config.gcc (powerpc-ibm-aix6.*): Delete extra_headers.
[official-gcc.git] / gcc / ChangeLog
blob12270855717f2cb5f040c93c88781f4f0517a681
1 2018-12-17  David Edelsohn  <dje.gcc@gmail.com>
3         * config.gcc (powerpc-ibm-aix6.*): Delete extra_headers.
4         (powerpc-ibm-aix7.1.*): Same.
5         (powerpc-ibm-aix[789].*): Same.
7 2018-12-17  H.J. Lu  <hongjiu.lu@intel.com>
9         PR debug/79342
10         * dwarf2out.c (find_AT_string_in_table): Add insert argument
11         defaulting to INSERT and replace INSERT.
12         (find_AT_string): Likewise.
13         (output_macinfo_op): Pass NO_INSERT to find_AT_string.
15 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
17         * coverage.c (struct conts_entry): Add n_counts.
18         (remap_counts_file): Record number of ocunts.
19         (get_coverage_counts): Verify that counts match.
20         * coverage.h (get_coverage_counts): Update prototype.
21         * profile.c (get_exec_counts. compute_value_histograms): Add
22         n_counts parametrs.
24 2018-12-17  Senthil Kumar Selvaraj  <senthilkumar.selvaraj@microchip.com>
26         PR rtl-optimization/88253
27         * combine.c (combine_simplify_rtx): Test for side-effects before
28         substituting by zero.
30 2018-12-17  Richard Sandiford  <richard.sandiford@arm.com>
32         * doc/invoke.texi (-fversion-loops-for-strides): Document
33         (loop-versioning-group-size, loop-versioning-max-inner-insns)
34         (loop-versioning-max-outer-insns): Document new --params.
35         * Makefile.in (OBJS): Add gimple-loop-versioning.o.
36         * common.opt (fversion-loops-for-strides): New option.
37         * opts.c (default_options_table): Enable fversion-loops-for-strides
38         at -O3.
39         * params.def (PARAM_LOOP_VERSIONING_GROUP_SIZE)
40         (PARAM_LOOP_VERSIONING_MAX_INNER_INSNS)
41         (PARAM_LOOP_VERSIONING_MAX_OUTER_INSNS): New parameters.
42         * passes.def: Add pass_loop_versioning.
43         * timevar.def (TV_LOOP_VERSIONING): New time variable.
44         * tree-ssa-propagate.h
45         (substitute_and_fold_engine::substitute_and_fold): Add an optional
46         block parameter.
47         * tree-ssa-propagate.c
48         (substitute_and_fold_engine::substitute_and_fold): Likewise.
49         When passed, only walk blocks dominated by that block.
50         * tree-vrp.h (range_includes_p): Declare.
51         (range_includes_zero_p): Turn into an inline wrapper around
52         range_includes_p.
53         * tree-vrp.c (range_includes_p): New function, generalizing...
54         (range_includes_zero_p): ...this.
55         * tree-pass.h (make_pass_loop_versioning): Declare.
56         * gimple-loop-versioning.cc: New file.
58 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
60         * ipa-fnsummary.c (remap_edge_change_prob): Do not ICE when changes
61         are not streamed in.
63 2018-12-15  H.J. Lu  <hongjiu.lu@intel.com>
65         PR target/88483
66         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Revert
67         reversion 267133.
69 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
71         * ipa-fnsummary.c (analyze_function_body): Do not loeak conds and
72         size_time_table.
73         (ipa_fn_summary_generate): Add prevails parameter; do not allocate
74         data when symbol is not prevailing.
75         (inline_read_section): Likewise.
77 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
79         * cgraph.h (cgraph_node): Add predicate prevailing_p.
80         (cgraph_edge): Add predicate possible_call_in_translation_unit_p.
81         * ipa-prop.c (ipa_write_jump_function): Optimize streaming of ADDR_EXPR.
82         (ipa_read_jump_function): Add prevails parameter; optimize streaming.
83         (ipa_read_edge_info): Break out from ...
84         (ipa_read_node_info): ... here; optimize streaming.
85         * cgraph.c (cgraph_edge::possibly_call_in_translation_unit_p): New
86         predicate.
88 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
90         * ipa-utils.c (ipa_merge_profiles): Do no merging when source function
91         has zero count.
93 2018-12-15  Jakub Jelinek  <jakub@redhat.com>
95         PR tree-optimization/88464
96         PR target/88498
97         * tree-vect-stmts.c (vect_build_gather_load_calls): For NARROWING
98         and mask with integral masktype, don't try to permute mask vectors,
99         instead emit VEC_UNPACK_{LO,HI}_EXPR.  Fix up NOP_EXPR operand.
100         (vectorizable_store): Handle masked scatters with decl and integral
101         mask type.
102         (permute_vec_elements): Allow scalar_dest to be NULL.
103         * config/i386/i386.c (ix86_get_builtin)
104         <case IX86_BUILTIN_GATHER3ALTDIV16SF>: Use lowpart_subreg for masks.
105         <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Don't assume mask and src have
106         to be the same.
108 2018-12-15  Jan Hubicka  <hubicka@ucw.cz>
110         * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
111         parameters.
112         (cgraph_build_static_cdtor): Update.
113         (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
114         callsed.
116 2018-12-15  Alan Modra  <amodra@gmail.com>
118         * config/rs6000/rs6000.c (generate_set_vrsave, rs6000_emit_savres_rtx),
119         (rs6000_emit_prologue, rs6000_call_aix, rs6000_call_sysv),
120         (rs6000_call_darwin_1): Use gen_hard_reg_clobber.
122 2018-12-15  Jakub Jelinek  <jakub@redhat.com>
124         PR target/88489
125         * config/i386/sse.md (UNSPEC_SFIXUPIMM): New unspec enumerator.
126         (avx512f_sfixupimm<mode><mask_name><round_saeonly_name>): Use it
127         instead of UNSPEC_FIXUPIMM.
129         PR rtl-optimization/88478
130         * cselib.c (cselib_record_sets): Move sets[i].src_elt tests
131         after REG_P (dest) test.
133 2018-12-14  Martin Sebor  <msebor@redhat.com>
135         PR tree-optimization/88372
136         * calls.c (maybe_warn_alloc_args_overflow): Handle function pointers.
137         * tree-object-size.c (alloc_object_size): Same.  Simplify.
138         * doc/extend.texi (Object Size Checking): Update.
139         (Other Builtins): Add __builtin_object_size.
140         (Common Type Attributes): Add alloc_size.
141         (Common Variable Attributes): Ditto.
143 2018-12-14  Martin Sebor  <msebor@redhat.com>
145         PR rtl-optimization/87096
146         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid
147         folding calls whose bound may exceed INT_MAX.  Diagnose bound ranges
148         that exceed the limit.
150 2018-12-14  Martin Sebor  <msebor@redhat.com>
152         PR web/79738
153         * doc/extend.texi (attribute const, pure): Clarify.
155 2018-12-14  H.J. Lu  <hongjiu.lu@intel.com>
157         * config/i386/i386.c (rest_of_insert_endbranch): Insert ENDBR
158         at the function entry only when -mmanual-endbr isn't used or
159         there is cf_check function attribute.
160         (ix86_attribute_table): Add cf_check.
161         * config/i386/i386.opt: Add -mmanual-endbr.
162         * doc/extend.texi: Document cf_check attribute.
163         * doc/invoke.texi: Document -mmanual-endbr.
165 2018-12-14  Thomas Schwinge  <thomas@codesourcery.com>
166             Cesar Philippidis  <cesar@codesourcery.com>
168         * omp-offload.c (inform_oacc_loop): New function.
169         (execute_oacc_device_lower): Use it to display loop parallelism.
171 2018-12-14  Jakub Jelinek  <jakub@redhat.com>
173         PR c++/82294
174         PR c++/87436
175         * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
176         * expr.c (categorize_ctor_elements_1): Likewise.  Compute it like
177         p_nz_elts, except don't multiply it by mult.  Adjust recursive call.
178         Fix up COMPLEX_CST handling.
179         (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
180         it and pass it through to categorize_ctor_elements_1.
181         (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
182         * gimplify.c (gimplify_init_constructor): Likewise.  Don't force
183         ctor into readonly data section if num_unique_nonzero_elements is
184         smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
185         bytes.
187 2018-12-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
189         * config/arm/arm.md (arm_movdi): Split if -mslow-flash-data and
190         source is a constant that would be loaded by literal pool.
191         (movsf expander): Generate a no_literal_pool_sf_immediate insn if
192         -mslow-flash-data is present, targeting hardfloat ABI and source is a
193         float constant that cannot be loaded via vmov.
194         (movdf expander): Likewise but generate a no_literal_pool_df_immediate
195         insn.
196         (arm_movsf_soft_insn): Split if -mslow-flash-data and source is a
197         float constant that would be loaded by literal pool.
198         (softfloat constant movsf splitter): Splitter for the above case.
199         (movdf_soft_insn): Split if -mslow-flash-data and source is a float
200         constant that would be loaded by literal pool.
201         (softfloat constant movdf splitter): Splitter for the above case.
202         * config/arm/constraints.md (Pz): Document existing constraint.
203         (Ha): Define constraint.
204         (Tu): Likewise.
205         * config/arm/predicates.md (hard_sf_operand): New predicate.
206         (hard_df_operand): Likewise.
207         * config/arm/thumb2.md (thumb2_movsi_insn): Split if
208         -mslow-flash-data and constant would be loaded by literal pool.
209         * constant/arm/vfp.md (thumb2_movsi_vfp): Likewise and disable constant
210         load in VFP register.
211         (movdi_vfp): Likewise.
212         (thumb2_movsf_vfp): Use hard_sf_operand as predicate for source to
213         prevent match for a constant load if -mslow-flash-data and constant
214         cannot be loaded via vmov.  Adapt constraint accordingly by
215         using Ha instead of E for generic floating-point constant load.
216         (thumb2_movdf_vfp): Likewise using hard_df_operand predicate instead.
217         (no_literal_pool_df_immediate): Add a clobber to use as the
218         intermediate general purpose register and also enable it after reload
219         but disable it constant is a valid FP constant.  Add constraints and
220         generate a DI intermediate load rather than 2 SI loads.
