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24255         handle rs6000_single_float and rs6000_double_float specially for
24256         e500 family CPUs.
24258 2018-02-16  Jeff Law  <law@redhat.com>
24260         * config/rx/rx.c (add_pop_cfi_notes): New function.;
24261         (pop_regs): Use it.
24263 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24265         PR ipa/84425
24266         * ipa-inline.c (inline_small_functions): Fix a typo.
24268 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24270         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24272 2018-02-16  Carl Love  <cel@us.ibm.com>
24274         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24275         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24276         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24277         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24278         expansion to P8V_BUILTIN_VEC_FLOAT2.
24280 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
24282         PR rtl-optimization/70023
24283         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24284         src_regno into account.
24286 2018-02-16  Carl Love  <cel@us.ibm.com>
24288         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24289         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24290         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24291         * config/rs6000/rs6000.c: Remove case statements for
24292         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24293         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24294         and P9V_BUILTIN_VEC_VINSERT4B.
24295         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24296         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24297         * config/rs6000/vsx.md:
24298         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24299         vec_insert4b.
24301 2018-02-16  Carl Love  <cel@us.ibm.com>
24303         * config/rs6000/altivec.h: Add builtin names vec_extract4b
24304         vec_insert4b.
24305         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24306         definitions.
24307         * config/rs6000/rs6000-c.c: Add the definitions for
24308         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24309         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24310         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24311         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
24312         definition for insert4b and define insn *insert3b_internal.
24313         * doc/extend.texi: Add documentation for vec_extract4b.
24315 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24317         * doc/extend.texi (Backwards Compatibility): Mention friend
24318         injection.  Note for-scope is deprecated.
24319         * doc/invoke.texi (-ffriend-injection): Deprecate.
24321 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
24323         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
24324         that moved to I2, also allow destinations that are a paradoxical
24325         subreg (instead of a normal reg).
24327 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
24329         PR target/83831
24330         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
24331         to QImode.
24333 2018-02-16  Richard Biener  <rguenther@suse.de>
24335         PR tree-optimization/84037
24336         PR tree-optimization/84016
24337         PR target/82862
24338         * config/i386/i386.c (ix86_builtin_vectorization_cost):
24339         Adjust vec_construct for the fact we need additional higher latency
24340         128bit inserts for AVX256 and AVX512 vector builds.
24341         (ix86_add_stmt_cost): Scale vector construction cost for
24342         elementwise loads.
24344 2018-02-16  Richard Biener  <rguenther@suse.de>
24346         PR tree-optimization/84417
24347         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
24348         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
24349         (non_rewritable_lvalue_p): Likewise, use poly-ints.
24351 2018-02-16  Martin Liska  <mliska@suse.cz>
24353         PR sanitizer/84307
24354         * internal-fn.def (ASAN_CHECK): Set proper flags.
24355         (ASAN_MARK): Likewise.
24357 2018-02-16  Julia Koval  <julia.koval@intel.com>
24359         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
24360         from PTA_CANNONLAKE.
24362 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24364         PR target/84272
24365         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
24366         Use ++iter rather than iter++ for std::list iterators.
24367         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
24368         defer deleting them until all nodes in the forest are processed.  Do
24369         free even leaf nodes.  Change to_process into auto_vec.
24371         PR bootstrap/84405
24372         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
24373         * vec.h (vec_default_construct): Use memset instead of placement new
24374         if BROKEN_VALUE_INITIALIZATION is defined.
24375         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
24376         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
24377         is defined.
24379         PR rtl-optimization/83723
24380         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
24381         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
24382         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
24383         recursive calls.
24384         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
24385         callers.
24386         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
24388 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
24390         PR rtl-optimization/81443
24391         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
24392         from inner REGs to paradoxical SUBREGs.
24394 2018-02-16  Richard Biener  <rguenther@suse.de>
24396         PR tree-optimization/84399
24397         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24398         For operands we can analyze at their definition make sure we can
24399         analyze them at each use as well.
24401 2018-02-16  Richard Biener  <rguenther@suse.de>
24403         PR tree-optimization/84190
24404         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
24405         volatile accesses if the decl isn't volatile.
24407 2018-02-15  Jason Merrill  <jason@redhat.com>
24409         PR c++/84314 - ICE with templates and fastcall attribute.
24410         * attribs.c (build_type_attribute_qual_variant): Don't clobber
24411         TYPE_CANONICAL on an existing type.
24413 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
24415         PR tree-optimization/84383
24416         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
24417         dstoff nor call operand_equal_p if dstbase is NULL.
24419         PR tree-optimization/84334
24420         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
24421         also a CONSTANT_CLASS_P, punt.
24423 2018-02-14  Jim Wilson  <jimw@sifive.com>
24425         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
24426         first SMALL_OPERAND check.  New local min_second_step.  Move assert
24427         to where locals are set.  Add TARGET_RVC support.
24428         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
24430 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
24432         * doc/invoke.texi: Correct -Wformat-overflow code sample.
24434 2018-02-14  Martin Sebor  <msebor@redhat.com>
24436         PR tree-optimization/83698
24437         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
24438         arrays constrain the offset range to their bounds.
24439         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
24440         (builtin_access::overlap): Avoid setting the size of overlap if it's
24441         already been set.
24442         (maybe_diag_overlap): Also consider arrays when deciding what values
24443         of offsets to include in diagnostics.
24445 2018-02-14  Martin Sebor  <msebor@redhat.com>
24447         PR c/84108
24448         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
24449         that correspond to the kind of a declaration.
24451 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
24453         PR target/83984
24454         * config/pa/pa.md: Load address of PIC label using the linkage table
24455         if the label is nonlocal.
24457 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24459         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
24460         warning message if user requests -maltivec=be.
24461         * doc/invoke.texi: Document deprecation of -maltivec=be.
24463 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
24465         PR target/84220
24466         * config/rs6000/rs6000-c.c: Update definitions for
24467         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
24468         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
24470 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24472         PR target/84239
24473         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
24474         add _get_ssp intrinsics. Remove argument from
24475         __builtin_ia32_rdssp[d|q].
24476         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
24477         * config/i386/i386-builtin.def: Remove argument from
24478         __builtin_ia32_rdssp[d|q].
24479         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
24480         ix86_expand_special_args_builtin for _rdssp[d|q].
24481         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
24482         Clear register before usage.
24483         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
24484         Add documentation for new _get_ssp and _inc_ssp intrinsics.
24486 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
24488         PR tree-optimization/84357
24489         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
24490         operand 1 of an ARRAY_REF too.
24492 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
24494         PR target/83831
24495         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
24496         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24497         declarations.
24498         (set_of_reg): New struct.
24499         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
24500         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
24501         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24502         functions.
24503         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
24504         Split into bitclr, bitset, bitinvert patterns if appropriate.
24505         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
24506         use rx_fuse_in_memory_bitop.
24507         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
24508         to named insn, correct maximum insn length.
24510 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
24512         PR target/79242
24513         * machmode.def: Define a complex mode for PARTIAL_INT.
24514         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
24515         MODE_PARTIAL_INT.
24516         * doc/rtl.texi: Document CSPImode.
24517         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
24518         handling.
24519         (msp430_hard_regno_nregs_with_padding): Likewise.
24521 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
24523         PR target/84279
24524         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
24526 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
24528         PR rtl-optimization/84169
24529         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
24530         we generated a parallel as new i3 and we split that to new i2 and i3
24531         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
24532         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
24533         those to i2, not i1.  Partially rewrite this scan code.
24535 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24537         PR c/82210
24538         * stor-layout.c (place_field): For variable length fields, adjust
24539         offset_align afterwards not just based on the field's alignment,
24540         but also on the size.
24542         PR middle-end/84309
24543         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
24544         of exps and logs in the use_exp2 case.
24546 2018-02-13  Jeff Law  <law@redhat.com>
24548         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
24549         entry for "vector".
24551         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
24552         ARGS as unused.
24554 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
24556         PR debug/84342
24557         PR debug/84319
24558         * common.opt (gas-loc-support, gas-locview-support): New.
24559         (ginline-points, ginternal-reset-location-views): New.
24560         * doc/invoke.texi: Document them.  Use @itemx where intended.
24561         (gvariable-location-views): Adjust.
24562         * target.def (reset_location_view): New.
24563         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
24564         (TARGET_RESET_LOCATION_VIEW): New.
24565         * doc/tm.texi: Rebuilt.
24566         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
24567         (dwarf2out_default_as_locview_support): New.
24568         (output_asm_line_debug_info): Use option variables.
24569         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
24570         (output_loc_list): Likewise.
24571         (add_high_low_attributes): Check option variables.
24572         Don't output entry view attribute in strict mode.
24573         (gen_inlined_subroutine_die): Check option variables.
24574         (dwarf2out_inline_entry): Likewise.
24575         (init_sections_and_labels): Likewise.
24576         (dwarf2out_early_finish): Likewise.
24577         (maybe_reset_location_view): New, from...
24578         (dwarf2out_var_location): ... here.  Call it.
24579         * debug.h (dwarf2out_default_as_loc_support): Declare.
24580         (dwarf2out_default_as_locview_support): Declare.
24581         * hooks.c (hook_int_rtx_insn_0): New.
24582         * hooks.h (hook_int_rtx_insn_0): Declare.
24583         * toplev.c (process_options): Take -gas-loc-support and
24584         -gas-locview-support from dwarf2out.  Enable
24585         -gvariable-location-views by default only with locview
24586         assembler support.  Enable -ginternal-reset-location-views by
24587         default only if the target defines the corresponding hook.
24588         Enable -ginline-points by default if location views are
24589         enabled; force it disabled if statement frontiers are
24590         disabled.
24591         * tree-inline.c (expand_call_inline): Check option variables.
24592         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
24594 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
24596         PR tree-optimization/84321
24597         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
24598         handling.  Also check whether the anti-range contains any values
24599         that satisfy the mask; switch to a VR_RANGE if not.
24601 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
24603         PR sanitizer/84340
24604         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
24606 2018-02-13  Martin Jambor  <mjambor@suse.cz>
24608         PR c++/83990
24609         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
24610         of call statements, also set location of a load to a temporary.
24612 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
24614         * config/rl78/rl78.c (add_vector_labels): New function.
24615         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
24616         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
24617         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
24618         which checks that no arguments are passed.
24619         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
24620         * doc/extend.texi: Documentation for the new attribute.
24622 2018-02-13  Andreas Schwab  <schwab@suse.de>
24624         * config/riscv/linux.h (CPP_SPEC): Define.
24626 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24628         PR target/84335
24629         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
24630         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
24631         OPTION_MASK_ISA_AES as first argument to def_builtin_const
24632         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
24633         instead of OPTION_MASK_ISA_PCLMUL as first argument to
24634         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
24635         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
24636         temporarily for AES and PCLMUL builtins.
24638         PR tree-optimization/84339
24639         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
24640         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
24641         Formatting fixes.
24643         PR middle-end/84309
24644         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
24645         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
24646         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
24647         inline function.
24648         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
24649         inline function.
24650         * omp-simd-clone.h: New file.
24651         * omp-simd-clone.c: Include omp-simd-clone.h.
24652         (expand_simd_clones): No longer static.
24653         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
24654         cgraph.h and omp-simd-clone.h.
24655         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
24656         (vect_recog_widen_shift_pattern): Formatting fix.
24657         (vect_pattern_recog_1): Don't check optab for calls.
24659         PR target/84336
24660         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
24661         operands[2] into a REG before using gen_lowpart on it.
24663 2018-02-12  Jeff Law  <law@redhat.com>
24665         PR target/83760
24666         * config/sh/sh.c (find_barrier): Consider a sibling call
24667         a barrier as well.
24669         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
24670         successfully back substituting a reg.
24672 2018-02-12  Richard Biener  <rguenther@suse.de>
24674         PR tree-optimization/84037
24675         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
24676         parameter, move visited init to caller.
24677         (vect_slp_analyze_operations): Separate cost from validity
24678         check, initialize visited once for all instances.
24679         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
24680         for all instances.
24681         * tree-vect-stmts.c (vect_model_simple_cost): Make early
24682         out an assert.
24683         (vect_model_promotion_demotion_cost): Likewise.
24684         (vectorizable_bswap): Guard cost modeling with !slp_node
24685         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
24686         SLP stmts.
24687         (vectorizable_call): Likewise.
24688         (vectorizable_conversion): Likewise.
24689         (vectorizable_assignment): Likewise.
24690         (vectorizable_shift): Likewise.
24691         (vectorizable_operation): Likewise.
24692         (vectorizable_store): Likewise.
24693         (vectorizable_load): Likewise.
24694         (vectorizable_condition): Likewise.
24695         (vectorizable_comparison): Likewise.
24697 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
24699         PR sanitizer/84307
24700         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
24701         (ASAN_MARK): Fix fnspec to account for return value, change pointer
24702         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
24704 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24706         PR middle-end/83665
24707         * params.def (inline-min-speedup): Increase from 8 to 15.
24708         (max-inline-insns-auto): Decrease from 40 to 30.
24709         * ipa-split.c (consider_split): Add some buffer for function to
24710         be considered inlining candidate.
24711         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
24712         default values.
24714 2018-02-12  Richard Biener  <rguenther@suse.de>
24716         PR tree-optimization/84037
24717         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
24718         matched stmts if we cannot swap the non-matched ones.
24720 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24722         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
24723         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
24724         _mm_maskz_scalef_round_ss): New intrinsics.
24725         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
24726         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
24727         __builtin_ia32_scalefss_round): Remove.
24728         (__builtin_ia32_scalefsd_mask_round,
24729         __builtin_ia32_scalefss_mask_round): New intrinsics.
24730         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
24731         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
24732         ((match_operand:VF_128 2 "<round_nimm_predicate>"
24733         "<round_constraint>")): Changed to ...
24734         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
24735         "<round_scalar_constraint>")): ... this.
24736         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
24737         %0, %1, %2<round_op3>}"): Changed to ...
24738         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
24739         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
24740         %2<round_scalar_mask_op3>}"): ... this.
24741         * config/i386/subst.md (round_scalar_nimm_predicate): New.
24743 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24745         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
24746         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
24747         (_mm_maskz_sqrt_round_ss): New intrinsics.
24748         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
24749         (__builtin_ia32_sqrtsd_mask_round)
24750         (__builtin_ia32_sqrtss_mask_round): New builtins.
24751         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
24752         (__builtin_ia32_sqrtss_round): Remove.
24753         (__builtin_ia32_sqrtsd_mask_round)
24754         (__builtin_ia32_sqrtss_mask_round): New builtins.
24755         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
24756         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
24757         ((match_operand:VF_128 1 "vector_operand"
24758         "xBm,<round_constraint>")): Changed to ...
24759         ((match_operand:VF_128 1 "vector_operand"
24760         "xBm,<round_scalar_constraint>")): ... this.
24761         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
24762         %0, %2, %<iptr>1<round_op3>}): Changed to ...
24763         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
24764         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
24765         %<iptr>1<round_scalar_mask_op3>}): ... this.
24766         ((set_attr "prefix" "<round_prefix>")): Changed to ...
24767         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
24769 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
24771         PR target/84266
24772         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
24773         Cast vec_cmpeq result to correct type.
24774         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
24775         Cast vec_cmpgt result to correct type.
24777 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
24779         * final.c (final_scan_insn_1): Renamed from...
24780         (final_scan_insn): ... this.  New wrapper, to recover
24781         seen from the outermost call in recursive ones.
24782         * config/sparc/sparc.c (output_return): Drop seen from call.
24783         (output_sibcall): Likewise.
24784         * config/visium/visium.c (output_branch): Likewise.
24786 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
24788         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
24789         function label.
24791 2018-02-10  Alan Modra  <amodra@gmail.com>
24793         PR target/84300
24794         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
24795         Specify LR as an input.
24797 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
24799         PR sanitizer/83987
24800         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
24801         remove_member_access_dummy_vars): New functions.
24802         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
24803         lower_omp_1, execute_lower_omp): Use them.
24805         PR rtl-optimization/84308
24806         * shrink-wrap.c (spread_components): Release todo vector.
24808 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
24810         PR rtl-optimization/57193
24811         * ira-color.c (struct allocno_color_data): Add member
24812         conflict_allocno_hard_prefs.
24813         (update_conflict_allocno_hard_prefs): New.
24814         (bucket_allocno_compare_func): Add a preference based on
24815         conflict_allocno_hard_prefs.
24816         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
24817         (color_allocnos): Remove a dead code.  Initiate
24818         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
24820 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
24822         PR target/84226
24823         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
24824         constraint from =wa to wa.  Avoid a subreg on the output operand,
24825         instead use a pseudo and subreg it in a move.
24826         (p9_xxbrd_<mode>): Changed to ...
24827         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
24828         (p9_xxbrd_v2df): New expander.
24829         (p9_xxbrw_<mode>): Changed to ...
24830         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
24831         (p9_xxbrw_v4sf): New expander.
24833 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
24835         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
24837 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
24839         PR target/83926
24840         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
24841         multiply in 32-bit mode.
24842         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
24843         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
24844         mode.
24846 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
24848         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
24849         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
24850         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
24851         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
24853 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
24855         PR lto/84213
24856         * dwarf2out.c (is_trivial_indirect_ref): New function.
24857         (dwarf2out_late_global_decl): Do not generate a location
24858         attribute for variables that have a non-trivial DECL_VALUE_EXPR
24859         and that are not defined in the current unit.
24861 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
24863         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
24864         instead of a libcall for UNORDERED.
24866 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
24868         PR target/82641
24869         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
24870         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
24872 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24874         PR target/PR84295
24875         * config/s390/s390.c (s390_set_current_function): Invoke
24876         s390_indirect_branch_settings also if fndecl didn't change.
24878 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
24880         * config/rs6000/rs6000.md (blockage): Set length to zero.
24882 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
24884         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
24886 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
24888         PR sanitizer/84285
24889         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
24890         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
24891         -static-lib*san.
24893         PR debug/84252
24894         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
24895         PARALLEL incoming that failed vt_get_decl_and_offset check.
24897         PR middle-end/84237
24898         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
24899         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
24900         TREE_READONLY bit.
24901         (get_variable_section): For decls in named .bss* sections pass true as
24902         second argument to bss_initializer_p.
24904 2018-02-09  Marek Polacek  <polacek@redhat.com>
24905             Jakub Jelinek  <jakub@redhat.com>
24907         PR c++/83659
24908         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
24909         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
24910         Sync some changes from cxx_fold_indirect_ref.
24912 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
24914         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
24915         markers.
24916         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
24917         (BLOCK_INLINE_ENTRY_LABEL): New.
24918         (dwarf2out_var_location): Disregard inline entry markers.
24919         (inline_entry_data): New struct.
24920         (inline_entry_data_hasher): New hashtable type.
24921         (inline_entry_data_hasher::hash): New.
24922         (inline_entry_data_hasher::equal): New.
24923         (inline_entry_data_table): New variable.
24924         (add_high_low_attributes): Add DW_AT_entry_pc and
24925         DW_AT_GNU_entry_view attributes if a pending entry is found
24926         in inline_entry_data_table.  Add old entry_pc attribute only
24927         if debug nonbinding markers are disabled.
24928         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
24929         markers are enabled.
24930         (block_within_block_p, dwarf2out_inline_entry): New.
24931         (dwarf2out_finish): Check that no entries remained in
24932         inline_entry_data_table.
24933         * final.c (reemit_insn_block_notes): Handle inline entry notes.
24934         (final_scan_insn, notice_source_line): Likewise.
24935         (rest_of_clean_state): Skip inline entry markers.
24936         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
24937         markers.
24938         * gimple.c (gimple_build_debug_inline_entry): New.
24939         * gimple.h (enum gimple_debug_subcode): Add
24940         GIMPLE_DEBUG_INLINE_ENTRY.
24941         (gimple_build_debug_inline_entry): Declare.
24942         (gimple_debug_inline_entry_p): New.
24943         (gimple_debug_nonbind_marker_p): Adjust.
24944         * insn-notes.def (INLINE_ENTRY): New.
24945         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
24946         inline entry marker notes.
24947         (print_insn): Likewise.
24948         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
24949         (INSN_DEBUG_MARKER_KIND): Likewise.
24950         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
24951         * tree-inline.c (expand_call_inline): Build and insert
24952         debug_inline_entry stmt.
24953         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
24954         inline entry blocks early, if nonbind markers are enabled.
24955         (dump_scope_block): Dump fragment info.
24956         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
24957         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
24958         (gimple_build_debug_inline_entry): New.
24959         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
24960         Enable/disable inline entry points too.
24961         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
24962         (DEBUG_INSN): Describe inline entry markers.
24964         * common.opt (gvariable-location-views): New.
24965         (gvariable-location-views=incompat5): New.
24966         * config.in: Rebuilt.
24967         * configure: Rebuilt.
24968         * configure.ac: Test assembler for view support.
24969         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
24970         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
24971         * dwarf2out.c (var_loc_view): New typedef.
24972         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
24973         (dwarf2out_locviews_in_attribute): New.
24974         (dwarf2out_locviews_in_loclist): New.
24975         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
24976         (enum dw_line_info_opcode): Add LI_adv_address.
24977         (struct dw_line_info_table): Add view.
24978         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
24979         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
24980         (zero_view_p): New variable.
24981         (ZERO_VIEW_P): New macro.
24982         (output_asm_line_debug_info): New.
24983         (struct var_loc_node): Add view.
24984         (add_AT_view_list, AT_loc_list): New.
24985         (add_var_loc_to_decl): Add view param.  Test it against last.
24986         (new_loc_list): Add view params.  Record them.
24987         (AT_loc_list_ptr): Handle loc and view lists.
24988         (view_list_to_loc_list_val_node): New.
24989         (print_dw_val): Handle dw_val_class_view_list.
24990         (size_of_die): Likewise.
24991         (value_format): Likewise.
24992         (loc_list_has_views): New.
24993         (gen_llsym): Set vl_symbol too.
24994         (maybe_gen_llsym, skip_loc_list_entry): New.
24995         (dwarf2out_maybe_output_loclist_view_pair): New.
24996         (output_loc_list): Output view list or entries too.
24997         (output_view_list_offset): New.
24998         (output_die): Handle dw_val_class_view_list.
24999         (output_dwarf_version): New.
25000         (output_compilation_unit_header): Use it.
25001         (output_skeleton_debug_sections): Likewise.
25002         (output_rnglists, output_line_info): Likewise.
25003         (output_pubnames, output_aranges): Update version comments.
25004         (output_one_line_info_table): Output view numbers in asm comments.
25005         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25006         Call maybe_gen_llsym.
25007         (loc_list_from_tree_1): Adjust.
25008         (add_AT_location_description): Create view list attribute if
25009         needed, check it's absent otherwise.
25010         (convert_cfa_to_fb_loc_list): Adjust.
25011         (maybe_emit_file): Call output_asm_line_debug_info for test.
25012         (dwarf2out_var_location): Reset views as needed.  Precompute
25013         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25014         attribute.  Set view.
25015         (new_line_info_table): Reset next view.
25016         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25017         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25018         the assembler, or select appropriate line info opcodes.
25019         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25020         (optimize_string_length): Catch it.  Adjust.
25021         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25022         dw_val_class_view_list, and remove it if no longer needed.
25023         (hash_loc_list): Hash view numbers.
25024         (loc_list_hasher::equal): Compare them.
25025         (optimize_location_lists): Check whether a view list symbol is
25026         needed, and whether the locview attribute is present, and
25027         whether they match.  Remove the locview attribute if no longer
25028         needed.
25029         (index_location_lists): Call skip_loc_list_entry for test.
25030         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25031         Use output_dwarf_version.
25032         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25033         (struct dw_val_node): Add val_view_list.
25034         * final.c (SEEN_NEXT_VIEW): New.
25035         (set_next_view_needed): New.
25036         (clear_next_view_needed): New.
25037         (maybe_output_next_view): New.
25038         (final_start_function): Rename to...
25039         (final_start_function_1): ... this.  Take pointer to FIRST,
25040         add SEEN parameter.  Emit param bindings in the initial view.
25041         (final_start_function): Reintroduce SEEN-less interface.
25042         (final): Rename to...
25043         (final_1): ... this.  Take SEEN parameter.  Output final pending
25044         next view at the end.
25045         (final): Reintroduce seen-less interface.
25046         (final_scan_insn): Output pending next view before switching
25047         sections or ending a block.  Mark the next view as needed when
25048         outputting variable locations.  Notify debug backend of section
25049         changes, and of location view changes.
25050         (rest_of_handle_final): Adjust.
25051         * toplev.c (process_options): Autodetect value for debug variable
25052         location views option.  Warn on incompat5 without -gdwarf-5.
25053         * doc/invoke.texi (gvariable-location-views): New.
25054         (gvariable-location-views=incompat5): New.
25055         (gno-variable-location-views): New.
25057 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
25059         PR tree-optimization/84136
25060         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25061         that the result of find_edge is non-NULL.
25063 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25065         PR target/83008
25066         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25067         storing integer register in SImode.  Fix cost of 256 and 512
25068         byte aligned SSE register store.
25070 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
25072         * config/i386/i386.c (ix86_multiplication_cost): Fix
25073         multiplication cost for TARGET_AVX512DQ.
25075 2018-02-08  Marek Polacek  <polacek@redhat.com>
25077         PR tree-optimization/84238
25078         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25079         get_range_strlen.
25081 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25083         PR tree-optimization/84265
25084         * tree-vect-stmts.c (vectorizable_store): Don't treat
25085         VMAT_CONTIGUOUS accesses as grouped.
25086         (vectorizable_load): Likewise.
25088 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25090         PR tree-optimization/81635
25091         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25092         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25093         (test_round_for_mask): New functions.
25094         (wide_int_cc_tests): Call test_round_for_mask.
25095         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25096         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25097         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25098         range returned by get_range_info.
25100 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25102         PR ipa/81360
25103         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25104         * symtab.c: Include builtins.h
25105         (symtab_node::output_to_lto_symbol_table_p): Move here
25106         from lto-streamer-out.c:output_symbol_p.
25107         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25108         (output_symbol_p): Move all logic to symtab.c
25109         (produce_symtab): Update.
25111 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25113         * config/s390/s390-opts.h (enum indirect_branch): Define.
25114         * config/s390/s390-protos.h (s390_return_addr_from_memory)
25115         (s390_indirect_branch_via_thunk)
25116         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25117         (enum s390_indirect_branch_type): Define.
25118         * config/s390/s390.c (struct s390_frame_layout, struct
25119         machine_function): Remove.
25120         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25121         (indirect_branch_table_label_no, indirect_branch_table_name):
25122         Define variables.
25123         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25124         (enum s390_indirect_branch_option): Define.
25125         (s390_return_addr_from_memory): New function.
25126         (s390_handle_string_attribute): New function.
25127         (s390_attribute_table): Add new attribute handler.
25128         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25129         (s390_indirect_branch_via_thunk): New function.
25130         (s390_indirect_branch_via_inline_thunk): New function.
25131         (s390_function_ok_for_sibcall): When jumping via thunk disallow
25132         sibling call optimization for non z10 compiles.
25133         (s390_emit_call): Force indirect branch target to be a single
25134         register.  Add r1 clobber for non-z10 compiles.
25135         (s390_emit_epilogue): Emit return jump via return_use expander.
25136         (s390_reorg): Handle JUMP_INSNs as execute targets.
25137         (s390_option_override_internal): Perform validity checks for the
25138         new command line options.
25139         (s390_indirect_branch_attrvalue): New function.
25140         (s390_indirect_branch_settings): New function.
25141         (s390_set_current_function): Invoke s390_indirect_branch_settings.
25142         (s390_output_indirect_thunk_function):  New function.
25143         (s390_code_end): Implement target hook.
25144         (s390_case_values_threshold): Implement target hook.
25145         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25146         macros.
25147         * config/s390/s390.h (struct s390_frame_layout)
25148         (struct machine_function): Move here from s390.c.
25149         (TARGET_INDIRECT_BRANCH_NOBP_RET)
25150         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25151         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25152         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25153         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25154         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25155         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25156         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25157         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25158         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25159         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25160         (mnemonic attribute): Add values which aren't recognized
25161         automatically.
25162         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25163         pattern for branch conversion.  Fix mnemonic attribute.
25164         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25165         indirect branch via thunk if requested.
25166         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25167         ("*indirect_jump"): Disable for branch conversion using out of
25168         line thunks.
25169         ("indirect_jump_via_thunk<mode>_z10")
25170         ("indirect_jump_via_thunk<mode>")
25171         ("indirect_jump_via_inlinethunk<mode>_z10")
25172         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25173         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25174         ("casesi_jump_via_inlinethunk<mode>_z10")
25175         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25176         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25177         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25178         ("*indirect2_jump"): Disable for branch conversion.
25179         ("casesi_jump"): Turn into expander and expand patterns for branch
25180         conversion.
25181         ("return_use"): New expander.
25182         ("*return"): Emit return via thunk and rename it to ...