221         (no_literal_pool_sf_immediate): Add a clobber to use as the
222         intermediate general purpose register and also enable it after
223         reload.
225 2018-12-14  Uros Bizjak  <ubizjak@gmail.com>
227         PR target/88474
228         * internal-fn.def (HYPOT): New.
229         * optabs.def (hypot_optab): New.
230         * config/i386/i386.md (hypot<mode>3): New expander.
232 2018-12-14  Jeff Law  <law@redhat.com>
234         * target.def (post_cfi_startproc): Fix text.
236 2018-12-14  Sam Tebbs  <sam.tebbs@arm.com>
238         * doc/tm.texi (TARGET_ASM_POST_CFI_STARTPROC): Define.
239         * doc/tm.texi.in (TARGET_ASM_POST_CFI_STARTPROC): Define.
240         * dwarf2out.c (dwarf2out_do_cfi_startproc): Trigger the hook.
241         * hooks.c (hook_void_FILEptr_tree): Define.
242         * hooks.h (hook_void_FILEptr_tree): Define.
243         * target.def (post_cfi_startproc): Define.
245 2018-12-14  Tom de Vries  <tdevries@suse.de>
247         * lto-cgraph.c (verify_node_partition): New function.
248         (input_overwrite_node, input_varpool_node): Use verify_node_partition.
250 2018-12-14  H.J. Lu  <hongjiu.lu@intel.com>
252         PR target/88483
253         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
254         use get_frame_size ().
256 2018-12-14  Andrew Stubbs  <ams@codesourcery.com>
258         * gcc/lra-int.h (lra_register_new_scratch_op): Add third parameter.
259         * gcc/lra-remat.c (update_scratch_ops): Pass icode to
260         lra_register_new_scratch_op.
261         * gcc/lra.c (struct sloc): Add icode field.
262         (lra_register_new_scratch_op): Add icode parameter.
263         Use icode to skip insns that have changed beyond recognition.
265 2018-12-14  Eric Botcazou  <ebotcazou@adacore.com>
267         * dwarf2out.c (analyze_discr_in_predicate): Simplify.
268         (analyze_variants_discr): Deal with naked boolean discriminants.
270 2018-12-14  Claudiu Zissulescu  <claziss@synopsys.com>
272         * config/arc/arc.c (arc_restore_callee_milli) Don't clobber off
273         variable.
275 2018-12-14  Claudiu Zissulescu  <claziss@synopsys.com>
277         * config/arc/arc.h (reg_class): Add SIBCALL_REGS.
279 2018-12-14  Richard Biener  <rguenther@suse.de>
281         * gimple-ssa-split-paths.c (is_feasible_trace): Remove
282         duplicated code block.
284 2018-12-14  Alan Modra  <amodra@gmail.com>
286         PR rtl-optimization/88311
287         * config/rs6000/rs6000.c (rs6000_call_sysv): Do not mask cookie.
288         (rs6000_sibcall_sysv): Likewise.
290 2018-12-14  Segher Boessenkool  <segher@kernel.crashing.org>
292         PR rtl-optimization/88001
293         * function.c (match_asm_constraints_1): Don't invalidly share RTL.
295 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
297         * auto-profile.c (afdo_annotate_cfg): Call update_max_bb_count even
298         if autofdo counts are all zeros.
300 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
302         * auto-profile.c (afdo_indirect_call): Skip generating histogram
303         value for internal call.
305 2018-12-14  Bin Cheng  <bin.cheng@linux.alibaba.com>
307         * auto-profile.c (AFDO_EINFO): New macro.
308         (class edge_info): New class.
309         (is_edge_annotated, set_edge_annotated): Delete.
310         (afdo_propagate_edge, afdo_propagate_circuit, afdo_propagate): Remove
311         parameter.  Adjust edge count computation and annotation using class
312         edge_info.
313         (afdo_calculate_branch_prob, afdo_annotate_cfg): Likewise.
315 2018-12-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
317         * ipa-cp.c (print_all_lattices): Skip cp clones.
319         * ipa-cp.c: Fix various comment typos.
321 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
323         PR tree-optimization/88444
324         * tree-inline.c (fold_marked_statements): Iterate up to
325         last_basic_block_for_fn rather than n_basic_blocks_for_fn.
327         PR rtl-optimization/88470
328         * cfgcleanup.c (outgoing_edges_match): If the function is
329         shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
330         edge to EXIT, return false.
332 2018-12-13  Martin Sebor  <msebor@redhat.com>
334         * doc/extend.texi: Consistently use @code for const and volatile
335         qualifiers, the true and false constants, and asm statements.
337 2018-12-13  Vladimir Makarov  <vmakarov@redhat.com>
339         PR rtl-optimization/88414
340         * lra-int.h (lra_asm_error_p): New.
341         * lra-assigns.c (lra_assign): Check lra_asm_error_p for checking
342         call crossed pseudo assignment correctness.
343         (lra_split_hard_reg_for): Set up lra_asm_error_p.
344         * lra-constraints.c (curr_insn_transform): Ditto.
345         * lra.c (lra_asm_error_p): New.
346         (lra): Reset lra_asm_error_p.
348 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
350         PR rtl-optimization/88416
351         * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
352         even if !AUTO_INC_DEC.
354 2018-12-13  David Edelsohn  <dje.gcc@gmail.com>
356         * config/rs6000/rs6000.c (rs6000_function_arg): Ensure type is
357         non-NULL.
358         (rs6000_arg_partial_bytes): Same.
360 2018-12-13  Jason Merrill  <jason@redhat.com>
362         * gdbinit.in (pp): New macro.
363         (pbb): Remove.
365 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
367         PR tree-optimization/88464
368         * tree-vect-stmts.c (vect_build_gather_load_calls): Handle INTEGER_TYPE
369         masktype if mask is non-NULL.
370         (vectorizable_load): Don't reject masked gather loads if masktype
371         in the decl is INTEGER_TYPE.
373 2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>
375         * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
376         s390_vec_double_s64 instead of s390_vcdgb.
377         (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
378         s390_vcdlgb.
380 2018-12-13  Andreas Krebbel  <krebbel@linux.ibm.com>
382         * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
383         ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
384         ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
386 2018-12-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
388         * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
389         behavior.
391 2018-12-13  Wei Xiao  <wei3.xiao@intel.com>
393         * common/config/i386/i386-common.c (processor_names): Add cascadelake.
394         (processor_alias_table): Add cascadelake.
395         * config.gcc: Add -march=cascadelake.
396         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
397         cascadelake.
398         * config/i386/i386.c (Add m_CASCADELAKE): New.
399         (processor_cost_table): Add cascadelake.
400         (get_builtin_code_for_version): Handle cascadelake.
401         * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New.
402         (PTA_CASCADELAKE): Ditto.
403         * doc/invoke.texi: Add -march=cascadelake.
405 2018-12-13  Jakub Jelinek  <jakub@redhat.com>
407         PR target/88465
408         * config/i386/i386.md (*movdi_internal, *movsi_internal,
409         *movhi_internal, *movqi_internal): Add alternative(s) to load
410         0 or -1 into k registers using kxor or kxnoq instructions.
412         PR target/88461
413         * config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New
414         mode iterators.
415         (<avx512>_testm<mode>3<mask_scalar_merge_name>,
416         <avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns
417         with VI12_AVX512VL and VI48_AVX512VL iterators into ones with
418         VI1248_AVX512VLBW iterator.
419         (*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext,
420         *<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask,
421         *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext,
422         *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New
423         define_insns.
425         PR target/88461
426         * config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2,
427         *zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives.
429 2018-12-12  Tom de Vries  <tdevries@suse.de>
430             Jakub Jelinek  <jakub@redhat.com>
432         * omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
433         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
434         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
435         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
436         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
437         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up
438         function types - remove one argument.
440 2018-12-12  Martin Sebor  <msebor@redhat.com>
442         * doc/extend.texi (attribute copy): Fix typos.
444 2018-12-12  Martin Sebor  <msebor@redhat.com>
446         * doc/extend.texi (Function Attributes): Clarify C++ aspects.
447         (Variable Attributes): Same.
448         (Type Attributes): Same.
450 2018-12-12  Olivier Hainque  <hainque@adacore.com>
452         * config/aarch64/aarch64.c (aarch64_override_options): Once arch,
453         cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if
454         defined.
456 2018-12-12  Olivier Hainque  <hainque@adacore.com>
458         * config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as
459         R9_REGNUM instead of 9.
460         (PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10.
462 2018-12-12  Eric Botcazou  <ebotcazou@adacore.com>
464         PR target/86806
465         * config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
466         (speculation_barrier): New instruction for V9.
468 2018-12-12 Segher Boessenkool  <segher@kernel.crashing.org>
469            Iain Sandoe  <iain@sandoe.co.uk>
471         PR target/88343
472         * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
473         unless it has been used.
474         (first_reg_to_save): Remove dead code.
476 2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>
478         * config/rs6000/rs6000-protos.h (rs6000_call_darwin): New.
479         (rs6000_sibcall_darwin): New. (macho_call_template): Remove.
480         * config/rs6000/rs6000.c (get_prev_label): Forward declaration. 
481         (rs6000_call_template_1): Handle Darwin.
482         (macho_call_template): Remove.
483         (rs6000_call_sysv): Remove handling for Darwin.
484         (rs6000_call_darwin_1, rs6000_call_darwin): New
485         (rs6000_sibcall_darwin): New.
486         * config/rs6000/rs6000.md (define_expand “call”): Handle Darwin
487         with its own expander. (define_expand “call_value”): Likewise.
488         (define_expand “sibcall”): Likewise.
489         (define_expand “sibcall_value”): Likewise.
490         (call_nonlocal_sysv): Remove Darwin special-casing.
491         (call_value_nonlocal_sysv): Likewise.
493 2018-12-12  Iain Sandoe  <iain@sandoe.co.uk>
495         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove.
496         (call_nonlocal_darwin64): Remove.
497         (call_value_indirect_nonlocal_darwin64): Remove.
498         (call_value_nonlocal_darwin64): Remove.
499         * config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with
500         the same asm output as AIX/ELFv2.  (rs6000_call_sysv): Preserve the
501         CALL_LONG flag when needed for Darwin.
502         * config/rs6000/rs6000.md (define expand “call”): Expand Darwin as
503         per sysv.
504         (define_expand “call_value”): Likewise.
505         (define_expand “sibcall”): Likewise.
506         (define_expand “sibcall_value”): Likewise.
507         (call_indirect_nonlocal_sysv): Mark the clobber mode P.
508         (call_nonlocal_sysv): Likewise.
509         (call_nonlocal_sysv_secure): Likewise.
510         (call_value_indirect_nonlocal_sysv): Likewise.
511         (call_value_nonlocal_sysv): Likewise.
512         (call_value_nonlocal_sysv_secure): Likewise.
513         (call_local64): Mark the clobber DI.
514         (call_value_local64): Likewise.
516 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
518         PR tree-optimization/88444
519         * tree-vrp.c (register_edge_assert_for_2): Only register assertions
520         for conversions if rhs1 is a SSA_NAME.
522 2018-12-11  Dimitar Dimitrov  <dimitar@dinux.eu>
524         * cfgexpand.c (asm_clobber_reg_is_valid): Also produce
525         error when stack pointer is clobbered.
526         (expand_asm_stmt): Refactor clobber check in separate function.
528 2018-12-11  Eric Botcazou  <ebotcazou@adacore.com>
530         * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
531         accounting for STACK_BOUNDARY 128.
532         (STACK_DYNAMIC_OFFSET): Likewise.
534 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
536         PR target/88425
537         * config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>):
538         New define_insn_and_split.
540 2018-12-11  Richard Biener  <rguenther@suse.de>
542         PR middle-end/88448
543         PR middle-end/88415
544         * gimple.c (gimple_assign_set_rhs_with_ops): Revert previous
545         change.
546         * tree-complex.c (update_complex_assignment): Properly transfer
547         or clean EH info around gimple_assign_set_rhs_with_ops.
549 2018-12-11  Richard Earnshaw  <rearnsha@arm.com>
551         PR target/87369
552         * config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode.
553         (Vbtype): Add SFmode mapping.
554         * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
555         (copysign<GPF:mode>3): New expand pattern.
556         (copysign<GPF:mode>3_insn): New insn pattern.
558 2018-12-11  Richard Biener  <rguenther@suse.de>
560         * ccmp.c (ccmp_candidate_p): Use GIMPLE API properly.
562 2018-12-11  Jakub Jelinek  <jakub@redhat.com>
564         PR lto/86004
565         * doc/sourcebuild.texi (lto_incremental): Document new effective
566         target.
568 2018-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
570         PR target/88145
571         * config/rs6000/rs6000.c (rs6000_expand_zeroop_builtin): Use
572         rs6000_isa_flags instead of rs6000_isa_flags_explicit to decide
573         whether soft float is enabled.
574         (rs6000_expand_mtfsb_builtin): Ditto.
575         (rs6000_expand_set_fpscr_rn_builtin): Ditto.
576         (rs6000_expand_set_fpscr_drn_builtin): Ditto.
578 2018-12-10  Jeff Law  <law@redhat.com>
580         PR tree-optimization/80520
581         * gimple-ssa-split-paths.c (is_feasible_trace): Recognize half
582         diamonds that are likely if convertable.
584 2018-12-10  Martin Sebor  <msebor@redhat.com>
586         PR tree-optimization/86196
587         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
588         base size only of arrays.
590 2018-12-10  Segher Boessenkool  <segher@kernel.crashing.org>
592         * config.gcc (Obsolete configurations): Delete powerpc*-*-*spe*.
593         (Unsupported targets): Add powerpc*-*-*spe*.
594         (powerpc*-*-*spe*): Delete.
595         (powerpc-*-eabispe*): Delete.
596         (powerpc-*-rtems*spe*): Delete.
597         (powerpc*-*-linux*spe*): Delete.
598         (powerpc*-*-linux*): Do not handle the linux*spe* targets.
599         (powerpc-wrs-vxworks*spe): Delete.
600         (with_cpu setting code): Delete powerpc*-*-*spe* handling.
601         * config.host (target powerpc*-*-*spe*): Delete.
602         * doc/invoke.texi (PowerPC SPE Options): Delete.
603         (PowerPC SPE Options): Delete.
604         * config/powerpcspe: Delete.
606 2018-12-10  Uros Bizjak  <ubizjak@gmail.com>
608         PR target/88418
609         * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
610         check operand 1 with vector_operand predicate.
611         (ix86_expand_sse_movcc): For vector modes, check op_true with
612         vector_operand, not nonimmediate_operand.
614 2018-12-10  Richard Biener  <rguenther@suse.de>
616         * tree-affine.c (tree_to_aff_combination): Remove unreachable
617         MEM_REF case.
618         (aff_combination_expand): Cache on SSA names, not possibly
619         on conversion trees.  Avoid expanding cases we do not handle.
621 2018-12-10  Richard Biener  <rguenther@suse.de>
623         PR tree-optimization/88427
624         * vr-values.c (vr_values::extract_range_from_phi_node):
625         Handle symbolic ranges conservatively when trying to drop
626         to Inf +- 1.
628 2018-12-10  Martin Jambor  <mjambor@suse.cz>
630         PR ipa/88214
631         * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
632         we check pointers against pointers.
634 2018-12-10  Richard Biener  <rguenther@suse.de>
636         PR middle-end/88415
637         * gimple.c (gimple_assign_set_rhs_with_ops): Transfer EH
638         info to a newly allocated stmt.
640 2018-12-10  Jerome Lambourg  <lambourg@adacore.com>
642         * config/vxworksae.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): Define.
644 2018-12-10  Fredrik Nyström  <fredrik@lysator.liu.se>
646         PR bootstrap/65725
647         * config/sol2.h: Only use libgcc-unwind.map if
648         ENABLE_SHARED_LIBGCC.
650 2018-12-10  Bin Cheng  <bin.cheng@linux.alibaba.com>
652         * auto-profile.c (afdo_calculate_branch_prob): Convert profile_count
653         to afdo precision quality.
654         (afdo_callsite_hot_enough_for_early_inline): Likewise.
655         (afdo_annotate_cfg): Set profile_count for exit basic block.
657 2018-12-10  Bin Cheng  <bin.cheng@linux.alibaba.com>
659         * auto-profile.c (string_table::get_index_by_decl)
660         (function_instance::get_function_instance_by_decl): Avoid infinite
661         recursion by using DECL_FROM_INLINE.
663 2018-12-09  John David Anglin  <danglin@gcc.gnu.org>
665         * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file.  Define
666         d_target_objs.
667         * config/pa/pa-protos.h (pa_d_target_versions): New prototype.
668         * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define.
669         * config/pa/pa-d.c: New file.
670         * config/pa/t-pa: New file.
672 2018-12-08  Bin Cheng  <bin.cheng@linux.alibaba.com>
674         * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for
675         zero profile count.
677 2018-12-08  Jakub Jelinek  <jakub@redhat.com>
679         PR fortran/88304
680         * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
681         lhs is not a decl normally, don't call use_pointer_in_frame on that
682         lhs.
684 2018-12-08  Eric Botcazou  <ebotcazou@adacore.com>
686         PR rtl-optimization/88390
687         * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
688         (cfi_row_equal_p): Compare it.
689         (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
690         If FAKE is false, set window_save of the current row.
691         (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
692         function with FAKE parameter set to true.
693         <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
694         (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
696 2018-12-07  Peter Bergner  <bergner@linux.ibm.com>
698         PR target/87496
699         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
700         -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
701         Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
702         * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
703         require -mlong-double-128.
705 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
707         * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
708         sve_width entry.
709         (aarch64_parse_sve_width_string): Define.
711 2018-12-07  Jeff Law  <law@redhat.com>
713         PR middle-end/87813
714         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
715         m_update_global_ranges member.  Add corresponding argument to ctor.
716         * gimple-ssa-evrp-analyze.c
717         (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
718         initialize m_update_global_ranges.
719         (evrp_range_analyzer::set_ssa_range_info): Assert that we are
720         updating global ranges.
721         (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
722         update global ranges if explicitly requested.
723         (evrp_range_analyzer::record_ranges_from_phis): Similarly.
724         (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
725         * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
726         evrp_range_analyzer ctor.
727         * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
728         * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
730 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
732         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
733         Add SVE_NOT_IMPLEMENTED value.
734         * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
735         field.
736         * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
737         cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
738         cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
739         thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
740         saphira_tunings, thunderx2t99_tunings, emag_tunings):
741         Specify sve_width.
742         (aarch64_estimated_poly_value): Define.
743         (TARGET_ESTIMATED_POLY_VALUE): Define.
745 2018-12-07  Paul A. Clarke  <pc@us.ibm.com>
747         PR target/88408
748         * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
750 2018-12-07  Vladimir Makarov  <vmakarov@redhat.com>
752         PR rtl-optimization/88349
753         * ira-costs.c (record_operand_costs): Check bigger reg class on
754         NO_REGS.
756 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
758         * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
759         Split the patterns after reload if we don't need the predicate
760         operand.
761         (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
763 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
765         * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
766         iterator.
767         (sve_fp_op): Handle minus and mult.
768         * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
769         (*mul<mode>3): Split the patterns after reload if we don't
770         need the predicate operand.
771         (*post_ra_<sve_fp_op><mode>3): New pattern.
773 2018-12-07  Bin Cheng  <bin.cheng@linux.alibaba.com>
775         * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
776         return condition when *this is precise zero.
778 2018-12-07  Jakub Jelinek  <jakub@redhat.com>
780         PR target/85593
781         * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
782         for functions with naked attribute.
784         PR c/88367
785         * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
786         with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
787         is non-NULL and offset is known to have most significant bit clear.
788         * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
789         of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
790         most significant bit clear.  If offset does have most significant bit
791         set and -fno-delete-null-pointer-checks, don't return true even if
792         the base pointer is non-NULL.