25183         ("*return<mode>"): ... this one.
25184         * config/s390/s390.opt: Add new options and and enum for the
25185         option values.
25187 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25189         * lra-constraints.c (match_reload): Unconditionally use
25190         gen_lowpart_SUBREG, rather than selecting between that
25191         and equivalent gen_rtx_SUBREG code.
25193 2018-02-08  Richard Biener  <rguenther@suse.de>
25195         PR tree-optimization/84233
25196         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25197         changed flag instead of boguously re-using phi_inserted.
25199 2018-02-08  Martin Jambor  <mjambor@suse.cz>
25201         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25202         static local variables.
25204 2018-02-08  Richard Biener  <rguenther@suse.de>
25206         PR tree-optimization/84278
25207         * tree-vect-stmts.c (vectorizable_store): When looking for
25208         smaller vector types to perform grouped strided loads/stores
25209         make sure the mode is supported by the target.
25210         (vectorizable_load): Likewise.
25212 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25214         * config/aarch64/aarch64.c (aarch64_components_for_bb):
25215         Increase LDP/STP opportunities by adding adjacent callee-saves.
25217 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25219         PR rtl-optimization/84068
25220         PR rtl-optimization/83459
25221         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25223 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
25225         PR tree-optimization/84224
25226         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25227         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25228         non-zero arguments.
25230 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
25232         PR target/84113
25233         * config/rs6000/altivec.md (*restore_world): Remove LR use.
25234         * config/rs6000/predicates.md (restore_world_operation): Adjust op
25235         count, remove one USE.
25237 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25239         * doc/install.texi (Configuration): Document the
25240         --with-long-double-format={ibm,ieee} PowerPC configuration
25241         options.
25243         PR target/84154
25244         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25245         Convert from define_expand to be define_insn_and_split.  Rework
25246         float/double/_Float128 conversions to QI/HI/SImode to work with
25247         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
25248         conversions to QI/HImode types did a store and then a load to
25249         truncate the value.  For conversions to VSX registers, don't split
25250         the insn, instead emit the code directly.  Use the code iterator
25251         any_fix to combine signed and unsigned conversions.
25252         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25253         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25254         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25255         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25256         (fix_<mode>di2_hw): Likewise.
25257         (fixuns_<mode>di2_hw): Likewise.
25258         (fix_<mode>si2_hw): Likewise.
25259         (fixuns_<mode>si2_hw): Likewise.
25260         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25261         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25262         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25263         fix<uns>_trunc<SFDF:mode>si2_p8.
25264         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25265         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25266         (fix<uns>_<mode>_mem): Likewise.
25267         (fctiw<u>z_<mode>_mem): Likewise.
25268         (fix<uns>_<mode>_mem): Likewise.
25269         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25270         the register allocator from doing a direct move to the GPRs to do
25271         a store, and instead use the ISA 3.0 store byte/half-word from
25272         vector register instruction.  For IEEE 128-bit floating point,
25273         also optimize stores of 32-bit ints.
25274         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25276 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
25278         * genextract.c (push_pathstr_operand): New function to support
25279         [a-zA-Z].
25280         (walk_rtx): Call push_pathstr_operand.
25281         (print_path): Support [a-zA-Z].
25283 2018-02-07  Richard Biener  <rguenther@suse.de>
25285         PR tree-optimization/84037
25286         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25287         (cse_and_gimplify_to_preheader): Declare.
25288         (vect_get_place_in_interleaving_chain): Likewise.
25289         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25290         ivexpr_map.
25291         (_loop_vec_info::~_loop_vec_info): Delete it.
25292         (cse_and_gimplify_to_preheader): New function.
25293         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25294         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25295         (vectorizable_load): Likewise.  For grouped stores always base
25296         the IV on the first element.
25297         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25298         condition before gimplifying.
25300 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25302         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25303         *DIV_EXPR and *MOD_EXPR.
25305 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
25307         PR target/84248
25308         * config/i386/i386.c (ix86_option_override_internal): Mask out
25309         the CF_SET bit when checking -fcf-protection.
25311 2018-02-07  Tom de Vries  <tom@codesourcery.com>
25313         PR libgomp/84217
25314         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
25315         enough.
25317 2018-02-07  Richard Biener  <rguenther@suse.de>
25319         PR tree-optimization/84204
25320         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
25321         this place.
25323         PR tree-optimization/84205
25324         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
25325         special-case isl_ast_op_zdiv_r.
25327         PR tree-optimization/84223
25328         * graphite-scop-detection.c (gather_bbs::before_dom_children):
25329         Only add conditions from within the region.
25330         (gather_bbs::after_dom_children): Adjust.
25332 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
25334         PR target/84209
25335         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
25336         * config/avr/avr.md: Only post-reload split REG-REG moves if
25337         either register is GENERAL_REG_P.
25339 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25341         PR tree-optimization/84235
25342         * tree-ssa-scopedtables.c
25343         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
25344         if the subtraction is performed in floating point type where NaNs are
25345         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
25346         build 1.  Formatting fix.
25348 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
25350         PR target/84146
25351         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
25352         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
25353         and skip it regardless of bb boundaries.  Use CALL_P macro,
25354         don't test INSN_P (insn) together with CALL_P or JUMP_P check
25355         unnecessarily, formatting fix.
25357 2018-02-06  Michael Collison  <michael.collison@arm.com>
25359         * config/arm/thumb2.md:
25360         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
25361         (*thumb_mov_notscc): Ditto.
25363 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
25365         PR target/84154
25366         * config/rs6000/rs6000.md (su code attribute): Use "u" for
25367         unsigned_fix, not "s".
25369 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25371         * configure.ac (gcc_fn_eh_frame_ro): New function.
25372         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
25373         correct .eh_frame permissions.
25374         * configure: Regenerate.
25376 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
25378         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
25379         irrelevant options.
25381 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25383         * config/rs6000/rs6000.c (rs6000_option_override_internal):
25384         Display warning message for -mno-speculate-indirect-jumps.
25386 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
25388         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
25389         Undocumented.
25390         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
25392 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
25394         PR tree-optimization/84225
25395         * tree-eh.c (find_trapping_overflow): Only call
25396         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
25398 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25400         PR target/84145
25401         * config/i386/i386.c: Reimplement the check of possible options
25402         -mibt/-mshstk conbination. Change error messages.
25403         * doc/invoke.texi: Fix a typo: remove extra '='.
25405 2018-02-06  Marek Polacek  <polacek@redhat.com>
25407         PR tree-optimization/84228
25408         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
25410 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
25412         PR target/82641
25413         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
25414         emitted arch directives.
25415         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
25416         __ARM_FEATURE_COPROC before changing architectures.
25418 2018-02-06  Richard Biener  <rguenther@suse.de>
25420         * config/i386/i386.c (print_reg): Fix typo.
25421         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
25423 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
25425         * configure: Regenerate.
25427 2018-02-05  Martin Sebor  <msebor@redhat.com>
25429         PR tree-optimization/83369
25430         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
25431         inlining context.
25433 2018-02-05  Martin Liska  <mliska@suse.cz>
25435         * doc/invoke.texi: Cherry-pick upstream r323995.
25437 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
25439         * ira.c (ira_init_register_move_cost): Adjust comment.
25441 2018-02-05  Martin Liska  <mliska@suse.cz>
25443         PR gcov-profile/84137
25444         * doc/gcov.texi: Fix typo in documentation.
25446 2018-02-05  Martin Liska  <mliska@suse.cz>
25448         PR gcov-profile/83879
25449         * doc/gcov.texi: Document necessity of --dynamic-list-data when
25450         using dlopen functionality.
25452 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
25454         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
25455         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
25456         _mm_maskz_range_ss, _mm_mask_range_round_ss,
25457         _mm_maskz_range_round_ss): New intrinsics.
25458         (__builtin_ia32_rangesd128_round)
25459         (__builtin_ia32_rangess128_round): Remove.
25460         (__builtin_ia32_rangesd128_mask_round,
25461         __builtin_ia32_rangess128_mask_round): New builtins.
25462         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
25463         __builtin_ia32_rangess128_round): Remove.
25464         (__builtin_ia32_rangesd128_mask_round,
25465         __builtin_ia32_rangess128_mask_round): New builtins.
25466         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
25467         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
25468         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
25469         "<round_saeonly_constraint>")): Changed to ...
25470         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
25471         "<round_saeonly_scalar_constraint>")): ... this.
25472         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25473         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
25474         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
25475         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
25476         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
25478 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
25480         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
25481         options.
25482         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
25483         Remove all values except native, 8540 and 8548.
25485 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
25487         * config/i386/i386.c (ix86_output_function_return): Pass
25488         INVALID_REGNUM, instead of -1, as invalid register number to
25489         indirect_thunk_name and output_indirect_thunk.
25491 2018-02-02  Julia Koval  <julia.koval@intel.com>
25493         * config.gcc: Add -march=icelake.
25494         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
25495         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
25496         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
25497         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
25498         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
25499         (processor_target_table): Add icelake.
25500         (ix86_option_override_internal): Handle new PTAs.
25501         (get_builtin_code_for_version): Handle icelake.
25502         (M_INTEL_COREI7_ICELAKE): New.
25503         (fold_builtin_cpu): Handle icelake.
25504         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
25505         * doc/invoke.texi: Add -march=icelake.
25507 2018-02-02  Julia Koval  <julia.koval@intel.com>
25509         * config/i386/i386.c (ix86_option_override_internal): Change flags type
25510         to wide_int_bitmask.
25511         * wide-int-bitmask.h: New.
25513 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25515         PR target/84066
25516         * config/i386/i386.md: Replace Pmode with word_mode in
25517         builtin_setjmp_setup and builtin_longjmp to support x32.
25519 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
25521         PR target/56010
25522         PR target/83743
25523         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
25524         #include "opts.h".
25525         (rs6000_supported_cpu_names): New static variable.
25526         (linux_cpu_translation_table): Likewise.
25527         (elf_platform) <cpu>: Define new static variable and use it.
25528         Translate kernel AT_PLATFORM name to canonical name if needed.
25529         Error if platform name is unknown.
25531 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
25533         PR target/84089
25534         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
25536 2018-02-01  Jeff Law  <law@redhat.com>
25538         PR target/84128
25539         * config/i386/i386.c (release_scratch_register_on_entry): Add new
25540         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
25541         the scratch if RELEASE_VIA_POP is false.
25542         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
25543         If we have to save a temporary register, decrement SIZE appropriately.
25544         Pass new arguments to release_scratch_register_on_entry.
25545         (ix86_adjust_stack_and_probe): Likewise.
25546         (ix86_emit_probe_stack_range): Pass new arguments to
25547         release_scratch_register_on_entry.
25549 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
25551         PR rtl-optimization/84157
25552         * combine.c (change_zero_ext): Use REG_P predicate in
25553         front of HARD_REGISTER_P predicate.
25555 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
25557         * config/avr/avr.c (avr_option_override): Move disabling of
25558         -fdelete-null-pointer-checks to...
25559         * common/config/avr/avr-common.c (avr_option_optimization_table):
25560         ...here.
25562 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25564         PR tree-optimization/81635
25565         * tree-data-ref.c (split_constant_offset_1): For types that
25566         wrap on overflow, try to use range info to prove that wrapping
25567         cannot occur.
25569 2018-02-01  Renlin Li  <renlin.li@arm.com>
25571         PR target/83370
25572         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
25573         TAILCALL_ADDR_REGS.
25574         (aarch64_register_move_cost): Likewise.
25575         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
25576         TAILCALL_ADDR_REGS.
25577         (REG_CLASS_NAMES): Likewise.
25578         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
25579         TAILCALL_ADDR_REGS. Remove IP registers.
25580         * config/aarch64/aarch64.md (Ucs): Update register constraint.
25582 2018-02-01  Richard Biener  <rguenther@suse.de>
25584         * domwalk.h (dom_walker::dom_walker): Add additional constructor
25585         for specifying RPO order and allow NULL for that.
25586         * domwalk.c (dom_walker::dom_walker): Likewise.
25587         (dom_walker::walk): Handle NULL RPO order.
25588         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
25589         in RPO order.
25590         (rewrite_update_dom_walker): Likewise.
25591         (mark_def_dom_walker): Likewise.
25593 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25595         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
25596         (aarch64_maybe_expand_sve_subreg_move): Declare.
25597         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
25598         * config/aarch64/predicates.md (aarch64_any_register_operand): New
25599         predicate.
25600         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
25601         that are semantically a reverse operation.
25602         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
25603         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
25604         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
25605         functions.
25606         (aarch64_can_change_mode_class): For big-endian, forbid changes
25607         between two SVE modes if they have different element sizes.
25609 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25611         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
25612         the TImode handling for big-endian targets.
25614 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25616         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
25617         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
25618         not just bytes.
25619         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
25620         Remove BSWAP handing for big-endian targets and use the form of
25621         LD1RQ appropariate for the mode.
25623 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25625         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
25626         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
25627         duplicated element.
25629 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25631         PR tearget/83845
25632         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
25633         check for operands that need to go through aarch64_sve_reload_be.
25635 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
25637         PR tree-optimization/81661
25638         PR tree-optimization/84117
25639         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
25640         * tree-eh.c: Include gimplify.h.
25641         (find_trapping_overflow, replace_trapping_overflow,
25642         rewrite_to_non_trapping_overflow): New functions.
25643         * tree-vect-loop.c: Include tree-eh.h.
25644         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
25645         * tree-data-ref.c: Include tree-eh.h.
25646         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
25648 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
25650         PR rtl-optimization/84123
25651         * combine.c (change_zero_ext): Check if hard register satisfies
25652         can_change_dest_mode before calling gen_lowpart_SUBREG.
25654 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
25656         PR target/82444
25657         * ira.c (ira_init_register_move_cost): Remove assert.
25659 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25661         PR rtl-optimization/84071
25662         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
25663         * doc/tm.texi: Regenerate.
25665 2018-01-31  Richard Biener  <rguenther@suse.de>
25667         PR tree-optimization/84132
25668         * tree-data-ref.c (analyze_miv_subscript): Properly
25669         check whether evolution_function_is_affine_multivariate_p
25670         before calling gcd_of_steps_may_divide_p.
25672 2018-01-31  Julia Koval  <julia.koval@intel.com>
25674         PR target/83618
25675         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
25676         * config/i386/i386.md (rdpid_rex64) New.
25677         (rdpid): Make 32bit only.
25679 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
25681         PR lto/84105
25682         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
25683         an IDENTIFIER_NODE for FUNCTION_TYPE's.
25685 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25687         Revert
25688         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25690         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25692 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25694         PR rtl-optimization/84071
25695         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
25696         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
25698 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25700         * config/arc/arc.c (arc_handle_aux_attribute): New function.
25701         (arc_attribute_table): Add 'aux' attribute.
25702         (arc_in_small_data_p): Consider aux like variables.
25703         (arc_is_aux_reg_p): New function.
25704         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
25705         (arc_get_aux_arg): New function.
25706         (prepare_move_operands): Handle aux-register access.
25707         (arc_handle_aux_attribute): New function.
25708         * doc/extend.texi (ARC Variable attributes): Add subsection.
25710 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25712         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
25713         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
25714         (arc_attribute_table): Add 'uncached' attribute.
25715         (arc_print_operand): Print '.di' flag for uncached memory
25716         accesses.
25717         (arc_in_small_data_p): Do not consider for small data the uncached
25718         types.
25719         (arc_is_uncached_mem_p): New function.
25720         * config/arc/predicates.md (compact_store_memory_operand): Check
25721         for uncached memory accesses.
25722         (nonvol_nonimm_operand): Likewise.
25723         * doc/extend.texi (ARC Type Attribute): New subsection.
25725 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
25727         PR c/84100
25728         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
25729         falign-loops=): Add Optimization flag.
25731 2018-01-30  Jeff Law  <law@redhat.com>
25733         PR target/84064
25734         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
25735         INT_REGISTERS_SAVED.  Check it prior to calling
25736         get_scratch_register_on_entry.
25737         (ix86_adjust_stack_and_probe): Similarly.
25738         (ix86_emit_probe_stack_range): Similarly.
25739         (ix86_expand_prologue): Corresponding changes.
25741 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25743         PR target/40411
25744         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
25745         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
25747 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
25749         PR target/84112
25750         * lra-constraints.c (curr_insn_transform): Process AND in the
25751         address.
25753 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
25755         PR rtl-optimization/83986
25756         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
25757         dependence against last_pending_memory_flush in addition to
25758         pending_jump_insns.
25760 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
25762         PR tree-optimization/81611
25763         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
25764         copies.
25766 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25768         PR target/83758
25769         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
25770         a reg rtx.
25772 2018-01-30  Richard Biener  <rguenther@suse.de>
25773             Jakub Jelinek  <jakub@redhat.com>
25775         PR tree-optimization/84111
25776         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
25777         inner loops added during recursion, as they don't have up-to-date
25778         SSA form.
25780 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
25782         PR ipa/81360
25783         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
25784         (can_inline_edge_by_limits_p): ... here.
25785         (can_early_inline_edge_p, check_callers,
25786         update_caller_keys, update_callee_keys, recursive_inlining,
25787         add_new_edges_to_heap, speculation_useful_p,
25788         inline_small_functions,
25789         inline_small_functions, flatten_function,
25790         inline_to_all_callers_1): Update.
25792 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
25794         * profile-count.c (profile_count::combine_with_ipa_count): Handle
25795         zeros correctly.
25797 2018-01-30  Richard Biener  <rguenther@suse.de>
25799         PR tree-optimization/83008
25800         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
25801         invariant and constant vector uses in stmts when they need
25802         more than one stmt.
25804 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25806         PR bootstrap/84017
25807         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
25808         * configure: Regenerate.
25810 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25812         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
25813         pattern.
25814         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
25815         Use gen_rtx_REG rather than gen_lowpart.
25817 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25819         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
25820         rather than 0 when creating partial subregs.
25822 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25824         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
25825         of usage.
25827 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
25829         PR target/81550
25830         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
25831         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
25832         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
25833         flags.  This restores the settings used before the 2017-07-24.
25834         Turning off pre increment/decrement/modify allows IVOPTS to
25835         optimize DF/SF loops where the index is an int.
25837 2018-01-29  Richard Biener  <rguenther@suse.de>
25838             Kelvin Nilsen  <kelvin@gcc.gnu.org>
25840         PR bootstrap/80867
25841         * tree-vect-stmts.c (vectorizable_call): Don't call
25842         targetm.vectorize_builtin_md_vectorized_function if callee is
25843         NULL.
25845 2018-01-22  Carl Love  <cel@us.ibm.com>
25847         * doc/extend.tex: Fix typo in second arg in
25848         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
25850 2018-01-29  Richard Biener  <rguenther@suse.de>
25852         PR tree-optimization/84086
25853         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
25854         (flush_ssaname_freelist): When SSA names were released reset
25855         the SCEV hash table.
25857 2018-01-29  Richard Biener  <rguenther@suse.de>
25859         PR tree-optimization/84057
25860         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
25861         removed paths when removing edges.
25863 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
25865         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
25866         -mfunction-return=@var{choice}.
25868 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
25870         PR diagnostic/84034
25871         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
25872         Handle CR like TAB.
25873         (layout::print_source_line): Likewise.
25874         (test_get_line_width_without_trailing_whitespace): Add test cases.
25876 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
25878         PR middle-end/84040
25879         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
25880         debug insns.
25882 2018-01-26  Jim Wilson  <jimw@sifive.com>
25884         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
25886         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
25887         specified.
25889 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25891         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
25892         and CMP + SUB-immediate -> SUBS.
25894 2018-01-26  Martin Sebor  <msebor@redhat.com>
25896         PR tree-optimization/83896
25897         * tree-ssa-strlen.c (get_string_len): Rename...
25898         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
25899         Avoid assuming length is constant.
25900         (handle_char_store): Use HOST_WIDE_INT for string length.
25902 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
25904         PR target/81763
25905         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
25906         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
25908 2018-01-26  Richard Biener  <rguenther@suse.de>
25910         PR rtl-optimization/84003
25911         * dse.c (record_store): Only record redundant stores when
25912         the earlier store aliases at least all accesses the later one does.
25914 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
25916         PR rtl-optimization/83985
25917         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
25918         REG_CFA_RESTORE insns.
25919         (delete_unmarked_insns): Don't ignore separate shrink wrapping
25920         REG_CFA_RESTORE insns here.
25922         PR c/83989
25923         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
25924         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
25926 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25928         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
25929         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
25930         (arc_init): Likewise.
25931         (arc_override_options): Likewise.
25932         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
25933         value.
25934         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
25935         support.
25936         * config/arc/arc.h (TARGET_DBNZ): Define.
25937         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
25938         properly set the tune attribute.
25939         (dbnz): Use TARGET_DBNZ guard.
25940         * config/arc/arc.opt (mtune): Add core3 option.
25942 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25944         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
25945         recognize new pic like addresses.
25946         (arc_delegitimize_address): Clean up.
25948 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25950         * config/arc/arc-arches.def: Option mrf16 valid for all
25951         architectures.
25952         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
25953         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
25954         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
25955         * config/arc/arc-tables.opt: Regenerate.
25956         * config/arc/arc.c (arc_conditional_register_usage): Handle
25957         reduced register file case.
25958         (arc_file_start): Set must have build attributes.
25959         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
25960         mrf16 option value.
25961         * config/arc/arc.opt (mrf16): Add new option.
25962         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
25963         * config/arc/genmultilib.awk: Handle new mrf16 option.
25964         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
25965         * config/arc/t-multilib: Regenerate.
25966         * doc/invoke.texi (ARC Options): Document mrf16 option.
25968 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25970         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
25971         * config/arc/arc.c (arc_handle_secure_attribute): New function.
25972         (arc_attribute_table): Add 'secure_call' attribute.
25973         (arc_print_operand): Print secure call operand.
25974         (arc_function_ok_for_sibcall): Don't optimize tail calls when
25975         secure.
25976         (arc_is_secure_call_p): New function.  * config/arc/arc.md
25977         (call_i): Add support for sjli instruction.
25978         (call_value_i): Likewise.
25979         * config/arc/constraints.md (Csc): New constraint.
25981 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25982             John Eric Martin  <John.Martin@emmicro-us.com>
25984         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
25985         * config/arc/arc.c (_arc_jli_section): New struct.
25986         (arc_jli_section): New type.
25987         (rc_jli_sections): New static variable.
25988         (arc_handle_jli_attribute): New function.
25989         (arc_attribute_table): Add jli_always and jli_fixed attribute.
25990         (arc_file_end): New function.
25991         (TARGET_ASM_FILE_END): Define.
25992         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
25993         (arc_add_jli_section): New function.
25994         (jli_call_scan): Likewise.
25995         (arc_reorg): Call jli_call_scan.
25996         (arc_output_addsi): Remove 'S' from printing asm operand.
25997         (arc_is_jli_call_p): New function.
25998         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
25999         operand.
26000         (movhi_insn): Likewise.
26001         (movsi_insn): Likewise.
26002         (movsi_set_cc_insn): Likewise.
26003         (loadqi_update): Likewise.
26004         (load_zeroextendqisi_update): Likewise.
26005         (load_signextendqisi_update): Likewise.
26006         (loadhi_update): Likewise.
26007         (load_zeroextendhisi_update): Likewise.
26008         (load_signextendhisi_update): Likewise.
26009         (loadsi_update): Likewise.
26010         (loadsf_update): Likewise.
26011         (movsicc_insn): Likewise.
26012         (bset_insn): Likewise.
26013         (bxor_insn): Likewise.
26014         (bclr_insn): Likewise.
26015         (bmsk_insn): Likewise.
26016         (bicsi3_insn): Likewise.
26017         (cmpsi_cc_c_insn): Likewise.
26018         (movsi_ne): Likewise.
26019         (movsi_cond_exec): Likewise.
26020         (clrsbsi2): Likewise.
26021         (norm_f): Likewise.
26022         (normw): Likewise.
26023         (swap): Likewise.
26024         (divaw): Likewise.
26025         (flag): Likewise.
26026         (sr): Likewise.
26027         (kflag): Likewise.
26028         (ffs): Likewise.
26029         (ffs_f): Likewise.
26030         (fls): Likewise.
26031         (call_i): Remove 'S' asm letter, add jli instruction.
26032         (call_value_i): Likewise.
26033         * config/arc/arc.op (mjli-always): New option.
26034         * config/arc/constraints.md (Cji): New constraint.
26035         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26036         operand.
26037         (subsf3_fpx): Likewise.
26038         (mulsf3_fpx): Likewise.
26039         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26040         asm operand.
26041         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26042         function attrbutes.
26043         * doc/invoke.texi (ARC): Document mjli-always option.
26045 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26047         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26048         avoid addition with 0 and use incw and decw where possible.
26050 2018-01-26  Richard Biener  <rguenther@suse.de>
26052         PR tree-optimization/81082
26053         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26054         association if it requires casting to unsigned.
26055         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26056         from fold_plusminus_mult_expr to catch important cases late when
26057         range info is available.
26059 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26061         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26062         * configure.ac (hidden_linkonce): New test.
26063         * configure: Regenerate.
26064         * config.in: Regenerate.
26066 2018-01-26  Julia Koval  <julia.koval@intel.com>
26068         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26069         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26070         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26071         _mm_mask_bitshuffle_epi64_mask): Fix type.
26072         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26073         USI_FTYPE_V4DI_V4DI_USI): Remove.
26074         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26075         __builtin_ia32_vpshufbitqmb256_mask,
26076         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26077         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26078         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26080 2018-01-26  Alan Modra  <amodra@gmail.com>
26082         PR target/84033
26083         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26084         UNSPEC_VBPERMQ.  Sort other unspecs.
26086 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
26088         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26090 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
26092         PR middle-end/83055
26093         * predict.c (drop_profile): Do not push/pop cfun; update also
26094         node->count.
26095         (handle_missing_profiles): Fix logic looking for zero profiles.
26097 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
26099         PR middle-end/83977
26100         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26101         on functions with #pragma omp declare simd or functions with simd
26102         attribute.
26103         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26104         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26105         Remove trailing \n from warning_at calls.
26107 2018-01-25  Tom de Vries  <tom@codesourcery.com>
26109         PR target/84028
26110         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26111         for neutered workers.
26113 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
26115         PR target/68467
26116         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26117         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26119 2018-01-24  Jeff Law  <law@redhat.com>
26121         PR target/83994
26122         * i386.c (get_probe_interval): Move to earlier point.
26123         (ix86_compute_frame_layout): If -fstack-clash-protection and
26124         the frame is larger than the probe interval, then use pushes
26125         to save registers rather than reg->mem moves.
26126         (ix86_expand_prologue): Remove conditional for int_registers_saved
26127         assertion.
26129 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
26131         PR target/84014
26132         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26133         min/max for never referenced object.
26135 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
26137         PR middle-end/83977
26138         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26139         here.
26140         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26141         attributes from DECL_ATTRIBUTES (decl) without affecting
26142         DECL_ATTRIBUTES (current_function_decl).
26143         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26144         functions with non-NULL DECL_ABSTRACT_ORIGIN.
26146 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
26148         PR tree-optimization/83979
26149         * fold-const.c (fold_comparison): Use constant_boolean_node
26150         instead of boolean_{true,false}_node.
26152 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
26154         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26155         with zero counts.
26157 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26159         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26160         Simplify the clause that sets the length attribute.
26161         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26162         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26163         clause that sets the length attribute.
26164         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26166 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26168         PR target/83589
26169         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26170         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26171         Add strict parameter.
26172         (prevent_branch_around_nothing): Insert dummy insn between branch to
26173         label and label with no ptx insn inbetween.
26174         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26176 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26178         PR target/81352
26179         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26180         for neutered threads in warp.
26181         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26183 2018-01-24  Richard Biener  <rguenther@suse.de>
26185         PR tree-optimization/83176
26186         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26187         operands.
26189 2018-01-24  Richard Biener  <rguenther@suse.de>
26191         PR tree-optimization/82819
26192         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26193         code generating pluses that are no-ops in the target precision.
26195 2018-01-24  Richard Biener  <rguenther@suse.de>
26197         PR middle-end/84000
26198         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26200 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26202         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26203         to merge probabilities.
26204         * predict.c (probably_never_executed): Also mark as cold functions
26205         with global 0 profile and guessed local profile.
26206         * profile-count.c (profile_probability::combine_with_count): New
26207         member function.
26208         * profile-count.h (profile_probability::operator*,
26209         profile_probability::operator*=, profile_probability::operator/,
26210         profile_probability::operator/=): Reduce precision to adjusted
26211         and set value to guessed on contradictory divisions.
26212         (profile_probability::combine_with_freq): Remove.
26213         (profile_probability::combine_wiht_count): Declare.
26214         (profile_count::force_nonzero):: Set to adjusted.
26215         (profile_count::probability_in):: Set quality to adjusted.
26216         * tree-ssa-tail-merge.c (replace_block_by): Use
26217         combine_with_count.