794 2018-12-06  Alexandre Oliva  <aoliva@redhat.com>
796         * cselib.c (cselib_record_sets): Skip strict low part sets
797         with NULL src_elt.
799 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
801         PR target/88316
802         * config/rs6000/smmintrin.h: New file.
803         * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
805 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
807         PR target/88316
808         * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
809         (_mm_unpacklo_pi8): Likewise.
810         (_mm_mulhi_pi16): Likewise.
811         (_mm_packs_pi16): Fix for big-endian. Use preferred API.
812         (_mm_packs_pi32): Likewise.
813         (_mm_packs_pu16): Likewise.
814         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
815         (_mm_cvtss_si64): Likewise.
816         (_mm_cvtpi32x2_ps): Likewise.
817         (_mm_shuffle_ps): Likewise.
818         (_mm_movemask_pi8): Likewise.
819         (_mm_mulhi_pu16): Likewise.
820         (_mm_sad_pu8): Likewise.
821         (_mm_sad_pu8): Likewise.
822         (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
823         (_mm_cvtpu8_ps): Likewise.
824         (_mm_movemask_ps): Better #else case for big-endian (no functional
825         change).
826         (_mm_shuffle_pi16): Likewise.
827         * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
828         Better #else case for big-endian (no functional change).
829         (_mm_movemask_epi8): Likewise.
830         (_mm_shufflehi_epi16): Likewise.
831         (_mm_shufflelo_epi16): Likewise.
832         (_mm_shuffle_epi32): Likewise.
833         (_mm_mul_epu32): Fix for big-endian.
834         (_mm_bsrli_si128): Likewise.
835         (_mm_cvtps_pd): Better #else case for big endian.
836         (_mm_mulhi_epi16): Likewise.
837         (_mm_mul_epu32): Likewise.
838         (_mm_slli_si128): Likewise.
839         (_mm_sll_epi16): Likewise.
840         (_mm_sll_epi32): Likewise.
841         (_mm_sra_epi16): Likewise.
842         (_mm_sra_epi32): Likewise.
843         (_mm_srl_epi16): Likewise.
844         (_mm_srl_epi32): Likewise.
845         (_mm_mulhi_epu16): Likewise.
846         (_mm_sad_epu8): Likewise.
847         * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
848         (_mm_sub_ps): Likewise.
849         * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
850         * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
851         macros consistently (no functional changes).
852         (_mm_alignr_pi8): Likewise.
854 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
856         PR c++/87380
857         * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
858         default.
859         * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
861 2018-12-06  Vladimir Makarov  <vmakarov@redhat.com>
863         PR target/88282
864         * ira.c (ira_init_register_move_cost): Use info from
865         hard_regno_mode_ok instead of contains_reg_of_mode.
866         * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
867         hard register class for some fixed hard registers.
869 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
871         * doc/extend.texi (Using Assembly Language with C): Document asm inline.
872         (Size of an asm): Fix typo.  Document asm inline.
873         * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
874         * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
875         (gimple_asm_set_volatile): Fix typo.
876         (gimple_asm_inline_p): New.
877         (gimple_asm_set_inline): New.
878         * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
879         tree to gimple.
880         * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
881         gimple_asm_inline_p flag, too.
882         * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
883         in an ASM_EXPR.
884         * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
885         a minimum size for an asm.
886         * tree.h (ASM_INLINE_P): New.
888 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
890         PR inline-asm/55681
891         * doc/extend.texi (Basic Asm): Update grammar.
892         (Extended Asm): Update grammar.
894 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
896         PR target/78444
897         * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
898         * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
899         stack alignment in non-leaf functions.
901 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
903         PR target/87598
904         * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
905         call output_operand_lossage on VOIDmode CONST_INTs.  After
906         output_operand_lossage do return false.
908 2018-12-06  Richard Biener  <rguenther@suse.de>
910         * df-problems.c (df_rd_local_compute): Use bitmap_release.
911         (df_live_free): Likewise.
912         (df_md_local_compute): Likewise.
913         (df_md_free): Release df_md_scratch bitmap.
914         * loop-invariant.c (calculate_loop_reg_pressure): Use
915         bitmap_release.
916         * sched-deps.c (true_dependency_cache, output_dependency_cache,
917         anti_dependency_cache, control_dependency_cache,
918         spec_dependency_cache): Use bitmap instead of bitmap_head *.
919         * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
920         dont_calc_deps as bitmap allocated from obstack not GC.
921         (schedule_ebbs_finish): Use bitmap_release.
922         * sched-rgn.c (schedule_insns): Initialize non-GTY
923         not_in_df as bitmap allocated from obstack not GC.
924         Use bitmap_release.
925         * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
926         (sel_region_init): Allocate forced_ebb_heads.
927         (sel_region_finish): Free forced_ebb_heads.
929 2018-12-06  Richard Biener  <rguenther@suse.de>
931         * bitmap.c (bitmap_head::crashme): Define.
932         * bitmap.h (bitmap_head): Add constexpr default constructor
933         poisoning the obstack member.
934         (bitmap_head::crashme): Declare.
935         (bitmap_release): New function clearing a bitmap and poisoning
936         the obstack member.
937         * gengtype.c (main): Make it recognize CONSTEXPR.
938         * lra-constraints.c (lra_inheritance): Use bitmap_release
939         instead of bitmap_clear.
940         * ira.c (ira): Work around class-memaccess warning.
941         * regrename.c (create_new_chain): Likewise.
943 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
945         PR tree-optimization/85726
946         * generic-match-head.c (optimize_successive_divisions_p): New function.
947         * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
948         * match.pd: Don't combine successive divisions if they aren't exact
949         and optimize_successive_divisions_p is false.
951         PR fortran/88304
952         * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
953         for non-local automatic decls.
955 2018-12-05  David Edelsohn  <dje.gcc@gmail.com>
957         * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
958         default. 
960 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
962         * dwarf2out.c (field_byte_offset): Remove
963         #ifdef PCC_BITFIELD_TYPE_MATTERS check.
965 2018-12-05  Richard Biener  <rguenther@suse.de>
966             Martin Sebor  <msebor@redhat.com>
968         PR c/87028
969         * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
970         SSA_NAME_VAR.
971         * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
972         * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
974 2018-12-05  Iain Sandoe  <iain@sandoe.co.uk>
976         * configure.ac (gcc_cv_otool): Set.
977         * configure: Regenerate.
979 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
981         * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
982         (GROUP_BITFIELDS_BY_ALIGN): Ditto.
984 2018-12-05  Richard Sandiford  <richard.sandiford@arm.com>
986         PR tree-optimization/88064
987         * tree-vect-patterns.c (vect_convert_input): Convert the result of
988         an existing cast if it has the right width but the wrong sign.
989         Do not test the signedness of the required result when
990         considering whether to split an existing cast; instead split to
991         a type with the same signedness as the source of the cast, then
992         convert it to the opposite signedness where necessary.
993         (vect_recog_over_widening_pattern): Handle sign changes between
994         the final PLUS_EXPR and the RSHIFT_EXPR.
995         (vect_recog_average_pattern): Use an unsigned operation when
996         truncating an addition, subtraction or multiplication.  Cast the
997         result back to the "real" signedness before promoting.
999 2018-12-05  Richard Biener  <rguenther@suse.de>
1001         PR tree-optimization/86637
1002         * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
1003         vect_location at the end.
1005 2018-12-05  Jakub Jelinek  <jakub@redhat.com>
1007         PR tree-optimization/87360
1008         * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1009         data dependencies, don't return false, just continue.  Formatting
1010         fixes.
1011         (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
1012         fuse_loops): Formatting fixes.
1014         PR sanitizer/88333
1015         * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
1016         aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
1018 2018-12-05  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1020         * config.gcc (csky-*-linux-gnu*): Force .init_array support.
1022 2018-12-05  Segher Boessenkool  <segher@kernel.crashing.org>
1024         * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
1025         aggregate handling on actual AIX, not on somewhat similar systems.
1026         (rs6000_arg_partial_bytes): Ditto.
1028 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
1030         PR rtl-optimization/88317
1031         * lra-constraints.c (split_reg): Don't set up check_only_regs if
1032         we are outside the inheritance pass.
1034 2018-12-04  Jan Hubicka  <hubicka@ucw.cz>
1036         * ipa-prop.c (jump_function_useful_p): New.
1037         (ipa_read_node_info): Do not allocated useless jump functions.
1039 2018-12-04  David Malcolm  <dmalcolm@redhat.com>
1041         PR c/82967
1042         * spellcheck.c (selftest::test_suggestions): Add another assertion.
1044 2018-12-04  Martin Sebor  <msebor@redhat.com>
1046         * doc/extend.texi (__builtin_has_attribute): Fix typo.
1048 2018-12-04  Jakub Jelinek  <jakub@redhat.com>
1050         PR target/88188
1051         * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
1052         failures.
1053         (print_operand): Use REG_P instead of GET_CODE == REG.
1054         <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
1055         <case 't'>: Likewise.  Remove GET_MODE check.
1056         * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
1057         return -1.
1059 2018-12-04  Martin Sebor  <msebor@redhat.com>
1061         * doc/extend.texi (attribute aligned): Expand.
1063 2018-12-04  David Edelsohn  <dje.gcc@gmail.com>
1065         PR target/61976
1066         * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
1067         in FPRs on AIX.
1068         (rs6000_arg_partial_bytes): Same.
1070 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
1072         PR target/88282
1073         * ira-costs.c (exec): Try bigger class to use smaller register
1074         move cost.
1076 2018-12-04  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1078         PR ipa/88297
1079         * ipa-cp.c (create_specialized_node): Track clone counters by
1080         node assembler names.
1081         (ipcp_driver): Change type of clone_num_suffixes key to const char*.
1083 2018-12-04  Richard Biener  <rguenther@suse.de>
1085         * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
1086         for epilogue vectorization.
1088 2018-12-04  Richard Biener  <rguenther@suse.de>
1090         PR tree-optimization/88301
1091         * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
1092         issues in last commit.
1094 2018-12-04  Richard Biener  <rguenther@suse.de>
1096         PR tree-optimization/88315
1097         * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
1098         and fix initialization vector for SAD and DOT_PROD SLP reductions.
1100 2018-12-03  Sandra Loosemore  <sandra@codesourcery.com>
1102         PR c/59039
1103         * doc/extend.texi (Nonlocal gotos): New section.
1105 2018-12-03  Uros Bizjak  <ubizjak@gmail.com>
1107         * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
1108         alternative to (v,vm,c).
1110 2018-12-03  Segher Boessenkool  <segher@kernel.crashing.org>
1112         * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
1113         (powerpc*-*-linux*): Don't handle it here.
1115 2018-12-03  Jeff Law  <law@redhat.com>
1117         * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
1118         warnings quiet.
1120         * config/h8300/h8300.md (low byte masking peepholes): Only accept
1121         nonimmediate_operand for the to-be-masked operand.
1123 2018-12-03  Richard Biener  <rguenther@suse.de>
1125         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1126         allow expected function-pointer cast re-instantiation.
1128 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
1130         * common/config/s390/s390-common.c (s390_option_init_struct):
1131         Use section anchors by default.
1132         * config/s390/s390.c (s390_check_symref_alignment): Handle
1133         anchors.
1134         (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
1135         * output.h (assemble_align): Pass `align' as unsigned int, so
1136         that the value 0x80000000, which corresponds to `aligned(1 <<
1137         28)', would pass the `align > BITS_PER_UNIT' check.
1138         * varasm.c (assemble_align): Likewise.
1140 2018-12-03  Julian Brown  <julian@codesourcery.com>
1142         * tree-pretty-print.c (dump_omp_clause): Make default case
1143         gcc_unreachable.
1145 2018-12-03  Richard Biener  <rguenther@suse.de>
1147         PR tree-optimization/88301
1148         * tree-vrp.c (register_edge_assert_for_2): Handle conversions
1149         that do not change the value by registering the same assert
1150         for the operand.
1152 2018-12-03  Richard Biener  <rguenther@suse.de>
1154         * tree-vect-loop.c (vect_transform_loop): Properly compute
1155         upper bound for the epilogue when doing epilogue vectorization.
1157 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
1159         PR target/80080
1160         * cfgcleanup.c (class pass_postreload_jump): New pass.
1161         (pass_postreload_jump::execute): Likewise.
1162         (make_pass_postreload_jump): Likewise.
1163         * passes.def: Add pass_postreload_jump before
1164         pass_postreload_cse.
1165         * tree-pass.h (make_pass_postreload_jump): New pass.
1167 2018-12-02  Uros Bizjak  <ubizjak@gmail.com>
1169         * config/i386/i386.c (inline_memory_move_cost): Correctly
1170         compute number of TFmode GPR moves.
1172 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
1174         PR target/88278
1175         * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
1177 2018-12-02  Jeff Law  <law@redhat.com>
1179         * config/h8300/h8300.md (call, call_value): Drop mode from
1180         operand containing number of bytes pushed.
1181         (call_insn, call_value_insn): Use mode iterator for number
1182         of bytes pushed operand.  Make pattern name vary appropriately.
1184 2018-12-02  Jan Hubicka  <jh@suse.cz>
1186         * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
1187         trust precise profiles.
1189 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
1191         * tree-nested.c (convert_nonlocal_omp_clauses,
1192         convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
1193         OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
1194         (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
1195         Convert clauses for GIMPLE_OMP_TASKGROUP.
1197         * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
1198         cancellation point with taskgroup clause inside of taskloop.  Consider
1199         a taskloop construct without nogroup clause as implicit taskgroup for
1200         diagnostics if cancel/cancellation point with taskgroup clause is
1201         closely nested inside of taskgroup region.
1203         PR sanitizer/88291
1204         * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
1205         to the start of the function.
1206         (asan_emit_stack_protection): When emitting clearing sequence for
1207         epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
1208         add last_size_aligned which is last_size padded to multiples of
1209         ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
1210         boundaries.
1212 2018-12-01  Jakub Jelinek  <jakub@redhat.com>
1214         PR sanitizer/88289
1215         * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
1216         an off-by-one for BYTES_BIG_ENDIAN.
1218         PR target/54589
1219         * combine.c (find_split_point): For invalid memory address
1220         nonobj + obj + const, if reg + obj + const is valid addressing
1221         mode, split at nonobj.  Use if rather than else if for the
1222         fallback.  Comment fixes.
1224 2018-11-30  Indu Bhagat  <indu.bhagat@oracle.com>
1226         * coverage.c (get_coverage_counts): Use from_function_decl for precise
1227         function location.
1228         * profile-count.c (profile_count::dump): Add handling for precise
1229         profile quality.
1230         * profile.c (compute_branch_probabilities): Rely on exec_counts instead
1231         of profile_info to set profile_status of function.
1232         (branch_prob): Do not set profile_status of function based on
1233         profile_info. Done above based on exec_counts.
1235 2018-11-30  Jakub Jelinek  <jakub@redhat.com>
1237         PR tree-optimization/88274
1238         * tree-ssa-reassoc.c (optimize_range_tests_xor,
1239         optimize_range_tests_diff): If type has smaller precision than
1240         corresponding mode or if it has non-standard min/max, compute
1241         everything in a standard type for the precision.
1243         PR testsuite/85368
1244         * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
1245         * tree-ssa-ifcombine.c (ifcombine_ifandif): If
1246         --param logical-op-non-short-circuit is present, override
1247         LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
1248         * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
1250 2018-11-30  Jeff Law  <law@redhat.com>
1252         * optabs.c (expand_binop): Use "machine_mode" rather than
1253         "enum machine mode" in most recent change.
1255 2018-11-30  Wilco Dijkstra  <wdijkstr@arm.com>
1257         PR middle-end/64242
1258         * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
1259         the frame pointer.
1260         (expand_builtin_nonlocal_goto): Likewise.
1262 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
1264         * diagnostic-core.h (emit_diagnostic): New decl.
1265         * diagnostic.c (emit_diagnostic): New overload, taking a
1266         rich_location *.
1268 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
1270         * pretty-print.c (class selftest::test_pretty_printer): New
1271         subclass of pretty_printer.
1272         (selftest::test_prefixes_and_wrapping): New test.
1273         (selftest::pretty_print_c_tests): Call it.
1275 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1277         Minimize clone counter memory usage in create_virtual_clone.
1278         * cgraph.h (clone_function_name): Add a variant that takes a
1279         tree decl.
1280         * cgraph.h (cgraph_node::create_virtual_clone): Add a new
1281         argument: num_suffix.
1282         * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
1283         num_suffix to clone_function_name.
1284         (clone_function_name): Add a variant that takes a tree decl.
1285         * ipa-cp.c (create_specialized_node): Keep track of clone
1286         counters in clone_num_suffixes hash map.
1287         (ipcp_driver): Free the counter hash map.
1288         * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
1289         clone per function.
1291 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1292         Make function assembly more independent.
1294         This is achieved by having clone_function_name assign unique clone
1295         numbers for each function independently.
1297         * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
1298         hash map.
1299         (clone_function_name_numbered): Use clone_fn_ids.
1301 2018-11-30  Vladimir Makarov  <vmakarov@redhat.com>
1303         PR rtl-optimization/88179
1304         * lra-constraints.c (address_eliminator::address_eliminator):
1305         Don't eleminate regs in illegitimate address.
1307 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
1309         PR preprocessor/88257
1310         * input.c (get_substring_ranges_for_loc): Fix indentation.
1311         Bulletproof against getting a different files back from
1312         linemap_resolve_location and expand_location_to_spelling_point.
1314 2018-11-30  Alexander Monakov  <amonakov@ispras.ru>
1316         PR gcov-profile/88279
1317         * tree-profile.c (init_ic_make_global_vars): Call
1318         decl_default_tls_model for variable and not it's type.
1320 2018-11-30  Sam Tebbs  <sam.tebbs@arm.com>
1322         * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
1324 2018-11-30  Martin Liska  <mliska@suse.cz>
1326         PR sanitizer/81715
1327         * asan.c (asan_shadow_cst): Remove, partially transform
1328         into flush_redzone_payload.
1329         (RZ_BUFFER_SIZE): New.
1330         (struct asan_redzone_buffer): New.
1331         (asan_redzone_buffer::emit_redzone_byte): Likewise.
1332         (asan_redzone_buffer::flush_redzone_payload): Likewise.
1333         (asan_redzone_buffer::flush_if_full): Likewise.
1334         (asan_emit_stack_protection): Use asan_redzone_buffer class
1335         that is responsible for proper aligned stores and flushing
1336         of shadow memory payload.
1337         * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
1338         (asan_var_and_redzone_size): Likewise.
1339         * cfgexpand.c (expand_stack_vars): Use smaller alignment
1340         (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
1341         for automatic variables more compact.
1343 2018-11-30  Alan Modra  <amodra@gmail.com>
1345         * config/rs6000/predicates.md (easy_fp_constant): Avoid long
1346         dependent insn sequences.
1347         * config/rs6000/rs6000.c (num_insns_constant): Support long
1348         double constants.
1349         * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
1350         length attribute.
1352 2018-11-30  Alan Modra  <amodra@gmail.com>
1354         * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
1355         num_insns_constant_wide.  Make static.  Revise comment.
1356         (num_insns_constant_multi): New function.
1357         (num_insns_constant): Formatting.  Correct CONST_WIDE_INT
1358         calculation.  Simplify and extract code common to both
1359         CONST_INT and CONST_DOUBLE.  Add gcc_unreachable for unhandled
1360         const_double modes.
1361         * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
1363 2018-11-30  Richard Biener  <rguenther@suse.de>
1365         PR tree-optimization/88274
1366         * tree-vrp.c (ranges_from_anti_range): Fix handling of
1367         TYPE_MIN/MAX_VALUE.
1369 2018-11-29  Andi Kleen  <ak@linux.intel.com>
1371         * config/i386/i386.c (current_fentry_section): Add.
1372         (x86_function_profiler): Handle fentry section.
1373         (ix86_attribute_table): Add fentry section.
1374         * config/i386/i386.opt: Add -mfentry-section.
1375         * doc/extend.texi: Document fentry_section attribute.