26219 2018-01-23  Andrew Waterman  <andrew@sifive.com>
26220             Jim Wilson  <jimw@sifive.com>
26222         * config/riscv/riscv.c (riscv_stack_boundary): New.
26223         (riscv_option_override): Set riscv_stack_boundary.  Handle
26224         riscv_preferred_stack_boundary_arg.
26225         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26226         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26227         (STACK_BOUNDARY): Set to riscv_stack_boundary.
26228         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26229         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26230         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26232 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
26234         PR target/83905
26235         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26236         of struct ix86_frame.
26237         (ix86_expand_epilogue): Likewise.  Add a local variable for
26238         the reg_save_offset field in struct ix86_frame.
26240 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
26242         PR tree-optimization/82604
26243         * tree-loop-distribution.c (enum partition_kind): New enum item
26244         PKIND_PARTIAL_MEMSET.
26245         (partition_builtin_p): Support above new enum item.
26246         (generate_code_for_partition): Ditto.
26247         (compute_access_range): Differentiate cases that equality can be
26248         proven at all loops, the innermost loops or no loops.
26249         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26250         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
26251         (finalize_partitions, distribute_loop): Don't fuse partition of
26252         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26253         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26254         parloop is enabled.
26256 2018-01-23  Martin Liska  <mliska@suse.cz>
26258         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26259         order to ignore the predictor.
26260         (PRED_POLYMORPHIC_CALL): Likewise.
26261         (PRED_RECURSIVE_CALL): Likewise.
26263 2018-01-23  Martin Liska  <mliska@suse.cz>
26265         * tree-profile.c (tree_profiling): Print function header to
26266         aware reader which function we are working on.
26267         * value-prof.c (gimple_find_values_to_profile): Do not print
26268         not interesting value histograms.
26270 2018-01-23  Martin Liska  <mliska@suse.cz>
26272         * profile-count.h (enum profile_quality): Add
26273         profile_uninitialized as the first value. Do not number values
26274         as they are zero based.
26275         (profile_count::verify): Update sanity check.
26276         (profile_probability::verify): Likewise.
26278 2018-01-23  Nathan Sidwell  <nathan@acm.org>
26280         * doc/invoke.texi (ffor-scope): Deprecate.
26282 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26284         PR tree-optimization/83510
26285         * domwalk.c (set_all_edges_as_executable): New function.
26286         (dom_walker::dom_walker): Convert bool param
26287         "skip_unreachable_blocks" to enum reachability.  Move setup of
26288         edge flags to set_all_edges_as_executable and only do it when
26289         reachability is REACHABLE_BLOCKS.
26290         * domwalk.h (enum dom_walker::reachability): New enum.
26291         (dom_walker::dom_walker): Convert bool param
26292         "skip_unreachable_blocks" to enum reachability.
26293         (set_all_edges_as_executable): New decl.
26294         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
26295         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26296         "reachability".
26297         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26298         but converting true to REACHABLE_BLOCKS.
26299         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26300         * tree-vrp.c
26301         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26302         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26303         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26304         REACHABLE_BLOCKS.
26305         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26306         if check_all_array_refs will be called.
26308 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26310         * tree.c (selftest::test_location_wrappers): Add more test
26311         coverage.
26313 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26315         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
26316         (selftest::test_bit_in_range): Likewise.
26318 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26320         PR testsuite/83888
26321         * doc/sourcebuild.texi (vect_float): Say that the selector
26322         only describes the situation when -funsafe-math-optimizations is on.
26323         (vect_float_strict): Document.
26325 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26327         PR tree-optimization/83965
26328         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
26329         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
26330         instead of checking only for a reduction.
26331         (vect_recog_widen_sum_pattern): Likewise.
26333 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26335         * predict.c (probably_never_executed): Only use precise profile info.
26336         (compute_function_frequency): Skip after inlining hack since we now
26337         have quality checking.
26339 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26341         * profile-count.h (profile_probability::very_unlikely,
26342         profile_probability::unlikely, profile_probability::even): Set
26343         precision to guessed.
26345 2018-01-23  Richard Biener  <rguenther@suse.de>
26347         PR tree-optimization/83963
26348         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
26349         Properly terminate dominator walk when crossing the exit edge not
26350         when visiting its source block.
26352 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
26354         PR c++/83918
26355         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
26356         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
26358 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
26360         PR tree-optimization/83957
26361         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
26362         semicolon after for body surrounded by braces.
26364         PR tree-optimization/83081
26365         * profile-count.h (profile_probability::split): New method.
26366         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
26367         Use profile_probability::split.
26368         (do_compare_rtx_and_jump): Fix adjustment of probabilities
26369         when splitting a single conditional jump into 2.
26371 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
26373         PR tree-optimization/69452
26374         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
26375         decl.
26377 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26379         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
26380         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
26381         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
26383 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26385         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
26386         declaration.
26387         * config/rl78/rl78.md (movdi): New define_expand.
26388         * config/rl78/rl78.c (rl78_split_movdi): New function.
26390 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
26392         PR target/83862
26393         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
26394         no longer used.
26395         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
26396         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
26397         128-bit to produce an UNSPEC move to get the double word with the
26398         signbit and then a shift directly to do signbit.
26399         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
26400         implementation with a new version that just does either a direct
26401         move or a regular move.  Move memory interface to separate insns.
26402         Move insns so they are next to the expander.
26403         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
26404         with signbit move.  Split big and little endian case.
26405         (signbit<mode>2_dm_mem_le): Likewise.
26406         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
26407         (signbit<mode>2_dm2): Likewise.
26409 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26411         * config/rl78/rl78.md (anddi3): New define_expand.
26413 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26415         * config/rl78/rl78.md (umindi3): New define_expand.
26417 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26419         * config/rl78/rl78.md (smindi3): New define_expand.
26421 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26423         * config/rl78/rl78.md (smaxdi3): New define_expand.
26425 2018-01-22  Carl Love  <cel@us.ibm.com>
26427         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
26428         LVX_V1TI): Add macro expansion.
26429         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
26430         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
26431         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
26432         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
26433         Change check to determine if the instruction is a byte reversing
26434         entry.  Fix typo in comment.
26435         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
26436         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
26437         Add def_builtin calls for new builtins.
26438         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
26439         Add define_insn expansion.
26441 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26443         * config/rl78/rl78.md (umaxdi3): New define_expand.
26445 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26447         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
26448         for non-QImode registers.
26450 2018-01-22  Richard Biener  <rguenther@suse.de>
26452         PR tree-optimization/83963
26453         * graphite-scop-detection.c (scop_detection::get_sese): Delay
26454         including the loop exit block.
26455         (scop_detection::merge_sese): Likewise.
26456         (scop_detection::add_scop): Do it here instead.
26458 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26460         * doc/sourcebuild.texi (arm_softfloat): Document.
26462 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
26464         PR gcc/77734
26465         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
26466         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
26467         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
26469 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26470             David Edelsohn  <dje.gcc@gmail.com>
26472         PR target/83946
26473         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26474         Change "crset eq" to "crset 2".
26475         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26476         (*call_indirect_aix<mode>_nospec): Likewise.
26477         (*call_value_indirect_aix<mode>_nospec): Likewise.
26478         (*call_indirect_elfv2<mode>_nospec): Likewise.
26479         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
26480         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
26481         change assembly output from . to $.
26482         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26483         (indirect_jump<mode>_nospec): Change assembly output from . to $.
26484         (*tablejump<mode>_internal1_nospec): Likewise.
26486 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
26488         PR target/80870
26489         * config/sh/sh_optimize_sett_clrt.cc:
26490         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
26492 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
26494         PR tree-optimization/83940
26495         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
26496         offset_dt to vect_constant_def rather than vect_unknown_def_type.
26497         (vect_check_load_store_mask): Add a mask_dt_out parameter and
26498         use it to pass back the definition type.
26499         (vect_check_store_rhs): Likewise rhs_dt_out.
26500         (vect_build_gather_load_calls): Add a mask_dt argument and use
26501         it instead of a call to vect_is_simple_use.
26502         (vectorizable_store): Update calls to vect_check_load_store_mask
26503         and vect_check_store_rhs.  Use the dt returned by the latter instead
26504         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
26505         instead of calls to vect_is_simple_use.  Pass the scalar rather
26506         than the vector operand to vect_is_simple_use when handling
26507         second and subsequent copies of an rhs value.
26508         (vectorizable_load): Update calls to vect_check_load_store_mask
26509         and vect_build_gather_load_calls.  Use the cached mask_dt and
26510         gs_info.offset_dt instead of calls to vect_is_simple_use.
26512 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
26514         PR middle-end/83945
26515         * tree-emutls.c: Include gimplify.h.
26516         (lower_emutls_2): New function.
26517         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
26518         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
26519         it before further processing.
26521         PR target/83930
26522         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
26523         UINTVAL (trueop1) instead of INTVAL (op1).
26525 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
26527         PR debug/81570
26528         PR debug/83728
26529         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
26530         INCOMING_FRAME_SP_OFFSET if not defined.
26531         (scan_trace): Add ENTRY argument.  If true and
26532         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
26533         emit a note to adjust the CFA offset.
26534         (create_cfi_notes): Adjust scan_trace callers.
26535         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
26536         INCOMING_FRAME_SP_OFFSET in the CIE.
26537         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
26538         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
26539         Likewise.
26540         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
26541         * doc/tm.texi: Regenerated.
26543 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26545         PR rtl-optimization/83147
26546         * lra-constraints.c (remove_inheritance_pseudos): Use
26547         lra_substitute_pseudo_within_insn.
26549 2018-01-19  Tom de Vries  <tom@codesourcery.com>
26550             Cesar Philippidis  <cesar@codesourcery.com>
26552         PR target/83920
26553         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
26555 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
26557         PR target/83790
26558         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
26559         spaces for function labels.
26561 2018-01-19  Martin Liska  <mliska@suse.cz>
26563         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
26564         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
26565         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
26566         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
26567         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
26568         (PRED_CONST_RETURN): Change from 69 to 65.
26569         (PRED_NULL_RETURN): Change from 91 to 71.
26570         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
26571         (PRED_LOOP_GUARD): Change from 66 to 73.
26573 2018-01-19  Martin Liska  <mliska@suse.cz>
26575         * predict.c (predict_insn_def): Add new assert.
26576         (struct branch_predictor): Change type to signed integer.
26577         (test_prediction_value_range): Amend test to cover
26578         PROB_UNINITIALIZED.
26579         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
26580         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
26581         (PRED_LOOP_ITERATIONS_MAX): Likewise.
26582         (PRED_LOOP_IV_COMPARE): Likewise.
26583         * predict.h (PROB_UNINITIALIZED): Define new constant.
26585 2018-01-19  Martin Liska  <mliska@suse.cz>
26587         * predict.c (dump_prediction): Add new format for
26588         analyze_brprob.py script which is enabled with -details
26589         suboption.
26590         * profile-count.h (precise_p): New function.
26592 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26594         PR tree-optimization/83922
26595         * tree-vect-loop.c (vect_verify_full_masking): Return false if
26596         there are no statements that need masking.
26597         (vect_active_double_reduction_p): New function.
26598         (vect_analyze_loop_operations): Use it when handling phis that
26599         are not in the loop header.
26601 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26603         PR tree-optimization/83914
26604         * tree-vect-loop.c (vectorizable_induction): Don't convert
26605         init_expr or apply the peeling adjustment for inductions
26606         that are nested within the vectorized loop.
26608 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26610         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
26611         instead of NEG.
26613 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26615         PR sanitizer/81715
26616         PR testsuite/83882
26617         * function.h (gimplify_parameters): Add gimple_seq * argument.
26618         * function.c: Include gimple.h and options.h.
26619         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
26620         for the added local temporaries if needed.
26621         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
26622         if there are any parameter cleanups, wrap whole body into a
26623         try/finally with the cleanups.
26625 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
26627         PR target/82964
26628         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
26629         Use GET_MODE_CLASS for scalar floating point.
26631 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26633         PR ipa/82256
26634         patch by PaX Team
26635         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
26636         Fix call of call_cgraph_insertion_hooks.
26638 2018-01-18  Martin Sebor  <msebor@redhat.com>
26640         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
26642 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26644         PR ipa/83619
26645         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
26646         frequencies.
26648 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
26650         PR other/70268
26651         * common.opt: (-ffile-prefix-map): New option.
26652         * opts.c (common_handle_option): Defer it.
26653         * opts-global.c (handle_common_deferred_options): Handle it.
26654         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
26655         * file-prefix-map.h: New file.
26656         (remap_debug_filename, add_debug_prefix_map): ...here.
26657         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
26658         * final.c (debug_prefix_map, add_debug_prefix_map
26659         remap_debug_filename): Move to...
26660         * file-prefix-map.c: New file.
26661         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
26662         generalize, get rid of alloca(), use strrchr() instead of strchr().
26663         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
26664         Implement in terms of add_prefix_map().
26665         (remap_macro_filename, remap_debug_filename): Implement in term of
26666         remap_filename().
26667         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
26668         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
26669         * dbxout.c: Include file-prefix-map.h.
26670         * varasm.c: Likewise.
26671         * vmsdbgout.c: Likewise.
26672         * xcoffout.c: Likewise.
26673         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
26674         * doc/cppopts.texi (-fmacro-prefix-map): Document.
26675         * doc/invoke.texi (-ffile-prefix-map): Document.
26676         (-fdebug-prefix-map): Update description.
26678 2018-01-18  Martin Liska  <mliska@suse.cz>
26680         * config/i386/i386.c (indirect_thunk_name): Document that also
26681         lfence is emitted.
26682         (output_indirect_thunk): Document why both instructions
26683         (pause and lfence) are generated.
26685 2018-01-18  Richard Biener  <rguenther@suse.de>
26687         PR tree-optimization/83887
26688         * graphite-scop-detection.c
26689         (scop_detection::get_nearest_dom_with_single_entry): Remove.
26690         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
26691         (scop_detection::merge_sese): Re-implement with a flood-fill
26692         algorithm that properly finds a SESE region if it exists.
26694 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26696         PR c/61240
26697         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
26698         pointer_diff optimizations use view_convert instead of convert.
26700 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26702         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26703         Generate different code for -mno-speculate-indirect-jumps.
26704         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26705         (*call_indirect_aix<mode>): Disable for
26706         -mno-speculate-indirect-jumps.
26707         (*call_indirect_aix<mode>_nospec): New define_insn.
26708         (*call_value_indirect_aix<mode>): Disable for
26709         -mno-speculate-indirect-jumps.
26710         (*call_value_indirect_aix<mode>_nospec): New define_insn.
26711         (*sibcall_nonlocal_sysv<mode>): Generate different code for
26712         -mno-speculate-indirect-jumps.
26713         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26715 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
26717         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
26718         long double type, set the flags for noting the default long double
26719         type, even if we don't pass or return a long double type.
26721 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
26723         PR ipa/83051
26724         * ipa-inline.c (flatten_function): Do not overwrite final inlining
26725         failure.
26727 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
26729         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
26730         support for merge[hl].
26731         (fold_mergehl_helper): New helper function.
26732         (tree-vector-builder.h): New #include for tree_vector_builder usage.
26733         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
26734         (altivec_vmrglw_direct): Add xxmrglw insn.
26736 2018-01-17  Andrew Waterman  <andrew@sifive.com>
26738         * config/riscv/riscv.c (riscv_conditional_register_usage): If
26739         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
26741 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
26743         PR lto/83121
26744         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
26745         call the lto_location_cache before reading the
26746         DECL_SOURCE_LOCATION of the types.
26748 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
26749             Richard Sandiford  <richard.sandiford@linaro.org>
26751         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
26752         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
26753         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
26754         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
26755         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
26756         Add declaration.
26757         * config/aarch64/constraints.md (aarch64_movti_operand):
26758         Limit immediates.
26759         * config/aarch64/predicates.md (Uti): Add new constraint.
26761 2018-01-17  Carl Love  <cel@us.ibm.com>
26763         * config/rs6000/vsx.md (define_expand xl_len_r,
26764         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
26765         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
26766         lxvll.
26767         (define_expand, define_insn): Move the shift left from  the
26768         define_insn to the define_expand for lxvl and stxvl instructions.
26769         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
26770         and XL_LEN_R definitions to PURE.
26772 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
26774         * config/i386/i386.c (indirect_thunk_name): Declare regno
26775         as unsigned int.  Compare regno with INVALID_REGNUM.
26776         (output_indirect_thunk): Ditto.
26777         (output_indirect_thunk_function): Ditto.
26778         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
26779         in the call to output_indirect_thunk_function.
26781 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
26783         PR middle-end/83884
26784         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
26785         rather than the size of inner_type to determine the stack slot size
26786         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
26788 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
26790         PR target/83546
26791         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
26792         to PTA_SILVERMONT.
26794 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
26796         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
26797         endian Linux systems to optionally enable multilibs for selecting
26798         the long double type if the user configured an explicit type.
26799         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
26800         have no long double multilibs if not defined.
26801         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
26802         warn if the user used -mabi={ieee,ibm}longdouble and we built
26803         multilibs for long double.
26804         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
26805         appropriate multilib option.
26806         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
26807         multilib options.
26808         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
26809         for building long double multilibs.
26810         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
26812 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
26814         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
26815         copies.
26817         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
26818         64 bits.
26819         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
26820         128 bits.
26822         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
26823         variables.
26825         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
26826         return value.
26828 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
26830         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
26831         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
26833 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26835         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
26836         different rtl trees depending on TARGET_64BIT.
26837         (rs6000_gen_lvx): Likewise.
26839 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
26841         * config/visium/visium.md (nop): Tweak comment.
26842         (hazard_nop): Likewise.
26844 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26846         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
26847         -mspeculate-indirect-jumps.
26848         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
26849         for -mno-speculate-indirect-jumps.
26850         (*call_indirect_elfv2<mode>_nospec): New define_insn.
26851         (*call_value_indirect_elfv2<mode>): Disable for
26852         -mno-speculate-indirect-jumps.
26853         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
26854         (indirect_jump): Emit different RTL for
26855         -mno-speculate-indirect-jumps.
26856         (*indirect_jump<mode>): Disable for
26857         -mno-speculate-indirect-jumps.
26858         (*indirect_jump<mode>_nospec): New define_insn.
26859         (tablejump): Emit different RTL for
26860         -mno-speculate-indirect-jumps.
26861         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
26862         (tablejumpsi_nospec): New define_expand.
26863         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
26864         (tablejumpdi_nospec): New define_expand.
26865         (*tablejump<mode>_internal1): Disable for
26866         -mno-speculate-indirect-jumps.
26867         (*tablejump<mode>_internal1_nospec): New define_insn.
26868         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
26869         option.
26871 2018-01-16  Artyom Skrobov tyomitch@gmail.com
26873         * caller-save.c (insert_save): Drop unnecessary parameter.  All
26874         callers updated.
26876 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26877             Richard Biener  <rguenth@suse.de>
26879         PR libgomp/83590
26880         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
26881         return early, inline manually is_gimple_sizepos.  Make sure if we
26882         call gimplify_expr we don't end up with a gimple constant.
26883         * tree.c (variably_modified_type_p): Don't return true for
26884         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
26885         * gimplify.h (is_gimple_sizepos): Remove.
26887 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26889         PR tree-optimization/83857
26890         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
26891         vectorizable_live_operation for pure SLP statements.
26892         (vectorizable_live_operation): Handle PHIs.
26894 2018-01-16  Richard Biener  <rguenther@suse.de>
26896         PR tree-optimization/83867
26897         * tree-vect-stmts.c (vect_transform_stmt): Precompute
26898         nested_in_vect_loop_p since the scalar stmt may get invalidated.
26900 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26902         PR c/83844
26903         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
26904         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
26905         If off is not INTEGER_CST, issue a may not be aligned warning
26906         rather than isn't aligned.  Use isn%'t rather than isn't.
26907         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
26908         into MULT_EXPR.
26909         <case MULT_EXPR>: Improve the case when bottom and one of the
26910         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
26911         operand, in that case check if the other operand is multiple of
26912         bottom divided by the INTEGER_CST operand.
26914 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26916         PR target/83858
26917         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
26918         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
26919         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
26920         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
26921         * config/pa/pa.c (pa_function_arg_advance): Likewise.
26922         (pa_function_arg, pa_arg_partial_bytes): Likewise.
26923         (pa_function_arg_size): New function.
26925 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26927         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
26928         in a separate statement.
26930 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26932         PR tree-optimization/83847
26933         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
26934         group gathers and scatters.
26936 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26938         PR rtl-optimization/86620
26939         * params.def (max-sched-ready-insns): Bump minimum value to 1.
26941         PR rtl-optimization/83213
26942         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
26943         to last if both are JUMP_INSNs.
26945         PR tree-optimization/83843
26946         * gimple-ssa-store-merging.c
26947         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
26948         store_immediate_info for bswap/nop orig_stores.
26950 2018-01-15  Andrew Waterman  <andrew@sifive.com>
26952         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
26953         !TARGET_MUL.
26954         <UDIV>: Increase cost if !TARGET_DIV.
26956 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
26958         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
26959         (define_attr "cr_logical_3op"): New.
26960         (cceq_ior_compare): Adjust.
26961         (cceq_ior_compare_complement): Adjust.
26962         (*cceq_rev_compare): Adjust.
26963         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
26964         (is_cracked_insn): Adjust.
26965         (insn_must_be_first_in_group): Adjust.
26966         * config/rs6000/40x.md: Adjust.
26967         * config/rs6000/440.md: Adjust.
26968         * config/rs6000/476.md: Adjust.
26969         * config/rs6000/601.md: Adjust.
26970         * config/rs6000/603.md: Adjust.
26971         * config/rs6000/6xx.md: Adjust.
26972         * config/rs6000/7450.md: Adjust.
26973         * config/rs6000/7xx.md: Adjust.
26974         * config/rs6000/8540.md: Adjust.
26975         * config/rs6000/cell.md: Adjust.
26976         * config/rs6000/e300c2c3.md: Adjust.
26977         * config/rs6000/e500mc.md: Adjust.
26978         * config/rs6000/e500mc64.md: Adjust.
26979         * config/rs6000/e5500.md: Adjust.
26980         * config/rs6000/e6500.md: Adjust.
26981         * config/rs6000/mpc.md: Adjust.
26982         * config/rs6000/power4.md: Adjust.
26983         * config/rs6000/power5.md: Adjust.
26984         * config/rs6000/power6.md: Adjust.
26985         * config/rs6000/power7.md: Adjust.
26986         * config/rs6000/power8.md: Adjust.
26987         * config/rs6000/power9.md: Adjust.
26988         * config/rs6000/rs64.md: Adjust.
26989         * config/rs6000/titan.md: Adjust.
26991 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26993         * config/i386/predicates.md (indirect_branch_operand): Rewrite
26994         ix86_indirect_branch_register logic.
26996 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26998         * config/i386/constraints.md (Bs): Update
26999         ix86_indirect_branch_register check.  Don't check
27000         ix86_indirect_branch_register with GOT_memory_operand.
27001         (Bw): Likewise.
27002         * config/i386/predicates.md (GOT_memory_operand): Don't check
27003         ix86_indirect_branch_register here.
27004         (GOT32_symbol_operand): Likewise.
27006 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27008         * config/i386/predicates.md (constant_call_address_operand):
27009         Rewrite ix86_indirect_branch_register logic.
27010         (sibcall_insn_operand): Likewise.
27012 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27014         * config/i386/constraints.md (Bs): Replace
27015         ix86_indirect_branch_thunk_register with
27016         ix86_indirect_branch_register.
27017         (Bw): Likewise.
27018         * config/i386/i386.md (indirect_jump): Likewise.
27019         (tablejump): Likewise.
27020         (*sibcall_memory): Likewise.
27021         (*sibcall_value_memory): Likewise.
27022         Peepholes of indirect call and jump via memory: Likewise.
27023         * config/i386/i386.opt: Likewise.
27024         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27025         (GOT_memory_operand): Likewise.
27026         (call_insn_operand): Likewise.
27027         (sibcall_insn_operand): Likewise.
27028         (GOT32_symbol_operand): Likewise.
27030 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27032         PR middle-end/83837
27033         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27034         type rather than type addr's type points to.
27035         (expand_omp_atomic_mutex): Likewise.
27036         (expand_omp_atomic): Likewise.
27038 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27040         PR target/83839
27041         * config/i386/i386.c (output_indirect_thunk_function): Use
27042         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27043         for  __x86_return_thunk.
27045 2018-01-15  Richard Biener  <rguenther@suse.de>
27047         PR middle-end/83850
27048         * expmed.c (extract_bit_field_1): Fix typo.
27050 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27052         PR target/83687
27053         * config/arm/iterators.md (VF): New mode iterator.
27054         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27055         Remove integer-related logic from pattern.
27056         (neon_vabd<mode>_3): Likewise.
27058 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27060         PR middle-end/82694
27061         * common.opt (fstrict-overflow): No longer an alias.
27062         (fwrapv-pointer): New option.
27063         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27064         also for pointer types based on flag_wrapv_pointer.
27065         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27066         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27067         opts->x_flag_wrapv got set.
27068         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27069         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27070         POINTER_TYPE_OVERFLOW_UNDEFINED.
27071         * match.pd: Likewise in address comparison pattern.
27072         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27074 2018-01-15  Richard Biener  <rguenther@suse.de>
27076         PR lto/83804
27077         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27078         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
27079         Reset type names to their identifier if their TYPE_DECL doesn't
27080         have linkage (and thus is used for ODR and devirt).
27081         (save_debug_info_for_decl): Remove.
27082         (save_debug_info_for_type): Likewise.
27083         (add_tree_to_fld_list): Adjust.
27084         * tree-pretty-print.c (dump_generic_node): Make dumping of
27085         type names more robust.
27087 2018-01-15  Richard Biener  <rguenther@suse.de>
27089         * BASE-VER: Bump to 8.0.1.
27091 2018-01-14  Martin Sebor  <msebor@redhat.com>
27093         PR other/83508
27094         * builtins.c (check_access): Avoid warning when the no-warning bit
27095         is set.
27097 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
27099         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27100         * ira-color (allocno_hard_regs_compare): Likewise.
27102 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
27104         PR target/83013
27105         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27106         Use .pushsection/.popsection.
27108 2018-01-14  Martin Sebor  <msebor@redhat.com>
27110         PR c++/81327
27111         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27113 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
27115         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27116         entry from extra_headers.
27117         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27118         extra_headers, make the list bitwise identical to the i?86-*-* one.
27120 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27122         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27123         -mcmodel=large with -mindirect-branch=thunk,
27124         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27125         -mfunction-return=thunk-extern.
27126         * doc/invoke.texi: Document -mcmodel=large is incompatible with
27127         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27128         -mfunction-return=thunk and -mfunction-return=thunk-extern.
27130 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27132         * config/i386/i386.c (print_reg): Print the name of the full
27133         integer register without '%'.
27134         (ix86_print_operand): Handle 'V'.
27135         * doc/extend.texi: Document 'V' modifier.
27137 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27139         * config/i386/constraints.md (Bs): Disallow memory operand for
27140         -mindirect-branch-register.
27141         (Bw): Likewise.
27142         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27143         (GOT_memory_operand): Likewise.
27144         (call_insn_operand): Likewise.
27145         (sibcall_insn_operand): Likewise.
27146         (GOT32_symbol_operand): Likewise.
27147         * config/i386/i386.md (indirect_jump): Call convert_memory_address
27148         for -mindirect-branch-register.
27149         (tablejump): Likewise.
27150         (*sibcall_memory): Likewise.
27151         (*sibcall_value_memory): Likewise.
27152         Disallow peepholes of indirect call and jump via memory for
27153         -mindirect-branch-register.
27154         (*call_pop): Replace m with Bw.
27155         (*call_value_pop): Likewise.
27156         (*sibcall_pop_memory): Replace m with Bs.
27157         * config/i386/i386.opt (mindirect-branch-register): New option.
27158         * doc/invoke.texi: Document -mindirect-branch-register option.
27160 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27162         * config/i386/i386-protos.h (ix86_output_function_return): New.
27163         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27164         set function_return_type.
27165         (indirect_thunk_name): Add ret_p to indicate thunk for function
27166         return.
27167         (output_indirect_thunk_function): Pass false to
27168         indirect_thunk_name.
27169         (ix86_output_indirect_branch_via_reg): Likewise.
27170         (ix86_output_indirect_branch_via_push): Likewise.
27171         (output_indirect_thunk_function): Create alias for function
27172         return thunk if regno < 0.
27173         (ix86_output_function_return): New function.
27174         (ix86_handle_fndecl_attribute): Handle function_return.
27175         (ix86_attribute_table): Add function_return.
27176         * config/i386/i386.h (machine_function): Add
27177         function_return_type.
27178         * config/i386/i386.md (simple_return_internal): Use
27179         ix86_output_function_return.
27180         (simple_return_internal_long): Likewise.