1376         * doc/invoke.texi: Document -mfentry-section.
1378 2018-11-29  Andi Kleen  <ak@linux.intel.com>
1380         * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
1381         (current_fentry_name): Add.
1382         (x86_function_profiler): Handle fentry_name attribute.
1383         (ix86_handle_fentry_name): Add.
1384         (ix86_attribute_table): Add fentry_name.
1385         * config/i386/i386.opt: Add -mfentry-name
1386         * doc/extend.texi: Document fentry_name.
1387         * doc/invoke.texi: Document minstrument-return.
1389 2018-11-29  Andi Kleen  <ak@linux.intel.com>
1391         * config/i386/i386-opts.h (enum instrument_return): Add.
1392         * config/i386/i386.c (output_return_instrumentation): Add.
1393         (ix86_output_function_return): Call output_return_instrumentation.
1394         (ix86_output_call_insn): Call output_return_instrumentation.
1395         * config/i386/i386.opt: Add -minstrument-return=.
1396         * doc/invoke.texi (-minstrument-return): Document.
1398 2018-11-29  Eric Botcazou  <ebotcazou@adacore.com>
1400         PR target/87807
1401         * config/sparc/sparc-modes.def: Minor tweak.
1402         * config/sparc/sparc.c: Minor reordering.
1403         (sparc_pass_by_reference): Move around.
1404         (traverse_record_type): Change offset from HOST_WIDE_INT to int.
1405         (classify_registers): Likewise for bitpos.
1406         (function_arg_slotno): Remove dead test and tweak comments.
1407         <MODE_RANDOM>: Remove useless assertion and test whether the
1408         parameter is named in order to pass it in FP registers.  Return
1409         the regno for floating-point vector types.
1410         (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
1411         (compute_fp_layout): Likewise.
1412         (count_registers): Likewise.
1413         (assign_int_registers): Likewise.
1414         (assign_fp_registers): Likewise.
1415         (assign_registers): Likewise.
1416         (function_arg_record_value): Change size from HOST_WIDE_INT to int
1417         and use CEIL_NWORDS to compute the number of registers.
1418         (function_arg_union_value): Minor tweaks.
1419         (function_arg_vector_value): Add slotno and named parameters, use
1420         CEIL_NWORDS to compute the number of registers.
1421         (sparc_function_arg_1): Rework handling of vector types.  Change
1422         size from HOST_WIDE_INT to int.
1423         (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
1424         with vector types for the 64-bt ABI.
1425         (sparc_function_arg_advance): Likewise.
1426         (sparc_return_in_memory): Add reference to -fpcc-struct-return.
1427         (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
1428         (sparc_function_value_1): Rework handling of vector types.  Change
1429         size from HOST_WIDE_INT to int.
1431 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
1433         PR target/88152
1434         * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
1435         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
1436         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
1437         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
1438         *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
1439         define_insn_and_split patterns.
1441         PR target/54700
1442         * config/i386/sse.md
1443         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
1444         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
1445         *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
1446         define_insn_and_split.
1448 2018-11-29  Peter Bergner  <bergner@linux.ibm.com>
1450         PR target/87496
1451         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1452         -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
1454 2018-11-29  Uros Bizjak  <ubizjak@gmail.com>
1456         * config/i386/i386.c (inline_memory_move_cost):
1457         Check "in" for 2 in MMX_CLASS_P case.
1458         * config/i386/mmx.md (*mov<mode>_internal): Correct
1459         TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
1460         alternatives in preferred_for_speed attribute calculation.
1462 2018-11-29  Martin Sebor  <msebor@redhat.com>
1464         PR c/88172
1465         PR testsuite/88208
1466         * doc/extend.texi (attribute constructor): Clarify.
1468 2018-11-29  Martin Liska  <mliska@suse.cz>
1470         PR middle-end/88246
1471         * builtins.c (expand_movstr): Fix thinko introduced
1472         when switching to the new enum.
1474 2018-11-29  qing zhao  <qing.zhao@oracle.com>
1476         * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
1477         * common.opt: Add -flive-patching flag.
1478         * doc/invoke.texi: Document -flive-patching.
1479         * flag-types.h (enum live_patching_level): New enum.
1480         * ipa-inline.c (can_inline_edge_p): Disable external functions from
1481         inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
1482         * opts.c (control_options_for_live_patching): New function.
1483         (finish_options): Make flag_live_patching incompatible with flag_lto.
1484         Control IPA optimizations based on different levels of
1485         flag_live_patching.
1487 2018-11-29  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1489         * match.pd (sinh (atanh (x))): New simplification rules.
1490         (cosh (atanh (x))): Likewise.
1492 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
1494         PR target/54700
1495         * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
1496         (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
1497         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
1498         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
1499         *<sse4_1_avx2>_pblendvb_lt): New define_insns.
1501         PR target/88152
1502         * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
1503         * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
1504         * match.pd (define_predicates): Add uniform_integer_cst_p.
1505         (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
1506         so that it works also for vector comparisons with uniform constants
1507         with INTEGER_CST element.
1509         PR target/88234
1510         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
1511         vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
1512         in unsigned_type_for instead of vector integral type where overflow
1513         doesn't wrap.
1515 2018-11-29  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1517         There can be at most one .resolver clone per function
1518         * config/rs6000/rs6000.c (make_resolver_func): Generate
1519         resolver symbol with clone_function_name instead of
1520         clone_function_name_numbered.
1522 2018-11-29  Richard Biener  <rguenther@suse.de>
1524         PR tree-optimization/88243
1525         * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
1526         type of all pattern-sequence stmts to vect_internal_def.
1528 2018-11-29  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1530         PR target/88224
1531         * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
1532         (cortex-r7, cortex-r8): Update default and add new configuration.
1533         * doc/invoke.texi (armv7-r): Add two new vfp options.
1534         (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
1535         support this option.
1537 2018-11-29  Alan Modra  <amodra@gmail.com>
1539         * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
1540         double split for targets other than Darwin.
1542 2018-11-29  Alan Modra  <amodra@gmail.com>
1544         * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
1546 2018-11-29  Alan Modra  <amodra@gmail.com>
1548         * config/rs6000/predicates.md (easy_fp_constant): Remove code
1549         dealing with integers in const_double.  Assert on unexpected
1550         modes.  Delete superfluous ABI_V4 test.
1551         * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
1552         Don't call easy_fp_constant for Pmode.
1553         (rs6000_hash_constant): Remove code dealing with integers in
1554         const_double.
1555         (rs6000_legitimate_constant_p): Likewise.
1556         (output_toc): Formatting, use CONST_DOUBLE_P.
1558 2018-11-29  Alan Modra  <amodra@gmail.com>
1560         * config/rs6000/constraints.md (G, H): Comment on purpose of
1561         constraint.  Correct mode comments and "G" mode test.
1562         * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
1563         alternative handling "F".  Add length attr.  Formatting.
1564         (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
1565         alternative, and move "F" to separate alternative.  Correct
1566         insn lengths.
1568 2018-11-29  Alan Modra  <amodra@gmail.com>
1570         * config/rs6000/predicates.md (easy_vector_constant): Don't call
1571         easy_fp_constant.
1573 2018-11-29  Alan Modra  <amodra@gmail.com>
1575         * config.in (HAVE_AS_PLTSEQ): Add.
1576         * config/rs6000/predicates.md (indirect_call_operand): New.
1577         * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
1578         (rs6000_sibcall_sysv): Declare.
1579         * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
1580         CALL_LONG for -fno-plt.
1581         (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
1582         (rs6000_indirect_call_template_1): Emit .reloc directives for
1583         UNSPEC_PLTSEQ calls.
1584         (rs6000_pltseq_template): New function.
1585         (rs6000_longcall_ref): Add arg parameter.  Use PLT16 insns if
1586         relocs supported by assembler.  Move SYMBOL_REF test to callers.
1587         (rs6000_call_aix): Adjust rs6000_longcall_ref call.  Package
1588         insns in UNSPEC_PLTSEQ, preserving original func_desc.
1589         (rs6000_call_sysv): Likewise.
1590         (rs6000_sibcall_sysv): New function.
1591         * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
1592         * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
1593         UNSPEC_PLT16_LO): New.
1594         (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
1595         (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
1596         cookie in constraints.  Test explicitly for flags in length attr.
1597         Handle unspec operand 1.
1598         (call_value_indirect_nonlocal_sysv): Likewise.
1599         (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
1600         (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
1601         (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
1602         (sibcall_indirect_nonlocal_sysv): New pattern.
1603         (sibcall_value_indirect_nonlocal_sysv): Likewise.
1604         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
1605         call alternatives.
1606         * configure.ac: Check for gas plt sequence marker support.
1607         * configure: Regenerate.
1609 2018-11-29  Alan Modra  <amodra@gmail.com>
1611         * config/rs6000/predicates.md (unspec_tls): New.
1612         * config/rs6000/rs6000-protos.h (rs6000_call_template),
1613         (rs6000_sibcall_template): Update prototype.
1614         (rs6000_longcall_ref): Delete.
1615         (rs6000_call_sysv): Declare.
1616         * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
1617         (global_tlsarg): New variable.
1618         (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
1619         handling.
1620         (print_operand): Extract UNSPEC_TLSGD address operand.
1621         (rs6000_call_template, rs6000_sibcall_template): Remove arg
1622         parameter, extract from second call operand instead.
1623         (rs6000_longcall_ref): Make static, localize vars.
1624         (rs6000_call_aix): Rename parameter to reflect new usage.  Take
1625         tlsarg from global_tlsarg.  Don't create unused rtl or nop insns.
1626         (rs6000_sibcall_aix): Rename parameter to reflect new usage.  Take
1627         tlsarg from global_tlsarg.
1628         (rs6000_call_sysv): New function.
1629         * config/rs6000/rs6000.md: Adjust rs6000_call_template and
1630         rs6000_sibcall_template throughout.
1631         (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
1632         (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
1633         (tls_gdld_nomark): New insn.
1634         (tls_gd): Swap operand order.  Simplify mode selection.
1635         (tls_gd_high, tls_gd_low): Swap operand order.
1636         (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
1637         Simplify mode selection.
1638         (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
1639         (call, call_value): Don't assert for second call operand.
1640         Use rs6000_call_sysv.
1642 2018-11-29  Alan Modra  <amodra@gmail.com>
1644         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
1645         (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
1646         (call_value_nonlocal_darwin64): Remove constraints from second call
1647         arg, the rounded_stack_size_rtx arg.
1648         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
1649         (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
1650         (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
1651         (call_value_local64, call_indirect_nonlocal_sysv),
1652         (call_nonlocal_sysv, call_nonlocal_sysv_secure),
1653         (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
1654         (call_value_nonlocal_sysv_secure, call_local_aix),
1655         (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
1656         (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
1657         (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
1658         (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
1659         (sibcall_value_aix): Likewise.
1661 2018-11-29  Alan Modra  <amodra@gmail.com>
1663         * config/rs6000/rs6000.md (TLSmode): Delete mode iterator.  Replace
1664         with P throughout except for call mems which should use SI.
1665         (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
1666         attributes.  Replace with bits, mode and ptrload respectively.
1668 2018-11-29  Alan Modra  <amodra@gmail.com>
1670         * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
1671         (rs6000_indirect_sibcall_template): Declare.
1672         * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
1673         (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
1674         New functions.
1675         * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
1676         (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
1677         (call_indirect_aix, call_value_indirect_aix): Use
1678         rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
1679         call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
1680         handle both speculation and non-speculation cases.
1681         (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
1682         (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
1684 2018-11-29  Alan Modra  <amodra@gmail.com>
1686         * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
1687         (rs6000_sibcall_template): Declare.
1688         (macho_call_template): Rename from output_call.
1689         * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
1690         (rs6000_call_template, rs6000_sibcall_template): Likewise.
1691         (macho_call_template): Rename from output_call.
1692         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
1693         (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
1694         (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
1695         (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
1696         (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
1697         (call_value_nonlocal_aix): Use rs6000_call_template and update
1698         occurrences of output_call to macho_call_template.
1699         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
1700         (sibcall_value_aix): Use rs6000_sibcall_template.
1702 2018-11-28  Aaron Sawdey  <acsawdey@linux.ibm.com>
1704         * config/rs6000/rs6000-string.c (expand_block_clear): Change how
1705         we determine if unaligned vsx is ok.
1707 2018-11-28  Jan Hubicka  <jh@suse.cz>
1709         * predict.c (determine_unlikely_bbs): Forward declare; also determine
1710         edges that are always known to be taken because it is only likely
1711         edge out of given BB.
1712         (tree_estimate_probability): Call before profile guessing to get bit
1713         of extra precision.
1715 2018-11-28  Jan Hubicka  <jh@suse.cz>
1717         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
1718         profile_probability::always better.
1720 2018-11-28  Jan Hubicka  <jh@suse.cz>
1722         * profile-count.h (profile_count::split): Give better result when
1723         splitting profile_probability::always.
1725 2018-11-28  Vladimir Makarov  <vmakarov@redhat.com>
1727         PR target/88207
1728         * ira-costs.c (scan_one_insn): Process subregs when updating costs
1729         for pseudos and allocnos from insn.
1731 2018-11-28  David Edelsohn  <dje.gcc@gmail.com>
1733         * config/rs6000/aix72.h: Update to match aix71.h changes.
1735 2018-11-28  Jeff Law  <law@redhat.com>
1737         * config/h8300/constraints.md: Add "C" constraint for call insns.
1738         * config/h8300/h8300.md (call, call_value): Turn into a define_expand
1739         and define_insn pair.  Move invalid call targets into a register in
1740         the expander and fix constraints in the matching pattern.
1741         * config/h8300/predicates.md (call_expander_operand): Renamed from
1742         call_insn_operand.  Reject things we shouldn't be trying to handle.
1743         (call_insn_operand): New predicate for use by the call/call_value
1744         insns.
1745         (small_call_insn_operand): Update appropriately.
1747 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1749         * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
1750         calls to strtok with strtok_r.
1752 2018-11-28  Richard Biener  <rguenther@suse.de>
1754         PR tree-optimization/88223
1755         * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
1756         over a stored-same value may-alias store make sure to consider
1757         partial overlaps which are valid when TBAA reasonings do not
1758         apply and byte-granular overlaps are possible at all.
1760 2018-11-28  Richard Biener  <rguenther@suse.de>
1762         PR tree-optimization/88217
1763         * vr-values.c (vr_values::extract_range_from_phi_node): Make
1764         sure to handle results > +INF and < -INF correctly when
1765         trying to drop down to +INF - 1 or -INF + 1.
1767 2018-11-28  Alan Modra  <amodra@gmail.com>
1769         * xcoffout.c (do_block): Signed/unsigned warning fix.
1771 2018-11-28  Richard Biener  <rguenther@suse.de>
1773         PR tree-optimization/79351
1774         * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
1775         empty CONSTRUCTORs ensure the store is at a constant position.
1777 2018-11-28  Richard Biener  <rguenther@suse.de>
1779         PR tree-optimization/88229
1780         * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
1781         is an INTEGER_CST before accessing it so.
1783 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1785         * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
1787 2018-11-28  Jakub Jelinek  <jakub@redhat.com>
1789         PR target/88189
1790         * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
1791         SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1.  Formatting fixes.
1792         * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
1794 2018-11-27  Martin Liska  <mliska@suse.cz>
1796         * gcov.c (generate_results): Append current_working_directory
1797         only when exists.
1799 2018-11-27  Mike Gulick  <mgulick@mathworks.com>
1801         PR preprocessor/83173
1802         * input.c (dump_location_info): Dump reason and included_from
1803         fields from line_map_ordinary struct.  Fix indentation when
1804         location > 5 digits.
1805         * diagnostic-show-locus.c (num_digits, num_digits): Move to
1806         diagnostic.c to allow it to be utilized by input.c.
1807         * diagnostic.c (num_digits, selftest::test_num_digits): Moved
1808         here.
1809         (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
1810         * diagnostic.h (num_digits): Add extern definition.
1812 2018-11-27  Fredrik Noring  <noring@nocrew.org>
1814         * config/mips/mips.c (mips_reorg_process_insns)
1815         (mips_option_override): Handle `-mfix-r5900'.
1816         * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
1817         `mno-fix-r5900'.
1818         * config/mips/mips.opt (mfix-r5900): New option.
1819         * doc/invoke.texi: Document the `r5900' processor name, and
1820         `-mfix-r5900' and `-mno-fix-r5900' options.
1822 2018-11-27  Jakub Jelinek  <jakub@redhat.com>
1824         PR target/88188
1825         * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
1826         output_operand_lossage instead of gcc_assert.
1827         <case 't'>: Likewise.
1828         <case 'z'>: Likewise.
1829         <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
1831 2018-11-27  Jeff Law  <law@redhat.com>
1833         * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
1834         in call to move_by_pieces.
1836         * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
1837         to move_by_pieces.
1839         * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
1840         call to move_by_pieces.
1842         * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
1843         in call to move_by_pieces.
1845         * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
1846         RETURN_BEGIN in call to move_by_pieces.
1847         (microblaze_expand_block_move): Likewise.
1849 2018-11-27  Tamar Christina  <tamar.christina@arm.com>
1851         * doc/md.texi (xorsign): Document it.
1853 2018-11-27  Martin Liska  <mliska@suse.cz>
1855         * config/gnu-user.h (TARGET_F951_OPTIONS): New.
1856         * gcc.c (find_fortran_preinclude_file): New function
1857         to handle Fortran pre-include.
1859 2018-11-27  Martin Liska  <mliska@suse.cz>
1861         * asan.c (asan_emit_stack_protection): Use new enum values
1862         instead of int constants.
1863         * builtins.c (expand_builtin_memory_copy_args): Replace int
1864         type with memop_ret enum type.
1865         (expand_builtin_mempcpy_args): Likewise.
1866         (expand_builtin_memcpy): Use new enum values
1867         instead of int constants. Likewise.
1868         (expand_builtin_mempcpy): Likewise.
1869         (expand_movstr): Likewise.
1870         (expand_builtin_strcpy_args): Likewise.
1871         (expand_builtin_stpcpy_1): Likewise.
1872         (expand_builtin_strncpy): Likewise.
1873         (expand_builtin_memset_args): Likewise.
1874         * expr.c (move_by_pieces_d::finish_endp): Rename to ...
1875         (move_by_pieces_d::finish_retmode): ... this.
1876         (move_by_pieces): Change last argument type to memop_ret.
1877         (store_by_pieces): Use new enum values
1878         instead of int constants.
1879         (emit_block_move_hints): Likewise.
1880         (emit_push_insn): Likewise.
1881         (store_expr): Likewise.
1882         * expr.h (store_by_pieces): Change int to newly added enum
1883         type.
1884         * rtl.h (enum memop_ret): Define.
1885         (move_by_pieces): Use the enum type.
1887 2018-11-27  Alan Modra  <amodra@gmail.com>
1889         * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
1890         cpu here.
1891         (ASM_CPU_SPEC): Do so here.  Rewrite using if .. else if .. specs
1892         form.  Error on missing -mcpu case.
1893         * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
1894         (asm_names <!_AIX>): Add missing cpus.  Update NULL case.  Apply
1895         PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
1896         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
1897         else if .. specs form.  Error on missing -mcpu case.  Don't output
1898         duplicate -maltivec.  Apply PR63177 fix for -mcpu=powerpc64le.
1900 2018-11-26  David Malcolm  <dmalcolm@redhat.com>
1902         * dump-context.h (dump_context::dump_loc): Convert 1st param from
1903         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1904         const dump_location_t & to const dump_user_location_t &.
1905         (dump_context::dump_loc_immediate): Convert 2nd param from
1906         const dump_location_t & to const dump_user_location_t &.
1907         (dump_context::dump_gimple_stmt): Convert 1st param from
1908         dump_flags_t to const dump_metadata_t &.
1909         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1910         2nd param from const dump_location_t & to
1911         const dump_user_location_t &.
1912         (dump_context::dump_gimple_expr): Convert 1st param from
1913         dump_flags_t to const dump_metadata_t &.