27181         * config/i386/i386.opt (mfunction-return=): New option.
27182         (indirect_branch): Mention -mfunction-return=.
27183         * doc/extend.texi: Document function_return function attribute.
27184         * doc/invoke.texi: Document -mfunction-return= option.
27186 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27188         * config/i386/i386-opts.h (indirect_branch): New.
27189         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27190         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27191         with local indirect jump when converting indirect call and jump.
27192         (ix86_set_indirect_branch_type): New.
27193         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27194         (indirectlabelno): New.
27195         (indirect_thunk_needed): Likewise.
27196         (indirect_thunk_bnd_needed): Likewise.
27197         (indirect_thunks_used): Likewise.
27198         (indirect_thunks_bnd_used): Likewise.
27199         (INDIRECT_LABEL): Likewise.
27200         (indirect_thunk_name): Likewise.
27201         (output_indirect_thunk): Likewise.
27202         (output_indirect_thunk_function): Likewise.
27203         (ix86_output_indirect_branch_via_reg): Likewise.
27204         (ix86_output_indirect_branch_via_push): Likewise.
27205         (ix86_output_indirect_branch): Likewise.
27206         (ix86_output_indirect_jmp): Likewise.
27207         (ix86_code_end): Call output_indirect_thunk_function if needed.
27208         (ix86_output_call_insn): Call ix86_output_indirect_branch if
27209         needed.
27210         (ix86_handle_fndecl_attribute): Handle indirect_branch.
27211         (ix86_attribute_table): Add indirect_branch.
27212         * config/i386/i386.h (machine_function): Add indirect_branch_type
27213         and has_local_indirect_jump.
27214         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27215         to true.
27216         (tablejump): Likewise.
27217         (*indirect_jump): Use ix86_output_indirect_jmp.
27218         (*tablejump_1): Likewise.
27219         (simple_return_indirect_internal): Likewise.
27220         * config/i386/i386.opt (mindirect-branch=): New option.
27221         (indirect_branch): New.
27222         (keep): Likewise.
27223         (thunk): Likewise.
27224         (thunk-inline): Likewise.
27225         (thunk-extern): Likewise.
27226         * doc/extend.texi: Document indirect_branch function attribute.
27227         * doc/invoke.texi: Document -mindirect-branch= option.
27229 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
27231         PR ipa/83051
27232         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27234 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
27236         * ipa-inline.c (want_inline_small_function_p): Return false if
27237         inlining has already failed with CIF_FINAL_ERROR.
27238         (update_caller_keys): Call want_inline_small_function_p before
27239         can_inline_edge_p.
27240         (update_callee_keys): Likewise.
27242 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27244         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27245         New function.
27246         (rs6000_quadword_masked_address_p): Likewise.
27247         (quad_aligned_load_p): Likewise.
27248         (quad_aligned_store_p): Likewise.
27249         (const_load_sequence_p): Add comment to describe the outer-most loop.
27250         (mimic_memory_attributes_and_flags): New function.
27251         (rs6000_gen_stvx): Likewise.
27252         (replace_swapped_aligned_store): Likewise.
27253         (rs6000_gen_lvx): Likewise.
27254         (replace_swapped_aligned_load): Likewise.
27255         (replace_swapped_load_constant): Capitalize argument name in
27256         comment describing this function.
27257         (rs6000_analyze_swaps): Add a third pass to search for vector loads
27258         and stores that access quad-word aligned addresses and replace
27259         with stvx or lvx instructions when appropriate.
27260         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27261         New function prototype.
27262         (rs6000_quadword_masked_address_p): Likewise.
27263         (rs6000_gen_lvx): Likewise.
27264         (rs6000_gen_stvx): Likewise.
27265         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27266         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27267         when memory address is aligned.
27268         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27269         this split to select lvx instruction when memory address is aligned.
27270         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27271         instruction when memory address is aligned.
27272         (*vsx_le_perm_load_v16qi): Likewise.
27273         (four unnamed splitters): Modify to select the stvx instruction
27274         when memory is aligned.
27276 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
27278         * predict.c (determine_unlikely_bbs): Handle correctly BBs
27279         which appears in the queue multiple times.
27281 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27282             Alan Hayward  <alan.hayward@arm.com>
27283             David Sherwood  <david.sherwood@arm.com>
27285         * tree-vectorizer.h (vec_lower_bound): New structure.
27286         (_loop_vec_info): Add check_nonzero and lower_bounds.
27287         (LOOP_VINFO_CHECK_NONZERO): New macro.
27288         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27289         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27290         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27291         fields.  Make seg_len the distance travelled, not including the
27292         access size.
27293         (dr_direction_indicator): Declare.
27294         (dr_zero_step_indicator): Likewise.
27295         (dr_known_forward_stride_p): Likewise.
27296         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27297         tree-ssanames.h.
27298         (runtime_alias_check_p): Allow runtime alias checks with
27299         variable strides.
27300         (operator ==): Compare access_size and align.
27301         (prune_runtime_alias_test_list): Rework for new distinction between
27302         the access_size and seg_len.
27303         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
27304         segment lengths.
27305         (get_segment_min_max): New function.
27306         (create_intersect_range_checks): Use it.
27307         (dr_step_indicator): New function.
27308         (dr_direction_indicator): Likewise.
27309         (dr_zero_step_indicator): Likewise.
27310         (dr_known_forward_stride_p): Likewise.
27311         * tree-loop-distribution.c (data_ref_segment_size): Return
27312         DR_STEP * (niters - 1).
27313         (compute_alias_check_pairs): Update call to the dr_with_seg_len
27314         constructor.
27315         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
27316         (vect_preserves_scalar_order_p): New function, split out from...
27317         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
27318         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
27319         (vect_vfa_access_size): New function.
27320         (vect_vfa_align): Likewise.
27321         (vect_compile_time_alias): Take access_size_a and access_b arguments.
27322         (dump_lower_bound): New function.
27323         (vect_check_lower_bound): Likewise.
27324         (vect_small_gap_p): Likewise.
27325         (vectorizable_with_step_bound_p): Likewise.
27326         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
27327         depencies if the vectorization factor is 1.  Convert the checks
27328         for nonzero steps into checks on the bounds of DR_STEP.  Try using
27329         a bunds check for variable steps if the minimum required step is
27330         relatively small. Update calls to the dr_with_seg_len
27331         constructor and to vect_compile_time_alias.
27332         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
27333         function.
27334         (vect_loop_versioning): Call it.
27335         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
27336         when retrying.
27337         (vect_estimate_min_profitable_iters): Account for any bounds checks.
27339 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27340             Alan Hayward  <alan.hayward@arm.com>
27341             David Sherwood  <david.sherwood@arm.com>
27343         * doc/sourcebuild.texi (vect_scatter_store): Document.
27344         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
27345         optabs.
27346         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
27347         Document.
27348         * genopinit.c (main): Add supports_vec_scatter_store and
27349         supports_vec_scatter_store_cached to target_optabs.
27350         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
27351         IFN_MASK_SCATTER_STORE.
27352         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
27353         functions.
27354         * internal-fn.h (internal_store_fn_p): Declare.
27355         (internal_fn_stored_value_index): Likewise.
27356         * internal-fn.c (scatter_store_direct): New macro.
27357         (expand_scatter_store_optab_fn): New function.
27358         (direct_scatter_store_optab_supported_p): New macro.
27359         (internal_store_fn_p): New function.
27360         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
27361         IFN_MASK_SCATTER_STORE.
27362         (internal_fn_mask_index): Likewise.
27363         (internal_fn_stored_value_index): New function.
27364         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
27365         for scatter stores.
27366         * optabs-query.h (supports_vec_scatter_store_p): Declare.
27367         * optabs-query.c (supports_vec_scatter_store_p): New function.
27368         * tree-vectorizer.h (vect_get_store_rhs): Declare.
27369         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
27370         true for scatter stores.
27371         (vect_gather_scatter_fn_p): Handle scatter stores too.
27372         (vect_check_gather_scatter): Consider using scatter stores if
27373         supports_vec_scatter_store_p.
27374         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
27375         scatter stores too.
27376         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27377         internal_fn_stored_value_index.
27378         (check_load_store_masking): Handle scatter stores too.
27379         (vect_get_store_rhs): Make public.
27380         (vectorizable_call): Use internal_store_fn_p.
27381         (vectorizable_store): Handle scatter store internal functions.
27382         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
27383         when deciding whether the end of the group has been reached.
27384         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
27385         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
27386         (mask_scatter_store<mode>): New insns.
27388 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27389             Alan Hayward  <alan.hayward@arm.com>
27390             David Sherwood  <david.sherwood@arm.com>
27392         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
27393         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
27394         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
27395         function.
27396         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
27397         Use vect_truncate_gather_scatter_offset if we can't treat the
27398         operation as a normal gather load or scatter store.
27399         (get_group_load_store_type): Take the gather_scatter_info
27400         as argument.  Try using a gather load or scatter store for
27401         single-element groups.
27402         (get_load_store_type): Update calls to get_group_load_store_type
27403         and vect_use_strided_gather_scatters_p.
27405 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27406             Alan Hayward  <alan.hayward@arm.com>
27407             David Sherwood  <david.sherwood@arm.com>
27409         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
27410         optional tree argument.
27411         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
27412         null target hooks.
27413         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
27414         but continue to use the current value as a fallback.
27415         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
27416         to compare the updates.
27417         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
27418         (get_load_store_type): Use it when handling a strided access.
27419         (vect_get_strided_load_store_ops): New function.
27420         (vect_get_data_ptr_increment): Likewise.
27421         (vectorizable_load): Handle strided gather loads.  Always pass
27422         a step to vect_create_data_ref_ptr and bump_vector_ptr.
27424 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27425             Alan Hayward  <alan.hayward@arm.com>
27426             David Sherwood  <david.sherwood@arm.com>
27428         * doc/md.texi (gather_load@var{m}): Document.
27429         (mask_gather_load@var{m}): Likewise.
27430         * genopinit.c (main): Add supports_vec_gather_load and
27431         supports_vec_gather_load_cached to target_optabs.
27432         * optabs-tree.c (init_tree_optimization_optabs): Use
27433         ggc_cleared_alloc to allocate target_optabs.
27434         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
27435         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
27436         functions.
27437         * internal-fn.h (internal_load_fn_p): Declare.
27438         (internal_gather_scatter_fn_p): Likewise.
27439         (internal_fn_mask_index): Likewise.
27440         (internal_gather_scatter_fn_supported_p): Likewise.
27441         * internal-fn.c (gather_load_direct): New macro.
27442         (expand_gather_load_optab_fn): New function.
27443         (direct_gather_load_optab_supported_p): New macro.
27444         (direct_internal_fn_optab): New function.
27445         (internal_load_fn_p): Likewise.
27446         (internal_gather_scatter_fn_p): Likewise.
27447         (internal_fn_mask_index): Likewise.
27448         (internal_gather_scatter_fn_supported_p): Likewise.
27449         * optabs-query.c (supports_at_least_one_mode_p): New function.
27450         (supports_vec_gather_load_p): Likewise.
27451         * optabs-query.h (supports_vec_gather_load_p): Declare.
27452         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
27453         and memory_type field.
27454         (NUM_PATTERNS): Bump to 15.
27455         * tree-vect-data-refs.c: Include internal-fn.h.
27456         (vect_gather_scatter_fn_p): New function.
27457         (vect_describe_gather_scatter_call): Likewise.
27458         (vect_check_gather_scatter): Try using internal functions for
27459         gather loads.  Recognize existing calls to a gather load function.
27460         (vect_analyze_data_refs): Consider using gather loads if
27461         supports_vec_gather_load_p.
27462         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
27463         (vect_get_gather_scatter_offset_type): Likewise.
27464         (vect_convert_mask_for_vectype): Likewise.
27465         (vect_add_conversion_to_patterm): Likewise.
27466         (vect_try_gather_scatter_pattern): Likewise.
27467         (vect_recog_gather_scatter_pattern): New pattern recognizer.
27468         (vect_vect_recog_func_ptrs): Add it.
27469         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27470         internal_fn_mask_index and internal_gather_scatter_fn_p.
27471         (check_load_store_masking): Take the gather_scatter_info as an
27472         argument and handle gather loads.
27473         (vect_get_gather_scatter_ops): New function.
27474         (vectorizable_call): Check internal_load_fn_p.
27475         (vectorizable_load): Likewise.  Handle gather load internal
27476         functions.
27477         (vectorizable_store): Update call to check_load_store_masking.
27478         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
27479         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
27480         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
27481         (aarch64_gather_scale_operand_d): New predicates.
27482         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
27483         (mask_gather_load<mode>): New insns.
27485 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27486             Alan Hayward  <alan.hayward@arm.com>
27487             David Sherwood  <david.sherwood@arm.com>
27489         * optabs.def (fold_left_plus_optab): New optab.
27490         * doc/md.texi (fold_left_plus_@var{m}): Document.
27491         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
27492         * internal-fn.c (fold_left_direct): Define.
27493         (expand_fold_left_optab_fn): Likewise.
27494         (direct_fold_left_optab_supported_p): Likewise.
27495         * fold-const-call.c (fold_const_fold_left): New function.
27496         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
27497         * tree-parloops.c (valid_reduction_p): New function.
27498         (gather_scalar_reductions): Use it.
27499         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
27500         (vect_finish_replace_stmt): Declare.
27501         * tree-vect-loop.c (fold_left_reduction_fn): New function.
27502         (needs_fold_left_reduction_p): New function, split out from...
27503         (vect_is_simple_reduction): ...here.  Accept reductions that
27504         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
27505         (vect_force_simple_reduction): Also store the reduction type in
27506         the assignment's STMT_VINFO_REDUC_TYPE.
27507         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
27508         (merge_with_identity): New function.
27509         (vect_expand_fold_left): Likewise.
27510         (vectorize_fold_left_reduction): Likewise.
27511         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
27512         scalar phi in place for it.  Check for target support and reject
27513         cases that would reassociate the operation.  Defer the transform
27514         phase to vectorize_fold_left_reduction.
27515         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
27516         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
27517         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
27519 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27521         * tree-if-conv.c (predicate_mem_writes): Remove redundant
27522         call to ifc_temp_var.
27524 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27525             Alan Hayward  <alan.hayward@arm.com>
27526             David Sherwood  <david.sherwood@arm.com>
27528         * target.def (legitimize_address_displacement): Take the original
27529         offset as a poly_int.
27530         * targhooks.h (default_legitimize_address_displacement): Update
27531         accordingly.
27532         * targhooks.c (default_legitimize_address_displacement): Likewise.
27533         * doc/tm.texi: Regenerate.
27534         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
27535         as an argument, moving assert of ad->disp == ad->disp_term to...
27536         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
27537         Try calling targetm.legitimize_address_displacement before expanding
27538         the address rather than afterwards, and adjust for the new interface.
27539         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
27540         Match the new hook interface.  Handle SVE addresses.
27541         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
27542         new hook interface.
27544 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27546         * Makefile.in (OBJS): Add early-remat.o.
27547         * target.def (select_early_remat_modes): New hook.
27548         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
27549         * doc/tm.texi: Regenerate.
27550         * targhooks.h (default_select_early_remat_modes): Declare.
27551         * targhooks.c (default_select_early_remat_modes): New function.
27552         * timevar.def (TV_EARLY_REMAT): New timevar.
27553         * passes.def (pass_early_remat): New pass.
27554         * tree-pass.h (make_pass_early_remat): Declare.
27555         * early-remat.c: New file.
27556         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
27557         function.
27558         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
27560 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27561             Alan Hayward  <alan.hayward@arm.com>
27562             David Sherwood  <david.sherwood@arm.com>
27564         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
27565         vfm1 with a bound_epilog parameter.
27566         (vect_do_peeling): Update calls accordingly, and move the prologue
27567         call earlier in the function.  Treat the base bound_epilog as 0 for
27568         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
27569         this base when peeling for gaps.
27570         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
27571         with fully-masked loops.
27572         (vect_estimate_min_profitable_iters): Handle the single peeled
27573         iteration in that case.
27575 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27576             Alan Hayward  <alan.hayward@arm.com>
27577             David Sherwood  <david.sherwood@arm.com>
27579         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
27580         single-element interleaving even if the size is not a power of 2.
27581         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
27582         accesses for single-element interleaving if the group size is
27583         not a power of 2.
27585 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27586             Alan Hayward  <alan.hayward@arm.com>
27587             David Sherwood  <david.sherwood@arm.com>
27589         * doc/md.texi (fold_extract_last_@var{m}): Document.
27590         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
27591         * optabs.def (fold_extract_last_optab): New optab.
27592         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
27593         * internal-fn.c (fold_extract_direct): New macro.
27594         (expand_fold_extract_optab_fn): Likewise.
27595         (direct_fold_extract_optab_supported_p): Likewise.
27596         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
27597         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27598         EXTRACT_LAST_REDUCTION.
27599         (get_initial_def_for_reduction): Do not create an initial vector
27600         for EXTRACT_LAST_REDUCTION reductions.
27601         (vectorizable_reduction): Leave the scalar phi in place for
27602         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
27603         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
27604         epilogue code for EXTRACT_LAST_REDUCTION and defer the
27605         transform phase to vectorizable_condition.
27606         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
27607         split out from...
27608         (vect_finish_stmt_generation): ...here.
27609         (vect_finish_replace_stmt): New function.
27610         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
27611         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
27612         pattern.
27613         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
27615 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27616             Alan Hayward  <alan.hayward@arm.com>
27617             David Sherwood  <david.sherwood@arm.com>
27619         * doc/md.texi (extract_last_@var{m}): Document.
27620         * optabs.def (extract_last_optab): New optab.
27621         * internal-fn.def (EXTRACT_LAST): New internal function.
27622         * internal-fn.c (cond_unary_direct): New macro.
27623         (expand_cond_unary_optab_fn): Likewise.
27624         (direct_cond_unary_optab_supported_p): Likewise.
27625         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
27626         loops using EXTRACT_LAST.
27627         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
27628         (extract_last_<mode>): ...this optab.
27629         (vec_extract<mode><Vel>): Update accordingly.
27631 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27632             Alan Hayward  <alan.hayward@arm.com>
27633             David Sherwood  <david.sherwood@arm.com>
27635         * target.def (empty_mask_is_expensive): New hook.
27636         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
27637         * doc/tm.texi: Regenerate.
27638         * targhooks.h (default_empty_mask_is_expensive): Declare.
27639         * targhooks.c (default_empty_mask_is_expensive): New function.
27640         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
27641         if the target says that empty masks are expensive.
27642         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
27643         New function.
27644         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
27646 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27647             Alan Hayward  <alan.hayward@arm.com>
27648             David Sherwood  <david.sherwood@arm.com>
27650         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
27651         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
27652         (vect_use_loop_mask_for_alignment_p): New function.
27653         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
27654         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
27655         niters_skip argument.  Make sure that the first niters_skip elements
27656         of the first iteration are inactive.
27657         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
27658         Update call to vect_set_loop_masks_directly.
27659         (get_misalign_in_elems): New function, split out from...
27660         (vect_gen_prolog_loop_niters): ...here.
27661         (vect_update_init_of_dr): Take a code argument that specifies whether
27662         the adjustment should be added or subtracted.
27663         (vect_update_init_of_drs): Likewise.
27664         (vect_prepare_for_masked_peels): New function.
27665         (vect_do_peeling): Skip prologue peeling if we're using a mask
27666         instead.  Update call to vect_update_inits_of_drs.
27667         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27668         mask_skip_niters.
27669         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
27670         alignment.  Do not include the number of peeled iterations in
27671         the minimum threshold in that case.
27672         (vectorizable_induction): Adjust the start value down by
27673         LOOP_VINFO_MASK_SKIP_NITERS iterations.
27674         (vect_transform_loop): Call vect_prepare_for_masked_peels.
27675         Take the number of skipped iterations into account when calculating
27676         the loop bounds.
27677         * tree-vect-stmts.c (vect_gen_while_not): New function.
27679 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27680             Alan Hayward  <alan.hayward@arm.com>
27681             David Sherwood  <david.sherwood@arm.com>
27683         * doc/sourcebuild.texi (vect_fully_masked): Document.
27684         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
27685         default value to 0.
27686         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
27687         split out from...
27688         (vect_analyze_loop_2): ...here. Don't check the vectorization
27689         factor against the number of loop iterations if the loop is
27690         fully-masked.
27692 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27693             Alan Hayward  <alan.hayward@arm.com>
27694             David Sherwood  <david.sherwood@arm.com>
27696         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
27697         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
27698         (dump_groups): Update accordingly.
27699         (iv_use::mem_type): New member variable.
27700         (address_p): New function.
27701         (record_use): Add a mem_type argument and initialize the new
27702         mem_type field.
27703         (record_group_use): Add a mem_type argument.  Use address_p.
27704         Remove obsolete null checks of base_object.  Update call to record_use.
27705         (find_interesting_uses_op): Update call to record_group_use.
27706         (find_interesting_uses_cond): Likewise.
27707         (find_interesting_uses_address): Likewise.
27708         (get_mem_type_for_internal_fn): New function.
27709         (find_address_like_use): Likewise.
27710         (find_interesting_uses_stmt): Try find_address_like_use before
27711         calling find_interesting_uses_op.
27712         (addr_offset_valid_p): Use the iv mem_type field as the type
27713         of the addressed memory.
27714         (add_autoinc_candidates): Likewise.
27715         (get_address_cost): Likewise.
27716         (split_small_address_groups_p): Use address_p.
27717         (split_address_groups): Likewise.
27718         (add_iv_candidate_for_use): Likewise.
27719         (autoinc_possible_for_pair): Likewise.
27720         (rewrite_groups): Likewise.
27721         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
27722         (determine_group_iv_cost): Update after split of USE_ADDRESS.
27723         (get_alias_ptr_type_for_ptr_address): New function.
27724         (rewrite_use_address): Rewrite address uses in calls that were
27725         identified by find_address_like_use.
27727 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27728             Alan Hayward  <alan.hayward@arm.com>
27729             David Sherwood  <david.sherwood@arm.com>
27731         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
27732         TARGET_MEM_REFs.
27733         * gimple-expr.h (is_gimple_addressable: Likewise.
27734         * gimple-expr.c (is_gimple_address): Likewise.
27735         * internal-fn.c (expand_call_mem_ref): New function.
27736         (expand_mask_load_optab_fn): Use it.
27737         (expand_mask_store_optab_fn): Likewise.
27739 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27740             Alan Hayward  <alan.hayward@arm.com>
27741             David Sherwood  <david.sherwood@arm.com>
27743         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
27744         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
27745         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
27746         (cond_umax@var{mode}): Document.
27747         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
27748         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
27749         (cond_umin_optab, cond_umax_optab): New optabs.
27750         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
27751         (COND_IOR, COND_XOR): New internal functions.
27752         * internal-fn.h (get_conditional_internal_fn): Declare.
27753         * internal-fn.c (cond_binary_direct): New macro.
27754         (expand_cond_binary_optab_fn): Likewise.
27755         (direct_cond_binary_optab_supported_p): Likewise.
27756         (get_conditional_internal_fn): New function.
27757         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
27758         Cope with reduction statements that are vectorized as calls rather
27759         than assignments.
27760         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
27761         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
27762         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
27763         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
27764         (UNSPEC_COND_EOR): New unspecs.
27765         (optab): Add mappings for them.
27766         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
27767         (sve_int_op, sve_fp_op): New int attributes.
27769 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27770             Alan Hayward  <alan.hayward@arm.com>
27771             David Sherwood  <david.sherwood@arm.com>
27773         * optabs.def (while_ult_optab): New optab.
27774         * doc/md.texi (while_ult@var{m}@var{n}): Document.
27775         * internal-fn.def (WHILE_ULT): New internal function.
27776         * internal-fn.h (direct_internal_fn_supported_p): New override
27777         that takes two types as argument.
27778         * internal-fn.c (while_direct): New macro.
27779         (expand_while_optab_fn): New function.
27780         (convert_optab_supported_p): Likewise.
27781         (direct_while_optab_supported_p): New macro.
27782         * wide-int.h (wi::udiv_ceil): New function.
27783         * tree-vectorizer.h (rgroup_masks): New structure.
27784         (vec_loop_masks): New typedef.
27785         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
27786         and fully_masked_p.
27787         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
27788         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
27789         (vect_max_vf): New function.
27790         (slpeel_make_loop_iterate_ntimes): Delete.
27791         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
27792         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
27793         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
27794         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
27795         internal-fn.h, stor-layout.h and optabs-query.h.
27796         (vect_set_loop_mask): New function.
27797         (add_preheader_seq): Likewise.
27798         (add_header_seq): Likewise.
27799         (interleave_supported_p): Likewise.
27800         (vect_maybe_permute_loop_masks): Likewise.
27801         (vect_set_loop_masks_directly): Likewise.
27802         (vect_set_loop_condition_masked): Likewise.
27803         (vect_set_loop_condition_unmasked): New function, split out from
27804         slpeel_make_loop_iterate_ntimes.
27805         (slpeel_make_loop_iterate_ntimes): Rename to..
27806         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
27807         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
27808         (vect_do_peeling): Update call accordingly.
27809         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
27810         loops.
27811         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27812         mask_compare_type, can_fully_mask_p and fully_masked_p.
27813         (release_vec_loop_masks): New function.
27814         (_loop_vec_info): Use it to free the loop masks.
27815         (can_produce_all_loop_masks_p): New function.
27816         (vect_get_max_nscalars_per_iter): Likewise.
27817         (vect_verify_full_masking): Likewise.
27818         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
27819         retries, and free the mask rgroups before retrying.  Check loop-wide
27820         reasons for disallowing fully-masked loops.  Make the final decision
27821         about whether use a fully-masked loop or not.
27822         (vect_estimate_min_profitable_iters): Do not assume that peeling
27823         for the number of iterations will be needed for fully-masked loops.
27824         (vectorizable_reduction): Disable fully-masked loops.
27825         (vectorizable_live_operation): Likewise.
27826         (vect_halve_mask_nunits): New function.
27827         (vect_double_mask_nunits): Likewise.
27828         (vect_record_loop_mask): Likewise.
27829         (vect_get_loop_mask): Likewise.
27830         (vect_transform_loop): Handle the case in which the final loop
27831         iteration might handle a partial vector.  Call vect_set_loop_condition
27832         instead of slpeel_make_loop_iterate_ntimes.
27833         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
27834         (check_load_store_masking): New function.
27835         (prepare_load_store_mask): Likewise.
27836         (vectorizable_store): Handle fully-masked loops.
27837         (vectorizable_load): Likewise.
27838         (supportable_widening_operation): Use vect_halve_mask_nunits for
27839         booleans.
27840         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
27841         (vect_gen_while): New function.
27842         * config/aarch64/aarch64.md (umax<mode>3): New expander.
27843         (aarch64_uqdec<mode>): New insn.
27845 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27846             Alan Hayward  <alan.hayward@arm.com>
27847             David Sherwood  <david.sherwood@arm.com>
27849         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
27850         (reduc_xor_scal_optab): New optabs.
27851         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
27852         (reduc_xor_scal_@var{m}): Document.
27853         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
27854         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
27855         internal functions.
27856         * fold-const-call.c (fold_const_call): Handle them.
27857         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
27858         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
27859         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
27860         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
27861         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
27862         (UNSPEC_XORV): New unspecs.
27863         (optab): Add entries for them.
27864         (BITWISEV): New int iterator.
27865         (bit_reduc_op): New int attributes.
27867 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27868             Alan Hayward  <alan.hayward@arm.com>
27869             David Sherwood  <david.sherwood@arm.com>
27871         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
27872         * internal-fn.def (VEC_SHL_INSERT): New internal function.
27873         * optabs.def (vec_shl_insert_optab): New optab.
27874         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
27875         (duplicate_and_interleave): Likewise.
27876         * tree-vect-loop.c: Include internal-fn.h.
27877         (neutral_op_for_slp_reduction): New function, split out from
27878         get_initial_defs_for_reduction.
27879         (get_initial_def_for_reduction): Handle option 2 for variable-length
27880         vectors by loading the neutral value into a vector and then shifting
27881         the initial value into element 0.
27882         (get_initial_defs_for_reduction): Replace the code argument with
27883         the neutral value calculated by neutral_op_for_slp_reduction.
27884         Use gimple_build_vector for constant-length vectors.
27885         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
27886         but the first group_size elements have a neutral value.
27887         Use duplicate_and_interleave otherwise.
27888         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
27889         Update call to get_initial_defs_for_reduction.  Handle SLP
27890         reductions for variable-length vectors by creating one vector
27891         result for each scalar result, with the elements associated
27892         with other scalar results stubbed out with the neutral value.
27893         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
27894         Require IFN_VEC_SHL_INSERT for double reductions on
27895         variable-length vectors, or SLP reductions that have
27896         a neutral value.  Require can_duplicate_and_interleave_p
27897         support for variable-length unchained SLP reductions if there
27898         is no neutral value, such as for MIN/MAX reductions.  Also require
27899         the number of vector elements to be a multiple of the number of
27900         SLP statements when doing variable-length unchained SLP reductions.
27901         Update call to vect_create_epilog_for_reduction.
27902         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
27903         and remove initial values.
27904         (duplicate_and_interleave): Make public.
27905         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
27906         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
27908 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27909             Alan Hayward  <alan.hayward@arm.com>
27910             David Sherwood  <david.sherwood@arm.com>
27912         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
27913         (can_duplicate_and_interleave_p): New function.
27914         (vect_get_and_check_slp_defs): Take the vector of statements
27915         rather than just the current one.  Remove excess parentheses.
27916         Restriction rejectinon of vect_constant_def and vect_external_def
27917         for variable-length vectors to boolean types, or types for which
27918         can_duplicate_and_interleave_p is false.
27919         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
27920         (duplicate_and_interleave): New function.
27921         (vect_get_constant_vectors): Use gimple_build_vector for
27922         constant-length vectors and suitable variable-length constant
27923         vectors.  Use duplicate_and_interleave for other variable-length
27924         vectors.  Don't defer the update when inserting new statements.
27926 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27927             Alan Hayward  <alan.hayward@arm.com>
27928             David Sherwood  <david.sherwood@arm.com>
27930         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
27931         min_profitable_iters doesn't go negative.
27933 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27934             Alan Hayward  <alan.hayward@arm.com>
27935             David Sherwood  <david.sherwood@arm.com>
27937         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
27938         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
27939         * optabs.def (vec_mask_load_lanes_optab): New optab.
27940         (vec_mask_store_lanes_optab): Likewise.
27941         * internal-fn.def (MASK_LOAD_LANES): New internal function.
27942         (MASK_STORE_LANES): Likewise.
27943         * internal-fn.c (mask_load_lanes_direct): New macro.
27944         (mask_store_lanes_direct): Likewise.
27945         (expand_mask_load_optab_fn): Handle masked operations.
27946         (expand_mask_load_lanes_optab_fn): New macro.
27947         (expand_mask_store_optab_fn): Handle masked operations.
27948         (expand_mask_store_lanes_optab_fn): New macro.
27949         (direct_mask_load_lanes_optab_supported_p): Likewise.
27950         (direct_mask_store_lanes_optab_supported_p): Likewise.
27951         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
27952         parameter.
27953         (vect_load_lanes_supported): Likewise.
27954         * tree-vect-data-refs.c (strip_conversion): New function.
27955         (can_group_stmts_p): Likewise.
27956         (vect_analyze_data_ref_accesses): Use it instead of checking
27957         for a pair of assignments.
27958         (vect_store_lanes_supported): Take a masked_p parameter.
27959         (vect_load_lanes_supported): Likewise.
27960         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
27961         vect_store_lanes_supported and vect_load_lanes_supported.
27962         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
27963         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
27964         parameter.  Don't allow gaps for masked accesses.
27965         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
27966         and vect_load_lanes_supported.
27967         (get_load_store_type): Take a masked_p parameter and update
27968         call to get_group_load_store_type.
27969         (vectorizable_store): Update call to get_load_store_type.
27970         Handle IFN_MASK_STORE_LANES.
27971         (vectorizable_load): Update call to get_load_store_type.
27972         Handle IFN_MASK_LOAD_LANES.
27974 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27975             Alan Hayward  <alan.hayward@arm.com>
27976             David Sherwood  <david.sherwood@arm.com>
27978         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
27979         modes for SVE.
27980         * config/aarch64/aarch64-protos.h
27981         (aarch64_sve_struct_memory_operand_p): Declare.
27982         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
27983         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
27984         (VPRED, vpred): Handle SVE structure modes.
27985         * config/aarch64/constraints.md (Utx): New constraint.
27986         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
27987         (aarch64_sve_struct_nonimmediate_operand): New predicates.
27988         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
27989         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
27990         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
27991         structure modes.  Split into pieces after RA.
27992         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
27993         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
27994         New patterns.
27995         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
27996         SVE structure modes.
27997         (aarch64_classify_address): Likewise.
27998         (sizetochar): Move earlier in file.
27999         (aarch64_print_operand): Handle SVE register lists.
28000         (aarch64_array_mode): New function.
28001         (aarch64_sve_struct_memory_operand_p): Likewise.
28002         (TARGET_ARRAY_MODE): Redefine.
28004 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28005             Alan Hayward  <alan.hayward@arm.com>
28006             David Sherwood  <david.sherwood@arm.com>
28008         * target.def (array_mode): New target hook.
28009         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28010         * doc/tm.texi: Regenerate.
28011         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28012         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28013         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28014         targetm.array_mode.
28015         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28016         type sizes.
28018 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28019             Alan Hayward  <alan.hayward@arm.com>
28020             David Sherwood  <david.sherwood@arm.com>
28022         * fold-const.c (fold_binary_loc): Check the argument types
28023         rather than the result type when testing for a vector operation.
28025 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28027         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28028         * doc/tm.texi: Regenerate.
28030 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28031             Alan Hayward  <alan.hayward@arm.com>
28032             David Sherwood  <david.sherwood@arm.com>
28034         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28035         (sve): Document new AArch64 extension.
28036         * doc/md.texi (w): Extend the description of the AArch64
28037         constraint to include SVE vectors.
28038         (Upl, Upa): Document new AArch64 predicate constraints.
28039         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28040         enum.
28041         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28042         (msve-vector-bits=): New option.
28043         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28044         SVE when these are disabled.
28045         (sve): New extension.
28046         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28047         modes.  Adjust their number of units based on aarch64_sve_vg.
28048         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28049         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28050         aarch64_addr_query_type.
28051         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28052         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28053         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28054         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28055         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28056         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28057         (aarch64_simd_imm_zero_p): Delete.
28058         (aarch64_check_zero_based_sve_index_immediate): Declare.
28059         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28060         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28061         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28062         (aarch64_sve_float_mul_immediate_p): Likewise.
28063         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28064         rather than an rtx.
28065         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28066         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28067         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28068         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28069         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28070         (aarch64_regmode_natural_size): Likewise.
28071         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28072         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28073         left one place.
28074         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28075         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28076         for VG and the SVE predicate registers.
28077         (V_ALIASES): Add a "z"-prefixed alias.
28078         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28079         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28080         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28081         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28082         (REG_CLASS_NAMES): Add entries for them.
28083         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
28084         and the predicate registers.
28085         (aarch64_sve_vg): Declare.
28086         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28087         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28088         (REGMODE_NATURAL_SIZE): Define.
28089         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28090         SVE macros.
28091         * config/aarch64/aarch64.c: Include cfgrtl.h.
28092         (simd_immediate_info): Add a constructor for series vectors,
28093         and an associated step field.
28094         (aarch64_sve_vg): New variable.
28095         (aarch64_dbx_register_number): Handle VG and the predicate registers.
28096         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28097         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28098         (VEC_ANY_DATA, VEC_STRUCT): New constants.
28099         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28100         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28101         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28102         (aarch64_get_mask_mode): New functions.
28103         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28104         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28105         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
28106         predicate modes and predicate registers.  Explicitly restrict
28107         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
28108         to store a vector mode if it is recognized by
28109         aarch64_classify_vector_mode.
28110         (aarch64_regmode_natural_size): New function.
28111         (aarch64_hard_regno_caller_save_mode): Return the original mode
28112         for predicates.
28113         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28114         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28115         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28116         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28117         functions.
28118         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
28119         does not overlap dest if the function is frame-related.  Handle
28120         SVE constants.
28121         (aarch64_split_add_offset): New function.
28122         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28123         them aarch64_add_offset.
28124         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28125         and update call to aarch64_sub_sp.
28126         (aarch64_add_cfa_expression): New function.
28127         (aarch64_expand_prologue): Pass extra temporary registers to the
28128         functions above.  Handle the case in which we need to emit new
28129         DW_CFA_expressions for registers that were originally saved
28130         relative to the stack pointer, but now have to be expressed
28131         relative to the frame pointer.
28132         (aarch64_output_mi_thunk): Pass extra temporary registers to the
28133         functions above.
28134         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
28135         IP0 and IP1 values for SVE frames.
28136         (aarch64_expand_vec_series): New function.
28137         (aarch64_expand_sve_widened_duplicate): Likewise.
28138         (aarch64_expand_sve_const_vector): Likewise.
28139         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28140         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
28141         into the register, rather than emitting a SET directly.
28142         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28143         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28144         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28145         (offset_9bit_signed_scaled_p): New functions.
28146         (aarch64_replicate_bitmask_imm): New function.
28147         (aarch64_bitmask_imm): Use it.
28148         (aarch64_cannot_force_const_mem): Reject expressions involving
28149         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
28150         (aarch64_classify_index): Handle SVE indices, by requiring
28151         a plain register index with a scale that matches the element size.
28152         (aarch64_classify_address): Handle SVE addresses.  Assert that
28153         the mode of the address is VOIDmode or an integer mode.
28154         Update call to aarch64_classify_symbol.
28155         (aarch64_classify_symbolic_expression): Update call to
28156         aarch64_classify_symbol.
28157         (aarch64_const_vec_all_in_range_p): New function.
28158         (aarch64_print_vector_float_operand): Likewise.
28159         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28160         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28161         and the FP immediates 1.0 and 0.5.
28162         (aarch64_print_address_internal): Handle SVE addresses.
28163         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28164         (aarch64_regno_regclass): Handle predicate registers.
28165         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28166         data modes.
28167         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28168         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28169         (aarch64_convert_sve_vector_bits): New function.
28170         (aarch64_override_options): Use it to handle -msve-vector-bits=.
28171         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28172         rather than an rtx.
28173         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28174         Handle SVE vector and predicate modes.  Accept VL-based constants
28175         that need only one temporary register, and VL offsets that require
28176         no temporary registers.
28177         (aarch64_conditional_register_usage): Mark the predicate registers
28178         as fixed if SVE isn't available.
28179         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28180         Return true for SVE vector and predicate modes.
28181         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28182         rather than an unsigned int.  Handle SVE modes.
28183         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
28184         SVE modes.
28185         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28186         if SVE is enabled.
28187         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28188         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28189         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28190         (aarch64_sve_float_mul_immediate_p): New functions.
28191         (aarch64_sve_valid_immediate): New function.
28192         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28193         Explicitly reject structure modes.  Check for INDEX constants.
28194         Handle PTRUE and PFALSE constants.
28195         (aarch64_check_zero_based_sve_index_immediate): New function.
28196         (aarch64_simd_imm_zero_p): Delete.
28197         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28198         vector modes.  Accept constants in the range of CNT[BHWD].
28199         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28200         ask for an Advanced SIMD mode.
28201         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28202         (aarch64_simd_vector_alignment): Handle SVE predicates.
28203         (aarch64_vectorize_preferred_vector_alignment): New function.
28204         (aarch64_simd_vector_alignment_reachable): Use it instead of
28205         the vector size.
28206         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28207         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28208         functions.
28209         (MAX_VECT_LEN): Delete.
28210         (expand_vec_perm_d): Add a vec_flags field.
28211         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28212         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28213         (aarch64_evpc_ext): Don't apply a big-endian lane correction
28214         for SVE modes.
28215         (aarch64_evpc_rev): Rename to...
28216         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
28217         (aarch64_evpc_rev_global): New function.
28218         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28219         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28220         MAX_VECT_LEN.
28221         (aarch64_evpc_sve_tbl): New function.
28222         (aarch64_expand_vec_perm_const_1): Update after rename of
28223         aarch64_evpc_rev.  Handle SVE permutes too, trying
28224         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28225         than aarch64_evpc_tbl.
28226         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28227         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28228         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28229         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28230         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28231         (aarch64_expand_sve_vcond): New functions.
28232         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28233         of aarch64_vector_mode_p.
28234         (aarch64_dwarf_poly_indeterminate_value): New function.
28235         (aarch64_compute_pressure_classes): Likewise.
28236         (aarch64_can_change_mode_class): Likewise.
28237         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28238         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28239         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28240         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28241         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28242         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28243         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28244         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28245         constraints.
28246         (Dn, Dl, Dr): Accept const as well as const_vector.
28247         (Dz): Likewise.  Compare against CONST0_RTX.
28248         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28249         of "vector" where appropriate.
28250         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28251         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28252         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28253         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28254         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28255         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28256         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28257         (v_int_equiv): Extend to SVE modes.
28258         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28259         mode attributes.
28260         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28261         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28262         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28263         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28264         (SVE_COND_FP_CMP): New int iterators.
28265         (perm_hilo): Handle the new unpack unspecs.
28266         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28267         attributes.
28268         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28269         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28270         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28271         (aarch64_equality_operator, aarch64_constant_vector_operand)
28272         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28273         (aarch64_sve_nonimmediate_operand): Likewise.
28274         (aarch64_sve_general_operand): Likewise.
28275         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28276         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28277         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28278         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28279         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28280         (aarch64_sve_float_arith_immediate): Likewise.
28281         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28282         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28283         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28284         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28285         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28286         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28287         (aarch64_sve_float_arith_operand): Likewise.
28288         (aarch64_sve_float_arith_with_sub_operand): Likewise.
28289         (aarch64_sve_float_mul_operand): Likewise.
28290         (aarch64_sve_vec_perm_operand): Likewise.
28291         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28292         (aarch64_mov_operand): Accept const_poly_int and const_vector.
28293         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28294         as well as const_vector.
28295         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28296         in file.  Use CONST0_RTX and CONSTM1_RTX.
28297         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
28298         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28299         Use aarch64_simd_imm_zero.
28300         * config/aarch64/aarch64-sve.md: New file.
28301         * config/aarch64/aarch64.md: Include it.
28302         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28303         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28304         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28305         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28306         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28307         (sve): New attribute.
28308         (enabled): Disable instructions with the sve attribute unless
28309         TARGET_SVE.
28310         (movqi, movhi): Pass CONST_POLY_INT operaneds through
28311         aarch64_expand_mov_immediate.
28312         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28313         CNT[BHSD] immediates.
28314         (movti): Split CONST_POLY_INT moves into two halves.
28315         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
28316         Split additions that need a temporary here if the destination
28317         is the stack pointer.
28318         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
28319         (*add<mode>3_poly_1): New instruction.
28320         (set_clobber_cc): New expander.
28322 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28324         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
28325         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
28326         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
28327         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
28328         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
28329         Change innermode from fixed_mode_size to machine_mode.
28330         (simplify_subreg): Update call accordingly.  Handle a constant-sized
28331         subreg of a variable-length CONST_VECTOR.
28333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28334             Alan Hayward  <alan.hayward@arm.com>
28335             David Sherwood  <david.sherwood@arm.com>
28337         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
28338         (add_offset_to_base): New function, split out from...
28339         (create_mem_ref): ...here.  When handling a scale other than 1,
28340         check first whether the address is valid without the offset.
28341         Add it into the base if so, leaving the index and scale as-is.
28343 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28345         PR c++/83778
28346         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
28347         fold_for_warn before checking if arg2 is INTEGER_CST.
28349 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
28351         * config/rs6000/predicates.md (load_multiple_operation): Delete.
28352         (store_multiple_operation): Delete.
28353         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
28354         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
28355         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
28356         guarded by TARGET_STRING.
28357         (rs6000_output_load_multiple): Delete.
28358         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28359         OPTION_MASK_STRING / TARGET_STRING handling.
28360         (print_operand) <'N', 'O'>: Add comment that these are unused now.
28361         (const rs6000_opt_masks) <"string">: Change mask to 0.
28362         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
28363         (MASK_STRING): Delete.
28364         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
28365         parts.  Simplify.
28366         (load_multiple): Delete.
28367         (*ldmsi8): Delete.
28368         (*ldmsi7): Delete.
28369         (*ldmsi6): Delete.
28370         (*ldmsi5): Delete.
28371         (*ldmsi4): Delete.
28372         (*ldmsi3): Delete.
28373         (store_multiple): Delete.
28374         (*stmsi8): Delete.
28375         (*stmsi7): Delete.
28376         (*stmsi6): Delete.
28377         (*stmsi5): Delete.
28378         (*stmsi4): Delete.
28379         (*stmsi3): Delete.
28380         (movmemsi_8reg): Delete.
28381         (corresponding unnamed define_insn): Delete.
28382         (movmemsi_6reg): Delete.
28383         (corresponding unnamed define_insn): Delete.
28384         (movmemsi_4reg): Delete.
28385         (corresponding unnamed define_insn): Delete.
28386         (movmemsi_2reg): Delete.
28387         (corresponding unnamed define_insn): Delete.
28388         (movmemsi_1reg): Delete.
28389         (corresponding unnamed define_insn): Delete.
28390         * config/rs6000/rs6000.opt (mno-string): New.
28391         (mstring): Replace by deprecation warning stub.
28392         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
28394 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28396         * regrename.c (regrename_do_replace): If replacing the same
28397         reg multiple times, try to reuse last created gen_raw_REG.
28399         PR debug/81155
28400         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
28401         main to workaround a bug in GDB.
28403 2018-01-12  Tom de Vries  <tom@codesourcery.com>
28405         PR target/83737
28406         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
28408 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
28410         PR rtl-optimization/80481
28411         * ira-color.c (get_cap_member): New function.
28412         (allocnos_conflict_by_live_ranges_p): Use it.
28413         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
28414         (setup_slot_coalesced_allocno_live_ranges): Ditto.
28416 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
28418         PR target/83628
28419         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
28420         (*saddl_se_1): Ditto.
28421         (*ssubsi_1): Ditto.
28422         (*ssubl_se_1): Ditto.
28424 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28426         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
28427         rather than wi::to_widest for DR_INITs.
28428         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
28429         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
28430         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
28431         INTEGER_CSTs.
28432         (vect_analyze_group_access_1): Note that here.
28434 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28436         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
28437         polynomial type sizes.
28439 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28441         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
28442         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
28443         (gimple_add_tmp_var): Likewise.
28445 2018-01-12  Martin Liska  <mliska@suse.cz>
28447         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
28448         (gimple_alloc_sizes): Likewise.
28449         (dump_gimple_statistics): Use PRIu64 in printf format.
28450         * gimple.h: Change uint64_t to int.
28452 2018-01-12  Martin Liska  <mliska@suse.cz>
28454         * tree-core.h: Use uint64_t instead of int.
28455         * tree.c (tree_node_counts): Likewise.
28456         (tree_node_sizes): Likewise.
28457         (dump_tree_statistics): Use PRIu64 in printf format.
28459 2018-01-12  Martin Liska  <mliska@suse.cz>
28461         * Makefile.in: As qsort_chk is implemented in vec.c, add
28462         vec.o to linkage of gencfn-macros.
28463         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
28464         passing the info to record_node_allocation_statistics.
28465         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
28466         and pass the info.
28467         * ggc-common.c (struct ggc_usage): Add operator== and use
28468         it in operator< and compare function.
28469         * mem-stats.h (struct mem_usage): Likewise.
28470         * vec.c (struct vec_usage): Remove operator< and compare
28471         function. Can be simply inherited.
28473 2018-01-12  Martin Jambor  <mjambor@suse.cz>
28475         PR target/81616
28476         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
28477         * tree-ssa-math-opts.c: Include domwalk.h.
28478         (convert_mult_to_fma_1): New function.
28479         (fma_transformation_info): New type.
28480         (fma_deferring_state): Likewise.
28481         (cancel_fma_deferring): New function.
28482         (result_of_phi): Likewise.
28483         (last_fma_candidate_feeds_initial_phi): Likewise.
28484         (convert_mult_to_fma): Added deferring logic, split actual
28485         transformation to convert_mult_to_fma_1.
28486         (math_opts_dom_walker): New type.
28487         (math_opts_dom_walker::after_dom_children): New method, body moved
28488         here from pass_optimize_widening_mul::execute, added deferring logic
28489         bits.
28490         (pass_optimize_widening_mul::execute): Moved most of code to
28491         math_opts_dom_walker::after_dom_children.
28492         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
28493         * config/i386/i386.c (ix86_option_override_internal): Added
28494         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
28496 2018-01-12  Richard Biener  <rguenther@suse.de>
28498         PR debug/83157
28499         * dwarf2out.c (gen_variable_die): Do not reset old_die for
28500         inline instance vars.
28502 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
28504         PR target/81819
28505         * config/rx/rx.c (rx_is_restricted_memory_address):
28506         Handle SUBREG case.
28508 2018-01-12  Richard Biener  <rguenther@suse.de>
28510         PR tree-optimization/80846
28511         * target.def (split_reduction): New target hook.
28512         * targhooks.c (default_split_reduction): New function.
28513         * targhooks.h (default_split_reduction): Declare.
28514         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
28515         target requests first reduce vectors by combining low and high
28516         parts.
28517         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
28518         (get_vectype_for_scalar_type_and_size): Export.
28519         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
28520         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
28521         * doc/tm.texi: Regenerate.
28522         * config/i386/i386.c (ix86_split_reduction): Implement
28523         TARGET_VECTORIZE_SPLIT_REDUCTION.
28525 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28527         PR target/83368
28528         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
28529         in PIC mode except for TARGET_VXWORKS_RTP.
28530         * config/sparc/sparc.c: Include cfgrtl.h.
28531         (TARGET_INIT_PIC_REG): Define.
28532         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
28533         (sparc_pic_register_p): New predicate.
28534         (sparc_legitimate_address_p): Use it.
28535         (sparc_legitimize_pic_address): Likewise.
28536         (sparc_delegitimize_address): Likewise.
28537         (sparc_mode_dependent_address_p): Likewise.
28538         (gen_load_pcrel_sym): Remove 4th parameter.
28539         (load_got_register): Adjust call to above.  Remove obsolete stuff.
28540         (sparc_expand_prologue): Do not call load_got_register here.
28541         (sparc_flat_expand_prologue): Likewise.
28542         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
28543         (sparc_use_pseudo_pic_reg): New function.
28544         (sparc_init_pic_reg): Likewise.
28545         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
28546         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
28548 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
28550         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28551         Add item for branch_cost.
28553 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28555         PR rtl-optimization/83565
28556         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
28557         not extend the result to a larger mode for rotate operations.
28558         (num_sign_bit_copies1): Likewise.
28560 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28562         PR target/40411
28563         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
28564         -symbolic.
28565         Use values-Xc.o for -pedantic.
28566         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
28568 2018-01-12  Martin Liska  <mliska@suse.cz>
28570         PR ipa/83054
28571         * ipa-devirt.c (final_warning_record::grow_type_warnings):
28572         New function.
28573         (possible_polymorphic_call_targets): Use it.
28574         (ipa_devirt): Likewise.
28576 2018-01-12  Martin Liska  <mliska@suse.cz>
28578         * profile-count.h (enum profile_quality): Use 0 as invalid
28579         enum value of profile_quality.
28581 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
28583         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
28584         -mext-string options.
28586 2018-01-12  Richard Biener  <rguenther@suse.de>
28588         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
28589         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
28590         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
28591         Likewise.
28592         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
28594 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
28596         * configure.ac (--with-long-double-format): Add support for the
28597         configuration option to change the default long double format on
28598         PowerPC systems.
28599         * config.gcc (powerpc*-linux*-*): Likewise.
28600         * configure: Regenerate.
28601         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
28602         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
28603         used without modification.
28605 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28607         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
28608         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
28609         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
28610         MISC_BUILTIN_SPEC_BARRIER.
28611         (rs6000_init_builtins): Likewise.
28612         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
28613         enum value.
28614         (speculation_barrier): New define_insn.
28615         * doc/extend.texi: Document __builtin_speculation_barrier.
28617 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28619         PR target/83203
28620         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
28621         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
28622         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
28623         iterators.
28624         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
28625         integral modes instead of "ss" and "sd".
28626         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
28627         vectors with 32-bit and 64-bit elements.
28628         (vecdupssescalarmodesuffix): New mode attribute.
28629         (vec_dup<mode>): Use it.
28631 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
28633         PR target/83330
28634         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
28635         frame if argument is passed on stack.
28637 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28639         PR target/82682
28640         * ree.c (combine_reaching_defs): Optimize also
28641         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
28642         reg2=any_extend(exp); reg1=reg2;, formatting fix.
28644 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28646         PR middle-end/83189
28647         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
28649 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28651         PR middle-end/83718
28652         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
28653         after they are computed.
28655 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
28657         PR tree-optimization/83695
28658         * gimple-loop-linterchange.cc
28659         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
28660         reset cached scev information after interchange.
28661         (pass_linterchange::execute): Remove call to scev_reset_htab.
28663 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28665         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
28666         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
28667         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
28668         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
28669         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
28670         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
28671         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
28672         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
28673         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
28674         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
28675         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
28676         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
28677         (V_lane_reg): Likewise.
28678         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
28679         New define_expand.
28680         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
28681         (vfmal_lane_low<mode>_intrinsic,
28682         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
28683         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
28684         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
28685         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
28686         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
28687         vfmsl_lane_high<mode>_intrinsic): New define_insns.
28689 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28691         * config/arm/arm-cpus.in (fp16fml): New feature.
28692         (ALL_SIMD): Add fp16fml.
28693         (armv8.2-a): Add fp16fml as an option.
28694         (armv8.3-a): Likewise.
28695         (armv8.4-a): Add fp16fml as part of fp16.
28696         * config/arm/arm.h (TARGET_FP16FML): Define.
28697         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
28698         when appropriate.
28699         * config/arm/arm-modes.def (V2HF): Define.
28700         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
28701         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
28702         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
28703         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
28704         vfmsl_low, vfmsl_high): New set of builtins.
28705         * config/arm/iterators.md (PLUSMINUS): New code iterator.
28706         (vfml_op): New code attribute.
28707         (VFMLHALVES): New int iterator.
28708         (VFML, VFMLSEL): New mode attributes.
28709         (V_reg): Define mapping for V2HF.
28710         (V_hi, V_lo): New mode attributes.
28711         (VF_constraint): Likewise.
28712         (vfml_half, vfml_half_selector): New int attributes.
28713         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
28714         define_expand.
28715         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
28716         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
28717         New define_insn.
28718         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
28719         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
28720         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
28721         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
28722         documentation.
28723         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
28724         Document new effective target and option set.
28726 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28728         * config/arm/arm-cpus.in (armv8_4): New feature.
28729         (ARMv8_4a): New fgroup.
28730         (armv8.4-a): New arch.
28731         * config/arm/arm-tables.opt: Regenerate.
28732         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
28733         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
28734         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
28735         Add matching rules for -march=armv8.4-a and extensions.
28736         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
28738 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
28740         PR target/81821
28741         * config/rx/rx.md (BW): New mode attribute.
28742         (sync_lock_test_and_setsi): Add mode suffix to insn output.
28744 2018-01-11  Richard Biener  <rguenther@suse.de>
28746         PR tree-optimization/83435
28747         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
28748         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
28749         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
28751 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28752             Alan Hayward  <alan.hayward@arm.com>
28753             David Sherwood  <david.sherwood@arm.com>
28755         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
28756         field.
28757         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
28758         (aarch64_print_address_internal): Use it to check for a zero offset.
28760 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28761             Alan Hayward  <alan.hayward@arm.com>
28762             David Sherwood  <david.sherwood@arm.com>
28764         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
28765         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
28766         Return a poly_int64 rather than a HOST_WIDE_INT.
28767         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
28768         rather than a HOST_WIDE_INT.
28769         * config/aarch64/aarch64.h (aarch64_frame): Protect with
28770         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
28771         hard_fp_offset, frame_size, initial_adjust, callee_offset and
28772         final_offset from HOST_WIDE_INT to poly_int64.
28773         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
28774         to_constant when getting the number of units in an Advanced SIMD
28775         mode.
28776         (aarch64_builtin_vectorized_function): Check for a constant number
28777         of units.
28778         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
28779         GET_MODE_SIZE.
28780         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
28781         attribute instead of GET_MODE_NUNITS.
28782         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
28783         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
28784         GET_MODE_SIZE for fixed-size registers.
28785         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
28786         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
28787         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
28788         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
28789         (aarch64_print_operand, aarch64_print_address_internal)
28790         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
28791         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
28792         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
28793         Handle polynomial GET_MODE_SIZE.
28794         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
28795         wider than SImode without modification.
28796         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
28797         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
28798         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
28799         passing and returning SVE modes.
28800         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
28801         rather than GEN_INT.
28802         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
28803         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
28804         (aarch64_allocate_and_probe_stack_space): Likewise.
28805         (aarch64_layout_frame): Cope with polynomial offsets.
28806         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
28807         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
28808         polynomial offsets.
28809         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
28810         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
28811         poly_int64 rather than a HOST_WIDE_INT.
28812         (aarch64_get_separate_components, aarch64_process_components)
28813         (aarch64_expand_prologue, aarch64_expand_epilogue)
28814         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
28815         (aarch64_anchor_offset): New function, split out from...
28816         (aarch64_legitimize_address): ...here.
28817         (aarch64_builtin_vectorization_cost): Handle polynomial
28818         TYPE_VECTOR_SUBPARTS.
28819         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
28820         GET_MODE_NUNITS.
28821         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
28822         number of elements from the PARALLEL rather than the mode.
28823         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
28824         rather than GET_MODE_BITSIZE.
28825         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
28826         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
28827         (aarch64_expand_vec_perm_const_1): Handle polynomial
28828         d->perm.length () and d->perm elements.
28829         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
28830         Apply to_constant to d->perm elements.
28831         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
28832         polynomial CONST_VECTOR_NUNITS.
28833         (aarch64_move_pointer): Take amount as a poly_int64 rather
28834         than an int.
28835         (aarch64_progress_pointer): Avoid temporary variable.
28836         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
28837         the mode attribute instead of GET_MODE.
28839 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28840             Alan Hayward  <alan.hayward@arm.com>
28841             David Sherwood  <david.sherwood@arm.com>
28843         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
28844         x exists before using it.
28845         (aarch64_add_constant_internal): Rename to...
28846         (aarch64_add_offset_1): ...this.  Replace regnum with separate
28847         src and dest rtxes.  Handle the case in which they're different,
28848         including when the offset is zero.  Replace scratchreg with an rtx.
28849         Use 2 additions if there is no spare register into which we can
28850         move a 16-bit constant.
28851         (aarch64_add_constant): Delete.
28852         (aarch64_add_offset): Replace reg with separate src and dest
28853         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
28854         Use aarch64_add_offset_1.
28855         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
28856         an rtx rather than an int.  Take the delta as a poly_int64
28857         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
28858         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
28859         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
28860         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
28861         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
28862         and aarch64_add_sp.
28863         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
28864         aarch64_add_constant.
28866 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28868         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
28869         Use scalar_float_mode.
28871 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28873         * config/aarch64/aarch64-simd.md
28874         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
28875         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
28876         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
28877         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
28878         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
28879         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
28880         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
28881         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
28882         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
28883         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
28885 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28887         PR target/83514
28888         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
28889         targ_options->x_arm_arch_string is non NULL.
28891 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
28893         * config/aarch64/aarch64.h
28894         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
28896 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
28898         PR target/82096
28899         * expmed.c (emit_store_flag_force): Swap if const op0
28900         and change VOIDmode to mode of op0.
28902 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28904         PR rtl-optimization/83761
28905         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
28906         than bytes to mode_for_size.
28908 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28910         PR middle-end/83189
28911         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
28912         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
28913         profile.
28915 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28917         PR middle-end/83575
28918         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
28919         when in layout mode.
28920         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
28921         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
28922         partition fixup.
28924 2018-01-10  Michael Collison  <michael.collison@arm.com>
28926         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
28927         * config/aarch64/aarch64-option-extension.def: Add
28928         AARCH64_OPT_EXTENSION of 'fp16fml'.
28929         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28930         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
28931         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
28932         * config/aarch64/constraints.md (Ui7): New constraint.
28933         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
28934         (VFMLA_SEL_W): Ditto.
28935         (f16quad): Ditto.
28936         (f16mac1): Ditto.
28937         (VFMLA16_LOW): New int iterator.
28938         (VFMLA16_HIGH): Ditto.
28939         (UNSPEC_FMLAL): New unspec.
28940         (UNSPEC_FMLSL): Ditto.
28941         (UNSPEC_FMLAL2): Ditto.
28942         (UNSPEC_FMLSL2): Ditto.
28943         (f16mac): New code attribute.
28944         * config/aarch64/aarch64-simd-builtins.def
28945         (aarch64_fmlal_lowv2sf): Ditto.
28946         (aarch64_fmlsl_lowv2sf): Ditto.
28947         (aarch64_fmlalq_lowv4sf): Ditto.
28948         (aarch64_fmlslq_lowv4sf): Ditto.
28949         (aarch64_fmlal_highv2sf): Ditto.
28950         (aarch64_fmlsl_highv2sf): Ditto.
28951         (aarch64_fmlalq_highv4sf): Ditto.
28952         (aarch64_fmlslq_highv4sf): Ditto.
28953         (aarch64_fmlal_lane_lowv2sf): Ditto.
28954         (aarch64_fmlsl_lane_lowv2sf): Ditto.
28955         (aarch64_fmlal_laneq_lowv2sf): Ditto.
28956         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
28957         (aarch64_fmlalq_lane_lowv4sf): Ditto.
28958         (aarch64_fmlsl_lane_lowv4sf): Ditto.
28959         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
28960         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
28961         (aarch64_fmlal_lane_highv2sf): Ditto.
28962         (aarch64_fmlsl_lane_highv2sf): Ditto.
28963         (aarch64_fmlal_laneq_highv2sf): Ditto.
28964         (aarch64_fmlsl_laneq_highv2sf): Ditto.
28965         (aarch64_fmlalq_lane_highv4sf): Ditto.
28966         (aarch64_fmlsl_lane_highv4sf): Ditto.
28967         (aarch64_fmlalq_laneq_highv4sf): Ditto.
28968         (aarch64_fmlsl_laneq_highv4sf): Ditto.
28969         * config/aarch64/aarch64-simd.md:
28970         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
28971         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28972         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
28973         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28974         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
28975         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
28976         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
28977         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
28978         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
28979         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
28980         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
28981         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
28982         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
28983         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
28984         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
28985         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
28986         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
28987         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
28988         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
28989         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
28990         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
28991         (vfmlsl_low_u32): Ditto.
28992         (vfmlalq_low_u32): Ditto.
28993         (vfmlslq_low_u32): Ditto.
28994         (vfmlal_high_u32): Ditto.
28995         (vfmlsl_high_u32): Ditto.
28996         (vfmlalq_high_u32): Ditto.
28997         (vfmlslq_high_u32): Ditto.
28998         (vfmlal_lane_low_u32): Ditto.
28999         (vfmlsl_lane_low_u32): Ditto.
29000         (vfmlal_laneq_low_u32): Ditto.
29001         (vfmlsl_laneq_low_u32): Ditto.
29002         (vfmlalq_lane_low_u32): Ditto.
29003         (vfmlslq_lane_low_u32): Ditto.
29004         (vfmlalq_laneq_low_u32): Ditto.
29005         (vfmlslq_laneq_low_u32): Ditto.
29006         (vfmlal_lane_high_u32): Ditto.
29007         (vfmlsl_lane_high_u32): Ditto.
29008         (vfmlal_laneq_high_u32): Ditto.
29009         (vfmlsl_laneq_high_u32): Ditto.
29010         (vfmlalq_lane_high_u32): Ditto.
29011         (vfmlslq_lane_high_u32): Ditto.
29012         (vfmlalq_laneq_high_u32): Ditto.
29013         (vfmlslq_laneq_high_u32): Ditto.
29014         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29015         (AARCH64_FL_FOR_ARCH8_4): New.
29016         (AARCH64_ISA_F16FML): New ISA flag.
29017         (TARGET_F16FML): New feature flag for fp16fml.
29018         (doc/invoke.texi): Document new fp16fml option.
29020 2018-01-10  Michael Collison  <michael.collison@arm.com>
29022         * config/aarch64/aarch64-builtins.c:
29023         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29024         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29025         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29026         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29027         (AARCH64_ISA_SHA3): New ISA flag.
29028         (TARGET_SHA3): New feature flag for sha3.
29029         * config/aarch64/iterators.md (sha512_op): New int attribute.
29030         (CRYPTO_SHA512): New int iterator.
29031         (UNSPEC_SHA512H): New unspec.
29032         (UNSPEC_SHA512H2): Ditto.
29033         (UNSPEC_SHA512SU0): Ditto.
29034         (UNSPEC_SHA512SU1): Ditto.
29035         * config/aarch64/aarch64-simd-builtins.def
29036         (aarch64_crypto_sha512hqv2di): New builtin.
29037         (aarch64_crypto_sha512h2qv2di): Ditto.
29038         (aarch64_crypto_sha512su0qv2di): Ditto.
29039         (aarch64_crypto_sha512su1qv2di): Ditto.
29040         (aarch64_eor3qv8hi): Ditto.
29041         (aarch64_rax1qv2di): Ditto.
29042         (aarch64_xarqv2di): Ditto.
29043         (aarch64_bcaxqv8hi): Ditto.
29044         * config/aarch64/aarch64-simd.md:
29045         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29046         (aarch64_crypto_sha512su0qv2di): Ditto.
29047         (aarch64_crypto_sha512su1qv2di): Ditto.
29048         (aarch64_eor3qv8hi): Ditto.
29049         (aarch64_rax1qv2di): Ditto.
29050         (aarch64_xarqv2di): Ditto.
29051         (aarch64_bcaxqv8hi): Ditto.
29052         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29053         (vsha512h2q_u64): Ditto.
29054         (vsha512su0q_u64): Ditto.
29055         (vsha512su1q_u64): Ditto.
29056         (veor3q_u16): Ditto.
29057         (vrax1q_u64): Ditto.
29058         (vxarq_u64): Ditto.
29059         (vbcaxq_u16): Ditto.
29060         * config/arm/types.md (crypto_sha512): New type attribute.
29061         (crypto_sha3): Ditto.
29062         (doc/invoke.texi): Document new sha3 option.
29064 2018-01-10  Michael Collison  <michael.collison@arm.com>
29066         * config/aarch64/aarch64-builtins.c:
29067         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29068         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29069         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29070         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29071         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29072         (AARCH64_ISA_SM4): New ISA flag.
29073         (TARGET_SM4): New feature flag for sm4.
29074         * config/aarch64/aarch64-simd-builtins.def
29075         (aarch64_sm3ss1qv4si): Ditto.
29076         (aarch64_sm3tt1aq4si): Ditto.
29077         (aarch64_sm3tt1bq4si): Ditto.
29078         (aarch64_sm3tt2aq4si): Ditto.
29079         (aarch64_sm3tt2bq4si): Ditto.
29080         (aarch64_sm3partw1qv4si): Ditto.
29081         (aarch64_sm3partw2qv4si): Ditto.
29082         (aarch64_sm4eqv4si): Ditto.
29083         (aarch64_sm4ekeyqv4si): Ditto.
29084         * config/aarch64/aarch64-simd.md:
29085         (aarch64_sm3ss1qv4si): Ditto.
29086         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29087         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29088         (aarch64_sm4eqv4si): Ditto.
29089         (aarch64_sm4ekeyqv4si): Ditto.
29090         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29091         (sm3part_op): Ditto.
29092         (CRYPTO_SM3TT): Ditto.
29093         (CRYPTO_SM3PART): Ditto.
29094         (UNSPEC_SM3SS1): New unspec.
29095         (UNSPEC_SM3TT1A): Ditto.
29096         (UNSPEC_SM3TT1B): Ditto.
29097         (UNSPEC_SM3TT2A): Ditto.
29098         (UNSPEC_SM3TT2B): Ditto.
29099         (UNSPEC_SM3PARTW1): Ditto.
29100         (UNSPEC_SM3PARTW2): Ditto.
29101         (UNSPEC_SM4E): Ditto.
29102         (UNSPEC_SM4EKEY): Ditto.
29103         * config/aarch64/constraints.md (Ui2): New constraint.
29104         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29105         * config/arm/types.md (crypto_sm3): New type attribute.
29106         (crypto_sm4): Ditto.
29107         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29108         (vsm3tt1aq_u32): Ditto.
29109         (vsm3tt1bq_u32): Ditto.
29110         (vsm3tt2aq_u32): Ditto.
29111         (vsm3tt2bq_u32): Ditto.
29112         (vsm3partw1q_u32): Ditto.
29113         (vsm3partw2q_u32): Ditto.
29114         (vsm4eq_u32): Ditto.
29115         (vsm4ekeyq_u32): Ditto.
29116         (doc/invoke.texi): Document new sm4 option.
29118 2018-01-10  Michael Collison  <michael.collison@arm.com>
29120         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29121         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29122         (AARCH64_FL_FOR_ARCH8_4): New.
29123         (AARCH64_FL_V8_4): New flag.
29124         (doc/invoke.texi): Document new armv8.4-a option.
29126 2018-01-10  Michael Collison  <michael.collison@arm.com>
29128         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29129         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29130         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29131         * config/aarch64/aarch64-option-extension.def: Add
29132         AARCH64_OPT_EXTENSION of 'sha2'.
29133         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29134         (crypto): Disable sha2 and aes if crypto disabled.
29135         (crypto): Enable aes and sha2 if enabled.
29136         (simd): Disable sha2 and aes if simd disabled.
29137         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29138         New flags.
29139         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29140         (TARGET_SHA2): New feature flag for sha2.
29141         (TARGET_AES): New feature flag for aes.
29142         * config/aarch64/aarch64-simd.md:
29143         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29144         conditional on TARGET_AES.
29145         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29146         (aarch64_crypto_sha1hsi): Make pattern conditional
29147         on TARGET_SHA2.
29148         (aarch64_crypto_sha1hv4si): Ditto.
29149         (aarch64_be_crypto_sha1hv4si): Ditto.
29150         (aarch64_crypto_sha1su1v4si): Ditto.
29151         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29152         (aarch64_crypto_sha1su0v4si): Ditto.
29153         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29154         (aarch64_crypto_sha256su0v4si): Ditto.
29155         (aarch64_crypto_sha256su1v4si): Ditto.
29156         (doc/invoke.texi): Document new aes and sha2 options.
29158 2018-01-10  Martin Sebor  <msebor@redhat.com>
29160         PR tree-optimization/83781
29161         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29162         as string arrays.
29164 2018-01-11  Martin Sebor  <msebor@gmail.com>
29165             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29167         PR tree-optimization/83501
29168         PR tree-optimization/81703
29170         * tree-ssa-strlen.c (get_string_cst): Rename...
29171         (get_string_len): ...to this.  Handle global constants.
29172         (handle_char_store): Adjust.
29174 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
29175             Jim Wilson  <jimw@sifive.com>
29177         * config/riscv/riscv-protos.h (riscv_output_return): New.
29178         * config/riscv/riscv.c (struct machine_function): New naked_p field.
29179         (riscv_attribute_table, riscv_output_return),
29180         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29181         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29182         (riscv_compute_frame_info): Only compute frame->mask if not a naked
29183         function.
29184         (riscv_expand_prologue): Add early return for naked function.
29185         (riscv_expand_epilogue): Likewise.
29186         (riscv_function_ok_for_sibcall): Return false for naked function.
29187         (riscv_set_current_function): New.
29188         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29189         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29190         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29191         * doc/extend.texi (RISC-V Function Attributes): New.
29193 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
29195         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29196         check for 128-bit long double before checking TCmode.
29197         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29198         128-bit long doubles before checking TFmode or TCmode.
29199         (FLOAT128_IBM_P): Likewise.
29201 2018-01-10  Martin Sebor  <msebor@redhat.com>
29203         PR tree-optimization/83671
29204         * builtins.c (c_strlen): Unconditionally return zero for the empty
29205         string.
29206         Use -Warray-bounds for warnings.
29207         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29208         for non-constant array indices with COMPONENT_REF, arrays of
29209         arrays, and pointers to arrays.
29210         (gimple_fold_builtin_strlen): Determine and set length range for
29211         non-constant character arrays.
29213 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
29215         PR middle-end/81897
29216         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29217         empty blocks.
29219 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
29221         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29223 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29225         PR target/83399
29226         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29227         VECTOR_MEM_ALTIVEC_OR_VSX_P.
29228         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29229         indexed_or_indirect_operand predicate.
29230         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29231         (*vsx_le_perm_load_v8hi): Likewise.
29232         (*vsx_le_perm_load_v16qi): Likewise.
29233         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29234         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29235         (*vsx_le_perm_store_v8hi): Likewise.
29236         (*vsx_le_perm_store_v16qi): Likewise.
29237         (eight unnamed splitters): Likewise.
29239 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29241         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29242         * config/rs6000/emmintrin.h: Likewise.
29243         * config/rs6000/mmintrin.h: Likewise.
29244         * config/rs6000/xmmintrin.h: Likewise.
29246 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
29248         PR c++/43486
29249         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29250         "public_flag".
29251         * tree.c (tree_nop_conversion): Return true for location wrapper
29252         nodes.
29253         (maybe_wrap_with_location): New function.
29254         (selftest::check_strip_nops): New function.
29255         (selftest::test_location_wrappers): New function.
29256         (selftest::tree_c_tests): Call it.
29257         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29258         (maybe_wrap_with_location): New decl.
29259         (EXPR_LOCATION_WRAPPER_P): New macro.
29260         (location_wrapper_p): New inline function.
29261         (tree_strip_any_location_wrapper): New inline function.
29263 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
29265         PR target/83735
29266         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29267         stack_realign_offset for the largest alignment of stack slot
29268         actually used.
29269         (ix86_find_max_used_stack_alignment): New function.
29270         (ix86_finalize_stack_frame_flags): Use it.  Set
29271         max_used_stack_alignment if we don't realign stack.
29272         * config/i386/i386.h (machine_function): Add
29273         max_used_stack_alignment.
29275 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
29277         * config/arm/arm.opt (-mbranch-cost): New option.
29278         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29279         account.
29281 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
29283         PR target/83629
29284         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29285         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29287 2018-01-10  Richard Biener  <rguenther@suse.de>
29289         PR debug/83765
29290         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29291         early out so it also covers the case where we have a non-NULL
29292         origin.
29294 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29296         PR tree-optimization/83753
29297         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29298         for non-strided grouped accesses if the number of elements is 1.
29300 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29302         PR target/81616
29303         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29304         * i386.h (TARGET_USE_GATHER): Define.
29305         * x86-tune.def (X86_TUNE_USE_GATHER): New.
29307 2018-01-10  Martin Liska  <mliska@suse.cz>
29309         PR bootstrap/82831
29310         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29311         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29312         partitioning.
29313         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
29314         CLEANUP_NO_PARTITIONING is not set.
29316 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29318         * doc/rtl.texi: Remove documentation of (const ...) wrappers
29319         for vectors, as a partial revert of r254296.
29320         * rtl.h (const_vec_p): Delete.
29321         (const_vec_duplicate_p): Don't test for vector CONSTs.
29322         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
29323         * expmed.c (make_tree): Likewise.
29325         Revert:
29326         * common.md (E, F): Use CONSTANT_P instead of checking for
29327         CONST_VECTOR.
29328         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
29329         checking for CONST_VECTOR.
29331 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29333         PR middle-end/83575
29334         * predict.c (force_edge_cold): Handle in more sane way edges
29335         with no prediction.
29337 2018-01-09  Carl Love  <cel@us.ibm.com>
29339         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
29340         V4SI, V4SF types.
29341         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
29342         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
29343         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
29344         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
29345         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
29346         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
29347         * config/rs6000/rs6000-protos.h: Add extern defition for
29348         rs6000_generate_float2_double_code.
29349         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
29350         function.
29351         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
29352         (float2_v2df): Add define_expand.
29354 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
29356         PR target/83628
29357         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
29358         op_mode in the force_to_mode call.
29360 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29362         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
29363         instead of checking each element individually.
29364         (aarch64_evpc_uzp): Likewise.
29365         (aarch64_evpc_zip): Likewise.
29366         (aarch64_evpc_ext): Likewise.
29367         (aarch64_evpc_rev): Likewise.
29368         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
29369         instead of checking each element individually.  Return true without
29370         generating rtl if
29371         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
29372         whether all selected elements come from the same input, instead of
29373         checking each element individually.  Remove calls to gen_rtx_REG,
29374         start_sequence and end_sequence and instead assert that no rtl is
29375         generated.
29377 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29379         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
29380         order of HIGH and CONST checks.
29382 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29384         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
29385         if the destination isn't an SSA_NAME.
29387 2018-01-09  Richard Biener  <rguenther@suse.de>
29389         PR tree-optimization/83668
29390         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
29391         move prologue...
29392         (canonicalize_loop_form): ... here, renamed from ...
29393         (canonicalize_loop_closed_ssa_form): ... this and amended to
29394         swap successor edges for loop exit blocks to make us use
29395         the RPO order we need for initial schedule generation.
29397 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
29399         PR tree-optimization/64811
29400         * match.pd: When optimizing comparisons with Inf, avoid
29401         introducing or losing exceptions from comparisons with NaN.
29403 2018-01-09  Martin Liska  <mliska@suse.cz>
29405         PR sanitizer/82517
29406         * asan.c (shadow_mem_size): Add gcc_assert.
29408 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
29410         Don't save registers in main().
29412         PR target/83738
29413         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
29414         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
29415         * config/avr/avr.c (avr_set_current_function): Don't error if
29416         naked, OS_task or OS_main are specified at the same time.
29417         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
29418         OS_main.
29419         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
29420         attribute.
29421         * common/config/avr/avr-common.c (avr_option_optimization_table):
29422         Switch on -mmain-is-OS_task for optimizing compilations.
29424 2018-01-09  Richard Biener  <rguenther@suse.de>
29426         PR tree-optimization/83572
29427         * graphite.c: Include cfganal.h.
29428         (graphite_transform_loops): Connect infinite loops to exit
29429         and remove fake edges at the end.
29431 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29433         * ipa-inline.c (edge_badness): Revert accidental checkin.
29435 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29437         PR ipa/80763
29438         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
29439         symbols; not inline clones.
29441 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
29443         PR target/83507
29444         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
29445         hard registers.  Formatting fixes.
29447         PR preprocessor/83722
29448         * gcc.c (try_generate_repro): Pass
29449         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
29450         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
29451         do_report_bug.
29453 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
29454             Kito Cheng  <kito.cheng@gmail.com>
29456         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
29457         (riscv_leaf_function_p): Delete.
29458         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
29460 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29462         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
29463         function.
29464         (do_ifelse): New function.
29465         (do_isel): New function.
29466         (do_sub3): New function.
29467         (do_add3): New function.
29468         (do_load_mask_compare): New function.
29469         (do_overlap_load_compare): New function.
29470         (expand_compare_loop): New function.
29471         (expand_block_compare): Call expand_compare_loop() when appropriate.
29472         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
29473         option description.
29474         (-mblock-compare-inline-loop-limit): New option.
29476 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29478         PR target/83677
29479         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
29480         Reverse order of second and third operands in first alternative.
29481         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
29482         of first and second elements in UNSPEC_VPERMR vector.
29483         (altivec_expand_vec_perm_le): Likewise.
29485 2018-01-08  Jeff Law  <law@redhat.com>
29487         PR rtl-optimizatin/81308
29488         * tree-switch-conversion.c (cfg_altered): New file scoped static.
29489         (process_switch): If group_case_labels makes a change, then set
29490         cfg_altered.
29491         (pass_convert_switch::execute): If a switch is converted, then
29492         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
29494         PR rtl-optimization/81308
29495         * recog.c (split_all_insns): Conditionally cleanup the CFG after
29496         splitting insns.
29498 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
29500         PR target/83663 - Revert r255946
29501         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
29502         generation for cases where splatting a value is not useful.
29503         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
29504         across a vec_duplicate and a paradoxical subreg forming a vector
29505         mode to a vec_concat.
29507 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29509         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
29510         -march=armv8.3-a variants.
29511         * config/arm/t-multilib: Likewise.
29512         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
29514 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29516         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
29517         to generate rtl.
29518         (cceq_ior_compare_complement): Give it a name so I can use it, and
29519         change boolean_or_operator predicate to boolean_operator so it can
29520         be used to generate a crand.
29521         (eqne): New code iterator.
29522         (bd/bd_neg): New code_attrs.
29523         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
29524         a single define_insn.
29525         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
29526         decrement (bdnzt/bdnzf/bdzt/bdzf).
29527         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
29528         with the new names of the branch decrement patterns, and added the
29529         names of the branch decrement conditional patterns.
29531 2018-01-08  Richard Biener  <rguenther@suse.de>
29533         PR tree-optimization/83563
29534         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
29535         cache.
29537 2018-01-08  Richard Biener  <rguenther@suse.de>
29539         PR middle-end/83713
29540         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
29542 2018-01-08  Richard Biener  <rguenther@suse.de>
29544         PR tree-optimization/83685
29545         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
29546         references to abnormals.
29548 2018-01-08  Richard Biener  <rguenther@suse.de>
29550         PR lto/83719
29551         * dwarf2out.c (output_indirect_strings): Handle empty
29552         skeleton_debug_str_hash.
29553         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
29555 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29557         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
29558         (emit_store_direct): Likewise.
29559         (arc_trampoline_adjust_address): Likewise.
29560         (arc_asm_trampoline_template): New function.
29561         (arc_initialize_trampoline): Use asm_trampoline_template.
29562         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
29563         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
29564         * config/arc/arc.md (flush_icache): Delete pattern.
29566 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29568         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
29569         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
29570         munaligned-access.
29572 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29574         PR target/83681
29575         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
29576         by not USED_FOR_TARGET.
29577         (make_pass_resolve_sw_modes): Likewise.
29579 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29581         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
29582         USED_FOR_TARGET.
29584 2018-01-08  Richard Biener  <rguenther@suse.de>
29586         PR middle-end/83580
29587         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
29589 2018-01-08  Richard Biener  <rguenther@suse.de>
29591         PR middle-end/83517
29592         * match.pd ((t * 2) / 2) -> t): Add missing :c.
29594 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
29596         PR middle-end/81897
29597         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
29598         basic blocks with a small number of successors.
29599         (convert_control_dep_chain_into_preds): Improve handling of
29600         forwarder blocks.
29601         (dump_predicates): Split apart into...
29602         (dump_pred_chain): ...here...
29603         (dump_pred_info): ...and here.
29604         (can_one_predicate_be_invalidated_p): Add debugging printfs.
29605         (can_chain_union_be_invalidated_p): Improve check for invalidation
29606         of paths.
29607         (uninit_uses_cannot_happen): Avoid unnecessary if
29608         convert_control_dep_chain_into_preds yielded nothing.
29610 2018-01-06  Martin Sebor  <msebor@redhat.com>
29612         PR tree-optimization/83640
29613         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
29614         subtracting negative offset from size.
29615         (builtin_access::overlap): Adjust offset bounds of the access to fall
29616         within the size of the object if possible.
29618 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
29620         PR rtl-optimization/83699
29621         * expmed.c (extract_bit_field_1): Restrict the vector usage of
29622         extract_bit_field_as_subreg to cases in which the extracted
29623         value is also a vector.
29625         * lra-constraints.c (process_alt_operands): Test for the equivalence
29626         substitutions when detecting a possible reload cycle.
29628 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
29630         PR debug/83480
29631         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
29632         by default if flag_selective_schedling{,2}.  Formatting fixes.
29634         PR rtl-optimization/83682
29635         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
29636         if it has non-VECTOR_MODE element mode.
29637         (vec_duplicate_p): Likewise.
29639         PR middle-end/83694
29640         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
29641         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
29643 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29645         PR target/83604
29646         * config/i386/i386-builtin.def
29647         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
29648         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
29649         Require also OPTION_MASK_ISA_AVX512F in addition to
29650         OPTION_MASK_ISA_GFNI.
29651         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
29652         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
29653         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
29654         to OPTION_MASK_ISA_GFNI.
29655         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
29656         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
29657         OPTION_MASK_ISA_AVX512BW.
29658         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
29659         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
29660         addition to OPTION_MASK_ISA_GFNI.
29661         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
29662         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
29663         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
29664         to OPTION_MASK_ISA_GFNI.
29665         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
29666         a requirement for all ISAs rather than any of them with a few
29667         exceptions.
29668         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
29669         processing.
29670         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
29671         bitmasks to be enabled with 3 exceptions, instead of requiring any
29672         enabled ISA with lots of exceptions.
29673         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
29674         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
29675         Change avx512bw in isa attribute to avx512f.
29676         * config/i386/sgxintrin.h: Add license boilerplate.
29677         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
29678         to __AVX512F__ and __AVX512VL to __AVX512VL__.
29679         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
29680         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
29681         defined.
29682         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
29683         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
29684         temporarily sse2 rather than sse if not enabled already.
29686         PR target/83604
29687         * config/i386/sse.md (VI248_VLBW): Rename to ...
29688         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
29689         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
29690         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
29691         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
29692         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
29693         mode iterator instead of VI248_VLBW.
29695 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
29697         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
29698         (record_modified): Skip clobbers; add debug output.
29699         (param_change_prob): Use sreal frequencies.
29701 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29703         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
29704         punt for user-aligned variables.
29706 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29708         * tree-chrec.c (chrec_contains_symbols): Return true for
29709         POLY_INT_CST.
29711 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
29713         PR target/82439
29714         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
29715         of (x|y) == x for BICS pattern.
29717 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29719         PR tree-optimization/83605
29720         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
29721         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
29722         can throw.
29724 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29726         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
29727         * config/epiphany/rtems.h: New file.
29729 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29730             Uros Bizjak  <ubizjak@gmail.com>
29732         PR target/83554
29733         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
29734         QIreg_operand instead of register_operand predicate.
29735         * config/i386/i386.c (ix86_rop_should_change_byte_p,
29736         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
29737         comments instead of -fmitigate[-_]rop.
29739 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29741         PR bootstrap/81926
29742         * cgraphunit.c (symbol_table::compile): Switch to text_section
29743         before calling assembly_start debug hook.
29744         * run-rtl-passes.c (run_rtl_passes): Likewise.
29745         Include output.h.
29747 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29749         * tree-vrp.c (extract_range_from_binary_expr_1): Check
29750         range_int_cst_p rather than !symbolic_range_p before calling
29751         extract_range_from_multiplicative_op_1.
29753 2018-01-04  Jeff Law  <law@redhat.com>
29755         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
29756         redundant test in assertion.
29758 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29760         * doc/rtl.texi: Document machine_mode wrapper classes.
29762 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29764         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
29765         using tree_to_uhwi.
29767 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29769         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
29770         the VEC_PERM_EXPR fold to fail.
29772 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29774         PR debug/83585
29775         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
29776         to switched_sections.
29778 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29780         PR target/83680
29781         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
29782         test for d.testing.
29784 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
29786         PR target/83387
29787         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
29788         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
29790 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29792         PR debug/83666
29793         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
29794         is BLKmode and bitpos not zero or mode change is needed.
29796 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29798         PR target/83675
29799         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
29800         TARGET_VIS2.
29802 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
29804         PR target/83628
29805         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
29806         instead of MULT rtx.  Update all corresponding splitters.
29807         (*saddl_se): Ditto.
29808         (*ssub<modesuffix>): Ditto.
29809         (*ssubl_se): Ditto.
29810         (*cmp_sadd_di): Update split patterns.
29811         (*cmp_sadd_si): Ditto.
29812         (*cmp_sadd_sidi): Ditto.
29813         (*cmp_ssub_di): Ditto.
29814         (*cmp_ssub_si): Ditto.
29815         (*cmp_ssub_sidi): Ditto.
29816         * config/alpha/predicates.md (const23_operand): New predicate.
29817         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
29818         Look for ASHIFT, not MULT inner operand.
29819         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
29821 2018-01-04  Martin Liska  <mliska@suse.cz>
29823         PR gcov-profile/83669
29824         * gcov.c (output_intermediate_file): Add version to intermediate
29825         gcov file.
29826         * doc/gcov.texi: Document new field 'version' in intermediate
29827         file format. Fix location of '-k' option of gcov command.
29829 2018-01-04  Martin Liska  <mliska@suse.cz>
29831         PR ipa/82352
29832         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
29834 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29836         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
29838 2018-01-03  Martin Sebor  <msebor@redhat.com>
29840         PR tree-optimization/83655
29841         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
29842         checking calls with invalid arguments.
29844 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29846         * tree-vect-stmts.c (vect_get_store_rhs): New function.
29847         (vectorizable_mask_load_store): Delete.
29848         (vectorizable_call): Return false for masked loads and stores.
29849         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
29850         instead of gimple_assign_rhs1.
29851         (vectorizable_load): Handle IFN_MASK_LOAD.
29852         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
29854 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29856         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
29857         split out from..,
29858         (vectorizable_mask_load_store): ...here.
29859         (vectorizable_load): ...and here.
29861 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29863         * tree-vect-stmts.c (vect_build_all_ones_mask)
29864         (vect_build_zero_merge_argument): New functions, split out from...
29865         (vectorizable_load): ...here.
29867 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29869         * tree-vect-stmts.c (vect_check_store_rhs): New function,
29870         split out from...
29871         (vectorizable_mask_load_store): ...here.
29872         (vectorizable_store): ...and here.
29874 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29876         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
29877         split out from...
29878         (vectorizable_mask_load_store): ...here.
29880 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29882         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
29883         (vect_model_store_cost): Take a vec_load_store_type instead of a
29884         vect_def_type.
29885         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
29886         (vect_model_store_cost): Take a vec_load_store_type instead of a
29887         vect_def_type.
29888         (vectorizable_mask_load_store): Update accordingly.
29889         (vectorizable_store): Likewise.
29890         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
29892 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29894         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
29895         IFN_MASK_LOAD calls here rather than...
29896         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
29898 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29899             Alan Hayward  <alan.hayward@arm.com>
29900             David Sherwood  <david.sherwood@arm.com>
29902         * expmed.c (extract_bit_field_1): For vector extracts,
29903         fall back to extract_bit_field_as_subreg if vec_extract
29904         isn't available.
29906 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29907             Alan Hayward  <alan.hayward@arm.com>
29908             David Sherwood  <david.sherwood@arm.com>
29910         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
29911         they are variable or constant sized.
29912         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
29913         slots for constant-sized data.
29915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29916             Alan Hayward  <alan.hayward@arm.com>
29917             David Sherwood  <david.sherwood@arm.com>
29919         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
29920         handling COND_EXPRs with boolean comparisons, try to find a better
29921         basis for the mask type than the boolean itself.
29923 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29925         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
29926         is calculated and how it can be overridden.
29927         * genmodes.c (max_bitsize_mode_any_mode): New variable.
29928         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
29929         if defined.
29930         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
29931         if nonzero.
29933 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29934             Alan Hayward  <alan.hayward@arm.com>
29935             David Sherwood  <david.sherwood@arm.com>
29937         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
29938         Remove the mode argument.
29939         (aarch64_simd_valid_immediate): Remove the mode and inverse
29940         arguments.
29941         * config/aarch64/iterators.md (bitsize): New iterator.
29942         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
29943         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
29944         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
29945         aarch64_simd_valid_immediate.
29946         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
29947         (aarch64_reg_or_bic_imm): Likewise.
29948         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
29949         with an insn_type enum and msl with a modifier_type enum.
29950         Replace element_width with a scalar_mode.  Change the shift
29951         to unsigned int.  Add constructors for scalar_float_mode and
29952         scalar_int_mode elements.
29953         (aarch64_vect_float_const_representable_p): Delete.
29954         (aarch64_can_const_movi_rtx_p)
29955         (aarch64_simd_scalar_immediate_valid_for_move)
29956         (aarch64_simd_make_constant): Update call to
29957         aarch64_simd_valid_immediate.
29958         (aarch64_advsimd_valid_immediate_hs): New function.
29959         (aarch64_advsimd_valid_immediate): Likewise.
29960         (aarch64_simd_valid_immediate): Remove mode and inverse
29961         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
29962         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
29963         and aarch64_float_const_representable_p on the result.
29964         (aarch64_output_simd_mov_immediate): Remove mode argument.
29965         Update call to aarch64_simd_valid_immediate and use of
29966         simd_immediate_info.
29967         (aarch64_output_scalar_simd_mov_immediate): Update call
29968         accordingly.
29970 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29971             Alan Hayward  <alan.hayward@arm.com>
29972             David Sherwood  <david.sherwood@arm.com>
29974         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
29975         (mode_nunits): Likewise CONST_MODE_NUNITS.
29976         * machmode.def (ADJUST_NUNITS): Document.
29977         * genmodes.c (mode_data::need_nunits_adj): New field.
29978         (blank_mode): Update accordingly.
29979         (adj_nunits): New variable.
29980         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
29981         parameter.
29982         (emit_mode_size_inline): Set need_bytesize_adj for all modes
29983         listed in adj_nunits.
29984         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
29985         listed in adj_nunits.  Don't emit case statements for such modes.
29986         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
29987         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
29988         nothing if adj_nunits is nonnull.
29989         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
29990         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
29991         (emit_mode_fbit): Update use of print_maybe_const_decl.
29992         (emit_move_size): Likewise.  Treat the array as non-const
29993         if adj_nunits.
29994         (emit_mode_adjustments): Handle adj_nunits.
29996 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29998         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
29999         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30000         (VECTOR_MODES): Use it.
30001         (make_vector_modes): Take the prefix as an argument.
30003 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30004             Alan Hayward  <alan.hayward@arm.com>
30005             David Sherwood  <david.sherwood@arm.com>
30007         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30008         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30009         for MODE_VECTOR_BOOL.
30010         * machmode.def (VECTOR_BOOL_MODE): Document.
30011         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30012         (make_vector_bool_mode): New function.
30013         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30014         MODE_VECTOR_BOOL.
30015         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30016         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30017         Likewise.
30018         * stor-layout.c (int_mode_for_mode): Likewise.
30019         * tree.c (build_vector_type_for_mode): Likewise.
30020         * varasm.c (output_constant_pool_2): Likewise.
30021         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30022         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30023         for MODE_VECTOR_BOOL.
30024         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30025         of mode class checks.
30026         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30027         instead of a list of mode class checks.
30028         (expand_vector_scalar_condition): Likewise.
30029         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30031 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30032             Alan Hayward  <alan.hayward@arm.com>
30033             David Sherwood  <david.sherwood@arm.com>
30035         * machmode.h (mode_size): Change from unsigned short to
30036         poly_uint16_pod.
30037         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30038         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30039         or if measurement_type is not polynomial.
30040         (fixed_size_mode::includes_p): Check for constant-sized modes.
30041         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30042         return a poly_uint16 rather than an unsigned short.
30043         (emit_mode_size): Change the type of mode_size from unsigned short
30044         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30045         (emit_mode_adjustments): Cope with polynomial vector sizes.
30046         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30047         for GET_MODE_SIZE.
30048         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30049         for GET_MODE_SIZE.
30050         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30051         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30052         * caller-save.c (setup_save_areas): Likewise.
30053         (replace_reg_with_saved_mem): Likewise.
30054         * calls.c (emit_library_call_value_1): Likewise.
30055         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30056         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30057         (gen_lowpart_for_combine): Likewise.
30058         * convert.c (convert_to_integer_1): Likewise.
30059         * cse.c (equiv_constant, cse_insn): Likewise.
30060         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30061         (cselib_subst_to_values): Likewise.
30062         * dce.c (word_dce_process_block): Likewise.
30063         * df-problems.c (df_word_lr_mark_ref): Likewise.
30064         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30065         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30066         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30067         (rtl_for_decl_location): Likewise.
30068         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30069         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30070         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30071         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30072         (expand_expr_real_1): Likewise.
30073         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30074         (pad_below): Likewise.
30075         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30076         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30077         * ira.c (get_subreg_tracking_sizes): Likewise.
30078         * ira-build.c (ira_create_allocno_objects): Likewise.
30079         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30080         (ira_sort_regnos_for_alter_reg): Likewise.
30081         * ira-costs.c (record_operand_costs): Likewise.
30082         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30083         (resolve_simple_move): Likewise.
30084         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30085         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30086         (lra_constraints): Likewise.
30087         (CONST_POOL_OK_P): Reject variable-sized modes.
30088         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30089         (add_pseudo_to_slot, lra_spill): Likewise.
30090         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30091         * optabs-query.c (get_best_extraction_insn): Likewise.
30092         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30093         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30094         (expand_mult_highpart, valid_multiword_target_p): Likewise.
30095         * recog.c (offsettable_address_addr_space_p): Likewise.
30096         * regcprop.c (maybe_mode_change): Likewise.
30097         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30098         * regrename.c (build_def_use): Likewise.
30099         * regstat.c (dump_reg_info): Likewise.
30100         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30101         (find_reloads, find_reloads_subreg_address): Likewise.
30102         * reload1.c (eliminate_regs_1): Likewise.
30103         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30104         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30105         (simplify_binary_operation_1, simplify_subreg): Likewise.
30106         * targhooks.c (default_function_arg_padding): Likewise.
30107         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30108         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30109         (verify_gimple_assign_ternary): Likewise.
30110         * tree-inline.c (estimate_move_cost): Likewise.
30111         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30112         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30113         (get_address_cost_ainc): Likewise.
30114         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30115         (vect_supportable_dr_alignment): Likewise.
30116         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30117         (vectorizable_reduction): Likewise.
30118         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30119         (vectorizable_operation, vectorizable_load): Likewise.
30120         * tree.c (build_same_sized_truth_vector_type): Likewise.
30121         * valtrack.c (cleanup_auto_inc_dec): Likewise.
30122         * var-tracking.c (emit_note_insn_var_location): Likewise.
30123         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30124         (ADDR_VEC_ALIGN): Likewise.
30126 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30127             Alan Hayward  <alan.hayward@arm.com>
30128             David Sherwood  <david.sherwood@arm.com>
30130         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30131         unsigned short.
30132         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30133         or if measurement_type is polynomial.
30134         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30135         * combine.c (make_extraction): Likewise.
30136         * dse.c (find_shift_sequence): Likewise.
30137         * dwarf2out.c (mem_loc_descriptor): Likewise.
30138         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30139         (extract_bit_field, extract_low_bits): Likewise.
30140         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30141         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30142         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30143         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30144         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30145         * reload.c (find_reloads): Likewise.
30146         * reload1.c (alter_reg): Likewise.
30147         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30148         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30149         * tree-if-conv.c (predicate_mem_writes): Likewise.
30150         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30151         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30152         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30153         * valtrack.c (dead_debug_insert_temp): Likewise.
30154         * varasm.c (mergeable_constant_section): Likewise.
30155         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30158             Alan Hayward  <alan.hayward@arm.com>
30159             David Sherwood  <david.sherwood@arm.com>
30161         * expr.c (expand_assignment): Cope with polynomial mode sizes
30162         when assigning to a CONCAT.
30164 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30165             Alan Hayward  <alan.hayward@arm.com>
30166             David Sherwood  <david.sherwood@arm.com>
30168         * machmode.h (mode_precision): Change from unsigned short to
30169         poly_uint16_pod.
30170         (mode_to_precision): Return a poly_uint16 rather than an unsigned
30171         short.
30172         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30173         or if measurement_type is not polynomial.
30174         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
30175         in which the mode is already known to be a scalar_int_mode.
30176         * genmodes.c (emit_mode_precision): Change the type of mode_precision
30177         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
30178         initializer.
30179         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30180         for GET_MODE_PRECISION.
30181         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30182         for GET_MODE_PRECISION.
30183         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30184         as polynomial.
30185         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30186         (expand_field_assignment, make_extraction): Likewise.
30187         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30188         (get_last_value): Likewise.
30189         * convert.c (convert_to_integer_1): Likewise.
30190         * cse.c (cse_insn): Likewise.
30191         * expr.c (expand_expr_real_1): Likewise.
30192         * lra-constraints.c (simplify_operand_subreg): Likewise.
30193         * optabs-query.c (can_atomic_load_p): Likewise.
30194         * optabs.c (expand_atomic_load): Likewise.
30195         (expand_atomic_store): Likewise.
30196         * ree.c (combine_reaching_defs): Likewise.
30197         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30198         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30199         * tree.h (type_has_mode_precision_p): Likewise.
30200         * ubsan.c (instrument_si_overflow): Likewise.
30202 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30203             Alan Hayward  <alan.hayward@arm.com>
30204             David Sherwood  <david.sherwood@arm.com>
30206         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30207         polynomial numbers of units.
30208         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30209         (valid_vector_subparts_p): New function.
30210         (build_vector_type): Remove temporary shim and take the number
30211         of units as a poly_uint64 rather than an int.
30212         (build_opaque_vector_type): Take the number of units as a
30213         poly_uint64 rather than an int.
30214         * tree.c (build_vector_from_ctor): Handle polynomial
30215         TYPE_VECTOR_SUBPARTS.
30216         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30217         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30218         (build_vector_from_val): If the number of units is variable,
30219         use build_vec_duplicate_cst for constant operands and
30220         VEC_DUPLICATE_EXPR otherwise.
30221         (make_vector_type): Remove temporary is_constant ().
30222         (build_vector_type, build_opaque_vector_type): Take the number of
30223         units as a poly_uint64 rather than an int.
30224         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30225         VECTOR_CST_NELTS.
30226         * cfgexpand.c (expand_debug_expr): Likewise.
30227         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30228         (store_constructor, expand_expr_real_1): Likewise.
30229         (const_scalar_mask_from_tree): Likewise.
30230         * fold-const-call.c (fold_const_reduction): Likewise.
30231         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30232         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30233         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30234         (fold_relational_const): Likewise.
30235         (native_interpret_vector): Likewise.  Change the size from an
30236         int to an unsigned int.
30237         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30238         TYPE_VECTOR_SUBPARTS.
30239         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30240         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30241         duplicating a non-constant operand into a variable-length vector.
30242         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30243         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30244         * ipa-icf.c (sem_variable::equals): Likewise.
30245         * match.pd: Likewise.
30246         * omp-simd-clone.c (simd_clone_subparts): Likewise.
30247         * print-tree.c (print_node): Likewise.
30248         * stor-layout.c (layout_type): Likewise.
30249         * targhooks.c (default_builtin_vectorization_cost): Likewise.
30250         * tree-cfg.c (verify_gimple_comparison): Likewise.
30251         (verify_gimple_assign_binary): Likewise.
30252         (verify_gimple_assign_ternary): Likewise.
30253         (verify_gimple_assign_single): Likewise.
30254         * tree-pretty-print.c (dump_generic_node): Likewise.
30255         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30256         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30257         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30258         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30259         (vect_shift_permute_load_chain): Likewise.
30260         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30261         (expand_vector_condition, optimize_vector_constructor): Likewise.
30262         (lower_vec_perm, get_compute_type): Likewise.
30263         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30264         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30265         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30266         (vect_recog_mask_conversion_pattern): Likewise.
30267         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30268         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30269         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30270         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30271         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30272         (vectorizable_shift, vectorizable_operation, vectorizable_store)
30273         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30274         (supportable_widening_operation): Likewise.
30275         (supportable_narrowing_operation): Likewise.
30276         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30277         Likewise.
30278         * varasm.c (output_constant): Likewise.
30280 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30281             Alan Hayward  <alan.hayward@arm.com>
30282             David Sherwood  <david.sherwood@arm.com>
30284         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30285         so that both the length == 3 and length != 3 cases set up their
30286         own permute vectors.  Add comments explaining why we know the
30287         number of elements is constant.
30288         (vect_permute_load_chain): Likewise.
30290 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30291             Alan Hayward  <alan.hayward@arm.com>
30292             David Sherwood  <david.sherwood@arm.com>
30294         * machmode.h (mode_nunits): Change from unsigned char to
30295         poly_uint16_pod.
30296         (ONLY_FIXED_SIZE_MODES): New macro.
30297         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30298         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30299         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30300         New typedefs.
30301         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30302         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30303         or if measurement_type is not polynomial.
30304         * genmodes.c (ZERO_COEFFS): New macro.
30305         (emit_mode_nunits_inline): Make mode_nunits_inline return a
30306         poly_uint16.
30307         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30308         Use ZERO_COEFFS when emitting initializers.
30309         * data-streamer.h (bp_pack_poly_value): New function.
30310         (bp_unpack_poly_value): Likewise.
30311         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30312         for GET_MODE_NUNITS.
30313         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30314         for GET_MODE_NUNITS.
30315         * tree.c (make_vector_type): Remove temporary shim and make
30316         the real function take the number of units as a poly_uint64
30317         rather than an int.
30318         (build_vector_type_for_mode): Handle polynomial nunits.
30319         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
30320         * emit-rtl.c (const_vec_series_p_1): Likewise.
30321         (gen_rtx_CONST_VECTOR): Likewise.
30322         * fold-const.c (test_vec_duplicate_folding): Likewise.
30323         * genrecog.c (validate_pattern): Likewise.
30324         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
30325         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30326         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
30327         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
30328         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
30329         * rtlanal.c (subreg_get_info): Likewise.
30330         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30331         (vect_grouped_load_supported): Likewise.
30332         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
30333         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
30334         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
30335         (simplify_const_unary_operation, simplify_binary_operation_1)
30336         (simplify_const_binary_operation, simplify_ternary_operation)
30337         (test_vector_ops_duplicate, test_vector_ops): Likewise.
30338         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
30339         instead of CONST_VECTOR_NUNITS.
30340         * varasm.c (output_constant_pool_2): Likewise.
30341         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
30342         explicit-encoded elements in the XVEC for variable-length vectors.
30344 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30346         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
30348 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30349             Alan Hayward  <alan.hayward@arm.com>
30350             David Sherwood  <david.sherwood@arm.com>
30352         * coretypes.h (fixed_size_mode): Declare.
30353         (fixed_size_mode_pod): New typedef.
30354         * builtins.h (target_builtins::x_apply_args_mode)
30355         (target_builtins::x_apply_result_mode): Change type to
30356         fixed_size_mode_pod.
30357         * builtins.c (apply_args_size, apply_result_size, result_vector)
30358         (expand_builtin_apply_args_1, expand_builtin_apply)
30359         (expand_builtin_return): Update accordingly.
30361 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30363         * cse.c (hash_rtx_cb): Hash only the encoded elements.
30364         * cselib.c (cselib_hash_rtx): Likewise.
30365         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
30366         CONST_VECTOR encoding.
30368 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30369             Jeff Law  <law@redhat.com>
30371         PR target/83641
30372         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
30373         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
30374         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
30375         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
30377         PR target/83641
30378         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
30379         explicitly probe *sp in a noreturn function if there were any callee
30380         register saves or frame pointer is needed.
30382 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30384         PR debug/83621
30385         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
30386         BLKmode for ternary, binary or unary expressions.
30388         PR debug/83645
30389         * var-tracking.c (delete_vta_debug_insn): New inline function.
30390         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
30391         insns from get_insns () to NULL instead of each bb separately.
30392         Use delete_vta_debug_insn.  No longer static.
30393         (vt_debug_insns_local, variable_tracking_main_1): Adjust
30394         delete_vta_debug_insns callers.
30395         * rtl.h (delete_vta_debug_insns): Declare.
30396         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
30397         instead of variable_tracking_main.
30399 2018-01-03  Martin Sebor  <msebor@redhat.com>
30401         PR tree-optimization/83603
30402         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
30403         arguments past the endof the argument list in functions declared
30404         without a prototype.
30405         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
30406         Avoid checking when arguments are null.
30408 2018-01-03  Martin Sebor  <msebor@redhat.com>
30410         PR c/83559
30411         * doc/extend.texi (attribute const): Fix a typo.
30412         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
30413         issuing -Wsuggest-attribute for void functions.
30415 2018-01-03  Martin Sebor  <msebor@redhat.com>
30417         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
30418         offset_int::from instead of wide_int::to_shwi.
30419         (maybe_diag_overlap): Remove assertion.
30420         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
30421         * gimple-ssa-sprintf.c (format_directive): Same.
30422         (parse_directive): Same.
30423         (sprintf_dom_walker::compute_format_length): Same.
30424         (try_substitute_return_value): Same.
30426 2018-01-03  Jeff Law  <law@redhat.com>
30428         PR middle-end/83654
30429         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
30430         non-constant residual for zero at runtime and avoid probing in
30431         that case.  Reorganize code for trailing problem to mirror handling
30432         of the residual.
30434 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30436         PR tree-optimization/83501
30437         * tree-ssa-strlen.c (get_string_cst): New.
30438         (handle_char_store): Call get_string_cst.
30440 2018-01-03  Martin Liska  <mliska@suse.cz>
30442         PR tree-optimization/83593
30443         * tree-ssa-strlen.c: Include tree-cfg.h.
30444         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
30445         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
30446         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
30447         to false.
30448         (strlen_dom_walker::before_dom_children): Call
30449         gimple_purge_dead_eh_edges. Dump tranformation with details
30450         dump flags.
30451         (strlen_dom_walker::before_dom_children): Update call by adding
30452         new argument cleanup_eh.
30453         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
30455 2018-01-03  Martin Liska  <mliska@suse.cz>
30457         PR ipa/83549
30458         * cif-code.def (VARIADIC_THUNK): New enum value.
30459         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
30460         thunks.
30462 2018-01-03  Jan Beulich  <jbeulich@suse.com>
30464         * sse.md (mov<mode>_internal): Tighten condition for when to use
30465         vmovdqu<ssescalarsize> for TI and OI modes.
30467 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30469         Update copyright years.
30471 2018-01-03  Martin Liska  <mliska@suse.cz>
30473         PR ipa/83594
30474         * ipa-visibility.c (function_and_variable_visibility): Skip
30475         functions with noipa attribure.
30477 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30479         * gcc.c (process_command): Update copyright notice dates.
30480         * gcov-dump.c (print_version): Ditto.
30481         * gcov.c (print_version): Ditto.
30482         * gcov-tool.c (print_version): Ditto.
30483         * gengtype.c (create_file): Ditto.
30484         * doc/cpp.texi: Bump @copying's copyright year.
30485         * doc/cppinternals.texi: Ditto.
30486         * doc/gcc.texi: Ditto.
30487         * doc/gccint.texi: Ditto.
30488         * doc/gcov.texi: Ditto.
30489         * doc/install.texi: Ditto.
30490         * doc/invoke.texi: Ditto.
30492 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30494         * vector-builder.h (vector_builder::m_full_nelts): Change from
30495         unsigned int to poly_uint64.
30496         (vector_builder::full_nelts): Update prototype accordingly.
30497         (vector_builder::new_vector): Likewise.
30498         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
30499         (vector_builder::operator ==): Likewise.
30500         (vector_builder::finalize): Likewise.
30501         * int-vector-builder.h (int_vector_builder::int_vector_builder):
30502         Take the number of elements as a poly_uint64 rather than an
30503         unsigned int.
30504         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
30505         from unsigned int to poly_uint64.
30506         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
30507         (vec_perm_indices::new_vector): Likewise.
30508         (vec_perm_indices::length): Likewise.
30509         (vec_perm_indices::nelts_per_input): Likewise.
30510         (vec_perm_indices::input_nelts): Likewise.
30511         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
30512         number of elements per input as a poly_uint64 rather than an
30513         unsigned int.  Use the original encoding for variable-length
30514         vectors, rather than clamping each individual element.
30515         For the second and subsequent elements in each pattern,
30516         clamp the step and base before clamping their sum.
30517         (vec_perm_indices::series_p): Handle polynomial element counts.
30518         (vec_perm_indices::all_in_range_p): Likewise.
30519         (vec_perm_indices_to_tree): Likewise.
30520         (vec_perm_indices_to_rtx): Likewise.
30521         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
30522         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
30523         (tree_vector_builder::new_binary_operation): Handle polynomial
30524         element counts.  Return false if we need to know the number
30525         of elements at compile time.
30526         * fold-const.c (fold_vec_perm): Punt if the number of elements
30527         isn't known at compile time.
30529 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30531         * vec-perm-indices.h (vec_perm_builder): Change element type
30532         from HOST_WIDE_INT to poly_int64.
30533         (vec_perm_indices::element_type): Update accordingly.
30534         (vec_perm_indices::clamp): Handle polynomial element_types.
30535         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30536         (vec_perm_indices::all_in_range_p): Likewise.
30537         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
30538         than shwi trees.
30539         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
30540         polynomial vec_perm_indices element types.
30541         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
30542         * fold-const.c (fold_vec_perm): Likewise.
30543         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
30544         * tree-vect-generic.c (lower_vec_perm): Likewise.
30545         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30546         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
30547         element type to HOST_WIDE_INT.
30549 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30550             Alan Hayward  <alan.hayward@arm.com>
30551             David Sherwood  <david.sherwood@arm.com>
30553         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
30554         rather than an int.  Use plus_constant.
30555         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
30556         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
30558 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30559             Alan Hayward  <alan.hayward@arm.com>
30560             David Sherwood  <david.sherwood@arm.com>
30562         * calls.c (emit_call_1, expand_call): Change struct_value_size from
30563         a HOST_WIDE_INT to a poly_int64.
30565 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30566             Alan Hayward  <alan.hayward@arm.com>
30567             David Sherwood  <david.sherwood@arm.com>
30569         * calls.c (load_register_parameters): Cope with polynomial
30570         mode sizes.  Require a constant size for BLKmode parameters
30571         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
30572         forces a parameter to be padded at the lsb end in order to
30573         fill a complete number of words, require the parameter size
30574         to be ordered wrt UNITS_PER_WORD.
30576 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30577             Alan Hayward  <alan.hayward@arm.com>
30578             David Sherwood  <david.sherwood@arm.com>
30580         * reload1.c (spill_stack_slot_width): Change element type
30581         from unsigned int to poly_uint64_pod.
30582         (alter_reg): Treat mode sizes as polynomial.
30584 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30585             Alan Hayward  <alan.hayward@arm.com>
30586             David Sherwood  <david.sherwood@arm.com>
30588         * reload.c (complex_word_subreg_p): New function.
30589         (reload_inner_reg_of_subreg, push_reload): Use it.
30591 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30592             Alan Hayward  <alan.hayward@arm.com>
30593             David Sherwood  <david.sherwood@arm.com>
30595         * lra-constraints.c (process_alt_operands): Reject matched
30596         operands whose sizes aren't ordered.
30597         (match_reload): Refer to this check here.
30599 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30600             Alan Hayward  <alan.hayward@arm.com>
30601             David Sherwood  <david.sherwood@arm.com>
30603         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
30604         that the mode size is in the set {1, 2, 4, 8, 16}.
30606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30607             Alan Hayward  <alan.hayward@arm.com>
30608             David Sherwood  <david.sherwood@arm.com>
30610         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
30611         Use plus_constant instead of gen_rtx_PLUS.
30613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30614             Alan Hayward  <alan.hayward@arm.com>
30615             David Sherwood  <david.sherwood@arm.com>
30617         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
30618         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
30619         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
30620         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
30621         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
30622         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
30623         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
30624         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
30625         * config/i386/i386.c (ix86_push_rounding): ...this new function.
30626         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
30627         a poly_int64.
30628         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
30629         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
30630         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
30631         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
30632         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
30633         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
30634         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
30635         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
30636         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
30637         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
30638         function.
30639         * expr.c (emit_move_resolve_push): Treat the input and result
30640         of PUSH_ROUNDING as a poly_int64.
30641         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
30642         (emit_push_insn): Likewise.
30643         * lra-eliminations.c (mark_not_eliminable): Likewise.
30644         * recog.c (push_operand): Likewise.
30645         * reload1.c (elimination_effects): Likewise.
30646         * rtlanal.c (nonzero_bits1): Likewise.
30647         * calls.c (store_one_arg): Likewise.  Require the padding to be
30648         known at compile time.
30650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30651             Alan Hayward  <alan.hayward@arm.com>
30652             David Sherwood  <david.sherwood@arm.com>
30654         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
30655         Use plus_constant instead of gen_rtx_PLUS.
30657 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30658             Alan Hayward  <alan.hayward@arm.com>
30659             David Sherwood  <david.sherwood@arm.com>
30661         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
30662         rather than an int.
30664 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30665             Alan Hayward  <alan.hayward@arm.com>
30666             David Sherwood  <david.sherwood@arm.com>
30668         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
30669         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
30670         via stack temporaries.  Treat the mode size as polynomial too.
30672 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30673             Alan Hayward  <alan.hayward@arm.com>
30674             David Sherwood  <david.sherwood@arm.com>
30676         * expr.c (expand_expr_real_2): When handling conversions involving
30677         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
30678         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
30679         as a poly_uint64 too.
30681 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30682             Alan Hayward  <alan.hayward@arm.com>
30683             David Sherwood  <david.sherwood@arm.com>
30685         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
30687 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30688             Alan Hayward  <alan.hayward@arm.com>
30689             David Sherwood  <david.sherwood@arm.com>
30691         * combine.c (can_change_dest_mode): Handle polynomial
30692         REGMODE_NATURAL_SIZE.
30693         * expmed.c (store_bit_field_1): Likewise.
30694         * expr.c (store_constructor): Likewise.
30695         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
30696         and polynomial REGMODE_NATURAL_SIZE.
30697         (gen_lowpart_common): Likewise.
30698         * reginfo.c (record_subregs_of_mode): Likewise.
30699         * rtlanal.c (read_modify_subreg_p): Likewise.
30701 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30702             Alan Hayward  <alan.hayward@arm.com>
30703             David Sherwood  <david.sherwood@arm.com>
30705         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
30706         numbers of elements.
30708 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30709             Alan Hayward  <alan.hayward@arm.com>
30710             David Sherwood  <david.sherwood@arm.com>
30712         * match.pd: Cope with polynomial numbers of vector elements.
30714 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30715             Alan Hayward  <alan.hayward@arm.com>
30716             David Sherwood  <david.sherwood@arm.com>
30718         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
30719         in a POINTER_PLUS_EXPR.
30721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30722             Alan Hayward  <alan.hayward@arm.com>
30723             David Sherwood  <david.sherwood@arm.com>
30725         * omp-simd-clone.c (simd_clone_subparts): New function.
30726         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
30727         (ipa_simd_modify_function_body): Likewise.
30729 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30730             Alan Hayward  <alan.hayward@arm.com>
30731             David Sherwood  <david.sherwood@arm.com>
30733         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
30734         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
30735         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
30736         (expand_vector_condition, vector_element): Likewise.
30737         (subparts_gt): New function.
30738         (get_compute_type): Use subparts_gt.
30739         (count_type_subparts): Delete.
30740         (expand_vector_operations_1): Use subparts_gt instead of
30741         count_type_subparts.
30743 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30744             Alan Hayward  <alan.hayward@arm.com>
30745             David Sherwood  <david.sherwood@arm.com>
30747         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
30748         (vect_compile_time_alias): ...this new function.  Do the calculation
30749         on poly_ints rather than trees.
30750         (vect_prune_runtime_alias_test_list): Update call accordingly.
30752 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30753             Alan Hayward  <alan.hayward@arm.com>
30754             David Sherwood  <david.sherwood@arm.com>
30756         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
30757         numbers of units.
30758         (vect_schedule_slp_instance): Likewise.
30760 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30761             Alan Hayward  <alan.hayward@arm.com>
30762             David Sherwood  <david.sherwood@arm.com>
30764         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
30765         constant and extern definitions for variable-length vectors.
30766         (vect_get_constant_vectors): Note that the number of units
30767         is known to be constant.
30769 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30770             Alan Hayward  <alan.hayward@arm.com>
30771             David Sherwood  <david.sherwood@arm.com>
30773         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
30774         of units as polynomial.  Choose between WIDE and NARROW based
30775         on multiple_p.
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         * tree-vect-stmts.c (simd_clone_subparts): New function.
30782         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
30784 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30785             Alan Hayward  <alan.hayward@arm.com>
30786             David Sherwood  <david.sherwood@arm.com>
30788         * tree-vect-stmts.c (vectorizable_call): Treat the number of
30789         vectors as polynomial.  Use build_index_vector for
30790         IFN_GOMP_SIMD_LANE.
30792 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30793             Alan Hayward  <alan.hayward@arm.com>
30794             David Sherwood  <david.sherwood@arm.com>
30796         * tree-vect-stmts.c (get_load_store_type): Treat the number of
30797         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
30798         for variable-length vectors.
30799         (vectorizable_mask_load_store): Treat the number of units as
30800         polynomial, asserting that it is constant if the condition has
30801         already been enforced.
30802         (vectorizable_store, vectorizable_load): Likewise.
30804 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30805             Alan Hayward  <alan.hayward@arm.com>
30806             David Sherwood  <david.sherwood@arm.com>
30808         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
30809         of units as polynomial.  Punt if we can't tell at compile time
30810         which vector contains the final result.
30812 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30813             Alan Hayward  <alan.hayward@arm.com>
30814             David Sherwood  <david.sherwood@arm.com>
30816         * tree-vect-loop.c (vectorizable_induction): Treat the number
30817         of units as polynomial.  Punt on SLP inductions.  Use an integer
30818         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
30819         cast of such a series for variable-length floating-point
30820         reductions.
30822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30823             Alan Hayward  <alan.hayward@arm.com>
30824             David Sherwood  <david.sherwood@arm.com>
30826         * tree.h (build_index_vector): Declare.
30827         * tree.c (build_index_vector): New function.
30828         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
30829         of units as polynomial, forcibly converting it to a constant if
30830         vectorizable_reduction has already enforced the condition.
30831         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
30832         to create a {1,2,3,...} vector.
30833         (vectorizable_reduction): Treat the number of units as polynomial.
30834         Choose vectype_in based on the largest scalar element size rather
30835         than the smallest number of units.  Enforce the restrictions
30836         relied on above.
30838 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30839             Alan Hayward  <alan.hayward@arm.com>
30840             David Sherwood  <david.sherwood@arm.com>
30842         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
30843         number of units as polynomial.
30845 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30846             Alan Hayward  <alan.hayward@arm.com>
30847             David Sherwood  <david.sherwood@arm.com>
30849         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
30850         * target.def (autovectorize_vector_sizes): Return the vector sizes
30851         by pointer, using vector_sizes rather than a bitmask.
30852         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
30853         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
30854         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
30855         Likewise.
30856         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
30857         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
30858         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
30859         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
30860         * omp-general.c (omp_max_vf): Likewise.
30861         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30862         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
30863         * tree-vect-loop.c (vect_analyze_loop): Likewise.
30864         * tree-vect-slp.c (vect_slp_bb): Likewise.
30865         * doc/tm.texi: Regenerate.
30866         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
30867         to a poly_uint64.
30868         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
30869         the vector size as a poly_uint64 rather than an unsigned int.
30870         (current_vector_size): Change from an unsigned int to a poly_uint64.
30871         (get_vectype_for_scalar_type): Update accordingly.
30872         * tree.h (build_truth_vector_type): Take the size and number of
30873         units as a poly_uint64 rather than an unsigned int.
30874         (build_vector_type): Add a temporary overload that takes
30875         the number of units as a poly_uint64 rather than an unsigned int.
30876         * tree.c (make_vector_type): Likewise.
30877         (build_truth_vector_type): Take the number of units as a poly_uint64
30878         rather than an unsigned int.
30880 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30881             Alan Hayward  <alan.hayward@arm.com>
30882             David Sherwood  <david.sherwood@arm.com>
30884         * target.def (get_mask_mode): Take the number of units and length
30885         as poly_uint64s rather than unsigned ints.
30886         * targhooks.h (default_get_mask_mode): Update accordingly.
30887         * targhooks.c (default_get_mask_mode): Likewise.
30888         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
30889         * doc/tm.texi: Regenerate.
30891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30892             Alan Hayward  <alan.hayward@arm.com>
30893             David Sherwood  <david.sherwood@arm.com>
30895         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
30896         * omp-general.c (omp_max_vf): Likewise.
30897         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
30898         (expand_omp_simd): Handle polynomial safelen.
30899         * omp-low.c (omplow_simd_context): Add a default constructor.
30900         (omplow_simd_context::max_vf): Change from int to poly_uint64.
30901         (lower_rec_simd_input_clauses): Update accordingly.
30902         (lower_rec_input_clauses): Likewise.
30904 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30905             Alan Hayward  <alan.hayward@arm.com>
30906             David Sherwood  <david.sherwood@arm.com>
30908         * tree-vectorizer.h (vect_nunits_for_cost): New function.
30909         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
30910         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
30911         (vect_analyze_slp_cost): Likewise.
30912         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
30913         (vect_model_load_cost): Likewise.
30915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30916             Alan Hayward  <alan.hayward@arm.com>
30917             David Sherwood  <david.sherwood@arm.com>
30919         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
30920         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
30921         from an unsigned int * to a poly_uint64_pod *.
30922         (calculate_unrolling_factor): New function.
30923         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
30925 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30926             Alan Hayward  <alan.hayward@arm.com>
30927             David Sherwood  <david.sherwood@arm.com>
30929         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
30930         from an unsigned int to a poly_uint64.
30931         (_loop_vec_info::slp_unrolling_factor): Likewise.
30932         (_loop_vec_info::vectorization_factor): Change from an int
30933         to a poly_uint64.
30934         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
30935         (vect_get_num_vectors): New function.
30936         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
30937         (vect_get_num_copies): Use vect_get_num_vectors.
30938         (vect_analyze_data_ref_dependences): Change max_vf from an int *
30939         to an unsigned int *.
30940         (vect_analyze_data_refs): Change min_vf from an int * to a
30941         poly_uint64 *.
30942         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30943         than an unsigned HOST_WIDE_INT.
30944         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
30945         (vect_analyze_data_ref_dependence): Change max_vf from an int *
30946         to an unsigned int *.
30947         (vect_analyze_data_ref_dependences): Likewise.
30948         (vect_compute_data_ref_alignment): Handle polynomial vf.
30949         (vect_enhance_data_refs_alignment): Likewise.
30950         (vect_prune_runtime_alias_test_list): Likewise.
30951         (vect_shift_permute_load_chain): Likewise.
30952         (vect_supportable_dr_alignment): Likewise.
30953         (dependence_distance_ge_vf): Take the vectorization factor as a
30954         poly_uint64 rather than an unsigned HOST_WIDE_INT.
30955         (vect_analyze_data_refs): Change min_vf from an int * to a
30956         poly_uint64 *.
30957         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
30958         vfm1 as a poly_uint64 rather than an int.  Make the same change
30959         for the returned bound_scalar.
30960         (vect_gen_vector_loop_niters): Handle polynomial vf.
30961         (vect_do_peeling): Likewise.  Update call to
30962         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
30963         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
30964         be constant.
30965         * tree-vect-loop.c (vect_determine_vectorization_factor)
30966         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
30967         (vect_get_known_peeling_cost): Likewise.
30968         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
30969         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
30970         (vect_transform_loop): Likewise.  Use the lowest possible VF when
30971         updating the upper bounds of the loop.
30972         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
30973         rather than an int.
30974         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
30975         polynomial unroll factors.
30976         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
30977         (vect_make_slp_decision): Likewise.
30978         (vect_supported_load_permutation_p): Likewise, and polynomial
30979         vf too.
30980         (vect_analyze_slp_cost): Handle polynomial vf.
30981         (vect_slp_analyze_node_operations): Likewise.
30982         (vect_slp_analyze_bb_1): Likewise.
30983         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30984         than an unsigned HOST_WIDE_INT.
30985         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
30986         (vectorizable_load): Handle polynomial vf.
30987         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
30988         a poly_uint64.
30989         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
30991 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30992             Alan Hayward  <alan.hayward@arm.com>
30993             David Sherwood  <david.sherwood@arm.com>
30995         * match.pd: Handle bit operations involving three constants
30996         and try to fold one pair.
30998 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31000         * tree-vect-loop-manip.c: Include gimple-fold.h.
31001         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31002         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31003         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31004         Add a path that uses a step of VF instead of 1, but disable it
31005         for now.
31006         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31007         and niters_no_overflow parameters.  Update calls to
31008         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31009         Create a new SSA name if the latter choses to use a ste other
31010         than zero, and return it via niters_vector_mult_vf_var.
31011         * tree-vect-loop.c (vect_transform_loop): Update calls to
31012         vect_do_peeling, vect_gen_vector_loop_niters and
31013         slpeel_make_loop_iterate_ntimes.
31014         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31015         (vect_gen_vector_loop_niters): Update declarations after above changes.
31017 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31019         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31020         128-bit round to integer instructions.
31021         (ceil<mode>2): Likewise.
31022         (btrunc<mode>2): Likewise.
31023         (round<mode>2): Likewise.
31025 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31027         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31028         unaligned VSX load/store on P8/P9.
31029         (expand_block_clear): Allow the use of unaligned VSX
31030         load/store on P8/P9.
31032 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31034         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31035         New function.
31036         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31037         swap associated with both a load and a store.
31039 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31041         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31042         * config/riscv/riscv.md (clear_cache): Use it.
31044 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31046         * web.c: Remove out-of-date comment.
31048 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31050         * expr.c (fixup_args_size_notes): Check that any existing
31051         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31052         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31053         (emit_single_push_insn): ...here.
31055 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31057         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31058         (const_vector_encoded_nelts): New function.
31059         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31060         (const_vector_int_elt, const_vector_elt): Declare.
31061         * emit-rtl.c (const_vector_int_elt_1): New function.
31062         (const_vector_elt): Likewise.
31063         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31064         of CONST_VECTOR_ELT.
31066 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31068         * expr.c: Include rtx-vector-builder.h.
31069         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31070         directly on the tree encoding.
31071         (const_vector_from_tree): Likewise.
31072         * optabs.c: Include rtx-vector-builder.h.
31073         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31074         sequence of "u" values.
31075         * vec-perm-indices.c: Include rtx-vector-builder.h.
31076         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31077         directly on the vec_perm_indices encoding.
31079 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31081         * doc/rtl.texi (const_vector): Describe new encoding scheme.
31082         * Makefile.in (OBJS): Add rtx-vector-builder.o.
31083         * rtx-vector-builder.h: New file.
31084         * rtx-vector-builder.c: Likewise.
31085         * rtl.h (rtx_def::u2): Add a const_vector field.
31086         (CONST_VECTOR_NPATTERNS): New macro.
31087         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31088         (CONST_VECTOR_DUPLICATE_P): Likewise.
31089         (CONST_VECTOR_STEPPED_P): Likewise.
31090         (CONST_VECTOR_ENCODED_ELT): Likewise.
31091         (const_vec_duplicate_p): Check for a duplicated vector encoding.
31092         (unwrap_const_vec_duplicate): Likewise.
31093         (const_vec_series_p): Check for a non-duplicated vector encoding.
31094         Say that the function only returns true for integer vectors.
31095         * emit-rtl.c: Include rtx-vector-builder.h.
31096         (gen_const_vec_duplicate_1): Delete.
31097         (gen_const_vector): Call gen_const_vec_duplicate instead of
31098         gen_const_vec_duplicate_1.
31099         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31100         (gen_const_vec_duplicate): Use rtx_vector_builder.
31101         (gen_const_vec_series): Likewise.
31102         (gen_rtx_CONST_VECTOR): Likewise.
31103         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31104         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31105         Build a new vector rather than modifying a CONST_VECTOR in-place.
31106         (handle_special_swappables): Update call accordingly.
31107         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31108         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31109         Build a new vector rather than modifying a CONST_VECTOR in-place.
31110         (handle_special_swappables): Update call accordingly.
31112 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31114         * simplify-rtx.c (simplify_const_binary_operation): Use
31115         CONST_VECTOR_ELT instead of XVECEXP.
31117 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31119         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31120         the selector elements to be different from the data elements
31121         if the selector is a VECTOR_CST.
31122         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31123         ssizetype for the selector.
31125 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31127         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31128         before testing each element individually.
31129         * tree-vect-generic.c (lower_vec_perm): Likewise.
31131 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31133         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31134         * selftest-run-tests.c (selftest::run_tests): Call it.
31135         * vector-builder.h (vector_builder::operator ==): New function.
31136         (vector_builder::operator !=): Likewise.
31137         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31138         (vec_perm_indices::all_from_input_p): New function.
31139         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31140         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31141         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31142         instead of reading the VECTOR_CST directly.  Detect whether both
31143         vector inputs are the same before constructing the vec_perm_indices,
31144         and update the number of inputs argument accordingly.  Use the
31145         utility functions added above.  Only construct sel2 if we need to.
31147 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31149         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31150         the broadcast of the low byte.
31151         (expand_mult_highpart): Use an explicit encoding for the permutes.
31152         * optabs-query.c (can_mult_highpart_p): Likewise.
31153         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31154         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31155         (vectorizable_bswap): Likewise.
31156         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31157         explicit encoding for the power-of-2 permutes.
31158         (vect_permute_store_chain): Likewise.
31159         (vect_grouped_load_supported): Likewise.
31160         (vect_permute_load_chain): Likewise.
31162 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31164         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31165         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31166         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31167         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31168         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31169         (vect_gen_perm_mask_any): Likewise.
31171 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31173         * int-vector-builder.h: New file.
31174         * vec-perm-indices.h: Include int-vector-builder.h.
31175         (vec_perm_indices): Redefine as an int_vector_builder.
31176         (auto_vec_perm_indices): Delete.
31177         (vec_perm_builder): Redefine as a stand-alone class.
31178         (vec_perm_indices::vec_perm_indices): New function.
31179         (vec_perm_indices::clamp): Likewise.
31180         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31181         (vec_perm_indices::new_vector): New function.
31182         (vec_perm_indices::new_expanded_vector): Update for new
31183         vec_perm_indices class.
31184         (vec_perm_indices::rotate_inputs): New function.
31185         (vec_perm_indices::all_in_range_p): Operate directly on the
31186         encoded form, without computing elided elements.
31187         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31188         encoding.  Update for new vec_perm_indices class.
31189         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31190         the given vec_perm_builder.
31191         (expand_vec_perm_var): Update vec_perm_builder constructor.
31192         (expand_mult_highpart): Use vec_perm_builder instead of
31193         auto_vec_perm_indices.
31194         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31195         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
31196         or double series encoding as appropriate.
31197         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31198         vec_perm_indices instead of auto_vec_perm_indices.
31199         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31200         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31201         (vect_permute_store_chain): Likewise.
31202         (vect_grouped_load_supported): Likewise.
31203         (vect_permute_load_chain): Likewise.
31204         (vect_shift_permute_load_chain): Likewise.
31205         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31206         (vect_transform_slp_perm_load): Likewise.
31207         (vect_schedule_slp_instance): Likewise.
31208         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31209         (vectorizable_mask_load_store): Likewise.
31210         (vectorizable_bswap): Likewise.
31211         (vectorizable_store): Likewise.
31212         (vectorizable_load): Likewise.
31213         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31214         vec_perm_indices instead of auto_vec_perm_indices.  Use
31215         tree_to_vec_perm_builder to read the vector from a tree.
31216         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31217         vec_perm_builder instead of a vec_perm_indices.
31218         (have_whole_vector_shift): Use vec_perm_builder and
31219         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
31220         truncation to calc_vec_perm_mask_for_shift.
31221         (vect_create_epilog_for_reduction): Likewise.
31222         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31223         from auto_vec_perm_indices to vec_perm_indices.
31224         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31225         instead of changing individual elements.
31226         (aarch64_vectorize_vec_perm_const): Use new_vector to install
31227         the vector in d.perm.
31228         * config/arm/arm.c (expand_vec_perm_d::perm): Change
31229         from auto_vec_perm_indices to vec_perm_indices.
31230         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31231         instead of changing individual elements.
31232         (arm_vectorize_vec_perm_const): Use new_vector to install
31233         the vector in d.perm.
31234         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31235         Update vec_perm_builder constructor.
31236         (rs6000_expand_interleave): Likewise.
31237         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31238         (rs6000_expand_interleave): Likewise.
31240 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31242         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31243         to qimode could truncate the indices.
31244         * optabs.c (expand_vec_perm_var): Likewise.
31246 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31248         * Makefile.in (OBJS): Add vec-perm-indices.o.
31249         * vec-perm-indices.h: New file.
31250         * vec-perm-indices.c: Likewise.
31251         * target.h (vec_perm_indices): Replace with a forward class
31252         declaration.
31253         (auto_vec_perm_indices): Move to vec-perm-indices.h.
31254         * optabs.h: Include vec-perm-indices.h.
31255         (expand_vec_perm): Delete.
31256         (selector_fits_mode_p, expand_vec_perm_var): Declare.
31257         (expand_vec_perm_const): Declare.
31258         * target.def (vec_perm_const_ok): Replace with...
31259         (vec_perm_const): ...this new hook.
31260         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31261         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31262         * doc/tm.texi: Regenerate.
31263         * optabs.def (vec_perm_const): Delete.
31264         * doc/md.texi (vec_perm_const): Likewise.
31265         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31266         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31267         expand_vec_perm for constant permutation vectors.  Assert that
31268         the mode of variable permutation vectors is the integer equivalent
31269         of the mode that is being permuted.
31270         * optabs-query.h (selector_fits_mode_p): Declare.
31271         * optabs-query.c: Include vec-perm-indices.h.
31272         (selector_fits_mode_p): New function.
31273         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31274         is defined, instead of checking whether the vec_perm_const_optab
31275         exists.  Use targetm.vectorize.vec_perm_const instead of
31276         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
31277         fit in the vector mode before using a variable permute.
31278         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31279         vec_perm_indices instead of an rtx.
31280         (expand_vec_perm): Replace with...
31281         (expand_vec_perm_const): ...this new function.  Take the selector
31282         as a vec_perm_indices rather than an rtx.  Also take the mode of
31283         the selector.  Update call to shift_amt_for_vec_perm_mask.
31284         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31285         Use vec_perm_indices::new_expanded_vector to expand the original
31286         selector into bytes.  Check whether the indices fit in the vector
31287         mode before using a variable permute.
31288         (expand_vec_perm_var): Make global.
31289         (expand_mult_highpart): Use expand_vec_perm_const.
31290         * fold-const.c: Includes vec-perm-indices.h.
31291         * tree-ssa-forwprop.c: Likewise.
31292         * tree-vect-data-refs.c: Likewise.
31293         * tree-vect-generic.c: Likewise.
31294         * tree-vect-loop.c: Likewise.
31295         * tree-vect-slp.c: Likewise.
31296         * tree-vect-stmts.c: Likewise.
31297         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31298         Delete.
31299         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31300         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31301         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31302         (aarch64_vectorize_vec_perm_const): ...this new function.
31303         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31304         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31305         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31306         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31307         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31308         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31309         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31310         into...
31311         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
31312         check for NEON modes.
31313         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
31314         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
31315         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
31316         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
31317         into...
31318         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
31319         the old VEC_PERM_CONST conditions.
31320         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
31321         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
31322         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
31323         (ia64_vectorize_vec_perm_const_ok): Merge into...
31324         (ia64_vectorize_vec_perm_const): ...this new function.
31325         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
31326         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
31327         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
31328         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
31329         * config/mips/mips.c (mips_expand_vec_perm_const)
31330         (mips_vectorize_vec_perm_const_ok): Merge into...
31331         (mips_vectorize_vec_perm_const): ...this new function.
31332         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
31333         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
31334         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
31335         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
31336         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
31337         (rs6000_expand_vec_perm_const): Delete.
31338         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
31339         Delete.
31340         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31341         (altivec_expand_vec_perm_const_le): Take each operand individually.
31342         Operate on constant selectors rather than rtxes.
31343         (altivec_expand_vec_perm_const): Likewise.  Update call to
31344         altivec_expand_vec_perm_const_le.
31345         (rs6000_expand_vec_perm_const): Delete.
31346         (rs6000_vectorize_vec_perm_const_ok): Delete.
31347         (rs6000_vectorize_vec_perm_const): New function.
31348         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31349         an element count and rtx array.
31350         (rs6000_expand_extract_even): Update call accordingly.
31351         (rs6000_expand_interleave): Likewise.
31352         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
31353         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
31354         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
31355         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
31356         (rs6000_expand_vec_perm_const): Delete.
31357         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31358         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31359         (altivec_expand_vec_perm_const_le): Take each operand individually.
31360         Operate on constant selectors rather than rtxes.
31361         (altivec_expand_vec_perm_const): Likewise.  Update call to
31362         altivec_expand_vec_perm_const_le.
31363         (rs6000_expand_vec_perm_const): Delete.
31364         (rs6000_vectorize_vec_perm_const_ok): Delete.
31365         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
31366         reference to the SPE evmerge intructions.
31367         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31368         an element count and rtx array.
31369         (rs6000_expand_extract_even): Update call accordingly.
31370         (rs6000_expand_interleave): Likewise.
31371         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
31372         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
31373         new function.
31374         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31376 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31378         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
31379         vector mode and that that mode matches the mode of the data
31380         being permuted.
31381         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
31382         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
31383         directly using expand_vec_perm_1 when forcing selectors into
31384         registers.
31385         (expand_vec_perm_var): New function, split out from expand_vec_perm.
31387 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31389         * optabs-query.h (can_vec_perm_p): Delete.
31390         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
31391         * optabs-query.c (can_vec_perm_p): Split into...
31392         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
31393         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
31394         particular selector is valid.
31395         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31396         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31397         (vect_grouped_load_supported): Likewise.
31398         (vect_shift_permute_load_chain): Likewise.
31399         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31400         (vect_transform_slp_perm_load): Likewise.
31401         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31402         (vectorizable_bswap): Likewise.
31403         (vect_gen_perm_mask_checked): Likewise.
31404         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
31405         implementations of variable permutation vectors into account
31406         when deciding which selector to use.
31407         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
31408         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
31409         with a false third argument.
31410         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
31411         to test whether the constant selector is valid and can_vec_perm_var_p
31412         to test whether a variable selector is valid.
31414 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31416         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
31417         * optabs-query.c (can_vec_perm_p): Likewise.
31418         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
31419         instead of vec_perm_indices.
31420         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
31421         (vect_gen_perm_mask_checked): Likewise,
31422         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
31423         (vect_gen_perm_mask_checked): Likewise,
31425 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31427         * optabs-query.h (qimode_for_vec_perm): Declare.
31428         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
31429         (qimode_for_vec_perm): ...this new function.
31430         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
31432 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31434         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
31435         does not have a conditional at the top.
31437 2018-01-02  Richard Biener  <rguenther@suse.de>
31439         * ipa-inline.c (big_speedup_p): Fix expression.
31441 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31443         PR target/81616
31444         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
31445         for generic 4->6.
31447 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31449         PR target/81616
31450         Generic tuning.
31451         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
31452         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
31453         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
31454         cond_taken_branch_cost 3->4.
31456 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
31458         PR tree-optimization/83581
31459         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
31460         TODO_cleanup_cfg if any changes have been made.
31462         PR middle-end/83608
31463         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
31464         convert_modes if target mode has the right side, but different mode
31465         class.
31467         PR middle-end/83609
31468         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
31469         last argument when extracting from CONCAT.  If either from_real or
31470         from_imag is NULL, use expansion through memory.  If result is not
31471         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
31472         the parts directly to inner mode, if even that fails, use expansion
31473         through memory.
31475         PR middle-end/83623
31476         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
31477         check for bswap in mode rather than HImode and use that in expand_unop
31478         too.
31480 Copyright (C) 2018 Free Software Foundation, Inc.
31482 Copying and distribution of this file, with or without modification,
31483 are permitted in any medium without royalty provided the copyright
31484 notice and this notice are preserved.