1914         (dump_context::dump_gimple_expr_loc): Likewise; convert
1915         2nd param from const dump_location_t & to
1916         const dump_user_location_t &.
1917         (dump_context::dump_generic_expr): Convert 1st param from
1918         dump_flags_t to const dump_metadata_t &.
1919         (dump_context::dump_generic_expr_loc): Likewise; convert
1920         2nd param from const dump_location_t & to
1921         const dump_user_location_t &.
1922         (dump_context::dump_printf_va): Convert 1st param from
1923         dump_flags_t to const dump_metadata_t &.
1924         (dump_context::dump_printf_loc_va): Likewise; convert
1925         2nd param from const dump_location_t & to
1926         const dump_user_location_t &.
1927         (dump_context::dump_dec): Convert 1st param from
1928         dump_flags_t to const dump_metadata_t &.
1929         (dump_context::dump_symtab_node): Likewise.
1930         (dump_context::begin_scope): Split out 2nd param into
1931         user and impl locations.
1932         (dump_context::ensure_pending_optinfo): Add metadata param.
1933         (dump_context::begin_next_optinfo): Replace dump_location_t param
1934         with metadata and user location.
1935         * dumpfile.c (dump_context::dump_loc): Convert 1st param from
1936         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1937         const dump_location_t & to const dump_user_location_t &.
1938         (dump_context::dump_loc_immediate): Convert 2nd param from
1939         const dump_location_t & to const dump_user_location_t &.
1940         (dump_context::dump_gimple_stmt): Convert 1st param from
1941         dump_flags_t to const dump_metadata_t &.
1942         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1943         2nd param from const dump_location_t & to
1944         const dump_user_location_t &.
1945         (dump_context::dump_gimple_expr): Convert 1st param from
1946         dump_flags_t to const dump_metadata_t &.
1947         (dump_context::dump_gimple_expr_loc): Likewise; convert
1948         2nd param from const dump_location_t & to
1949         const dump_user_location_t &.
1950         (dump_context::dump_generic_expr): Convert 1st param from
1951         dump_flags_t to const dump_metadata_t &.
1952         (dump_context::dump_generic_expr_loc): Likewise; convert
1953         2nd param from const dump_location_t & to
1954         const dump_user_location_t &.
1955         (dump_context::dump_printf_va): Convert 1st param from
1956         dump_flags_t to const dump_metadata_t &.
1957         (dump_context::dump_printf_loc_va): Likewise; convert
1958         2nd param from const dump_location_t & to
1959         const dump_user_location_t &.
1960         (dump_context::dump_dec): Convert 1st param from
1961         dump_flags_t to const dump_metadata_t &.
1962         (dump_context::dump_symtab_node): Likewise.
1963         (dump_context::begin_scope): Split out 2nd param into
1964         user and impl locations.
1965         (dump_context::ensure_pending_optinfo): Add metadata param.
1966         (dump_context::begin_next_optinfo): Replace dump_location_t param
1967         with metadata and user location.
1968         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
1969         const dump_metadata_t &.
1970         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
1971         const dump_location_t & to const dump_user_location_t &.
1972         (dump_gimple_expr): Convert 1st param from dump_flags_t to
1973         const dump_metadata_t &.
1974         (dump_gimple_expr_loc): Likewise; convert 2nd param from
1975         const dump_location_t & to const dump_user_location_t &.
1976         (dump_generic_expr): Convert 1st param from dump_flags_t to
1977         const dump_metadata_t &.
1978         (dump_generic_expr_loc): Likewise; convert 2nd param from
1979         const dump_location_t & to const dump_user_location_t &.
1980         (dump_printf): Convert 1st param from dump_flags_t to
1981         const dump_metadata_t &.
1982         (dump_printf_loc): Likewise; convert 2nd param from
1983         const dump_location_t & to const dump_user_location_t &.
1984         (dump_dec): Convert 1st param from dump_flags_t to
1985         const dump_metadata_t &.
1986         (dump_symtab_node): Likewise.
1987         (dump_begin_scope): Split out 2nd param into user and impl
1988         locations.
1989         (selftest::assert_impl_location_eq): New function.
1990         (ASSERT_IMPL_LOCATION_EQ): New macro.
1991         (selftest::test_impl_location): Update to use
1992         ASSERT_IMPL_LOCATION_EQ.
1993         (selftest::test_capture_of_dump_calls): Convert "loc" to
1994         dump_user_location_t.  Add ASSERT_IMPL_LOCATION_EQ throughout,
1995         verifying line numbers of dump emissions.
1996         * dumpfile.h (class dump_metadata_t): New class.
1997         (dump_printf): Convert 1st param from dump_flags_t to
1998         const dump_metadata_t &.
1999         (dump_printf_loc): Likewise; convert 2nd param from
2000         const dump_location_t & to const dump_user_location_t &.
2001         (dump_generic_expr_loc): Likewise.
2002         (dump_generic_expr): Convert 1st param from dump_flags_t to
2003         const dump_metadata_t &.
2004         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2005         const dump_location_t & to const dump_user_location_t &.
2006         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2007         const dump_metadata_t &.
2008         (dump_gimple_expr_loc): Likewise; convert 2nd param from
2009         const dump_location_t & to const dump_user_location_t &.
2010         (dump_gimple_expr): Convert 1st param from dump_flags_t to
2011         const dump_metadata_t &.
2012         (dump_symtab_node): Likewise.
2013         (dump_dec): Likewise.
2014         (dump_begin_scope): Split out 2nd param into user and impl
2015         locations.
2016         (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
2017         location and impl_location, and capture the impl_location.
2018         (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
2019         * loop-unroll.c (report_unroll): Update for changes to
2020         dump_printf_loc and dump_printf.
2021         * opt-problem.cc (opt_problem::opt_problem): Update for change to
2022         dump_loc.
2023         * optinfo-emit-json.cc
2024         (selftest::test_building_json_from_dump_calls): Convert "loc" from
2025         dump_location_t to dump_user_location_t.
2026         * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
2027         dump_loc_immediate.
2028         * profile.c (compute_branch_probabilities): Update for change to
2029         dump_printf_loc.
2030         * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
2031         * tree-vect-slp.c (vect_print_slp_tree): Update for change to
2032         dump_printf_loc.
2034 2018-11-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2036         * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
2037         and check_effective_target_ptr_eq_long.
2039 2018-11-27  Alan Modra  <amodra@gmail.com>
2041         * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
2042         (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
2043         (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
2044         (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
2045         (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
2046         * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
2047         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2048         (ASM_APP_ON, ASM_APP_OFF): Don't define.
2049         * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2050         * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2051         * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2052         * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
2053         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2054         * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2055         * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
2056         (GNU_USER_TARGET_CRTN): Redefine.
2057         (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
2058         (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
2059         (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
2060         (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
2061         (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
2062         (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
2063         (LINK_EH_SPEC): Don't define.
2065 2018-11-26  Martin Sebor  <msebor@redhat.com>
2067         PR 87756
2068         * expr.c (string_constant): Handle top-level decls of all character
2069         types and subobjects of narrow character type.
2071 2018-11-27  Alan Modra  <amodra@gmail.com>
2073         * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
2074         * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
2075         GNU_USER_TARGET_OS_CPP_BUILTINS.
2076         (RS6000_ABI_NAME): Define.
2077         * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
2078         GNU_USER_TARGET_OS_CPP_BUILTINS.
2079         (MUSL_DYNAMIC_LINKER32): Undef before defining.
2080         (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
2081         (CHOOSE_DYNAMIC_LINKER): Don't define.
2082         (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
2083         * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
2084         (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
2085         not already defined.
2086         (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
2087         TARGET_OS_CPP_BUILTINS.
2089 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2091         * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
2093 2018-11-26  Jakub Jelinek  <jakub@redhat.com>
2095         PR target/88195
2096         * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
2097         and !TARGET_64BIT, return NULL_TREE.
2099         PR c++/86900
2100         * dwarf2out.c (secname_for_decl): For functions with
2101         DECL_SECTION_NAME if in_cold_section_p, try to return
2102         current_function_section's name if it is a named section.
2104 2018-11-26  Maya Rashish  <coypu@sdf.org>
2106         PR target/58397
2107         * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
2108         * config/x-netbsd: New file.
2109         * config/host-netbsd: New file.
2111 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2113         * doc/sourcebuild.texi: Document check_effective_target_size20plus.
2114         Clarify documentation for check_effective_target_size32plus.
2116 2018-11-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
2118         * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
2119         and modify expand_strncmp_vec_sequence.
2120         (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
2121         (generate_6432_conversion): New function.
2122         (expand_block_compare): Add support for vsx.
2123         (expand_block_compare_gpr): New function.
2124         * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
2125         default limit to 63 because of more compact vsx code.
2127 2018-11-26  Uros Bizjak  <ubizjak@gmail.com>
2129         PR target/88178
2130         * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
2131         for registers for which no debug information can be generated.
2132         Use INVALID_REGNUM for non-existent registers.
2133         (dbx64_register_map): Ditto.
2134         (svr4_dbx_register_map): Ditto.
2136 2018-11-26  Sandra Loosemore  <sandra@codesourcery.com>
2138         PR c/57182
2139         * doc/invoke.texi (Optimize Options): Document that
2140         -freorder-functions also uses "hot" and "cold" attributes.
2142 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2144         * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
2146 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2148         * config/arm/types.md (mul64): Rename to...
2149         (widen_mul64): ... This.
2150         * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
2151         * config/arm/cortex-a15.md: Likewise.
2152         * config/arm/cortex-a5.md: Likewise.
2153         * config/arm/cortex-a53.md: Likewise.
2154         * config/arm/cortex-a57.md: Likewise.
2155         * config/arm/cortex-a7.md: Likewise.
2156         * config/arm/cortex-m4.md: Likewise.
2157         * config/arm/exynos-m1.md: Likewise.
2158         * config/arm/marvell-pj4.md: Likewise.
2159         * config/arm/xgene1.md: Likewise.
2161 2018-11-26  Richard Biener  <rguenther@suse.de>
2163         PR tree-optimization/88182
2164         * tree-vect-loop.c (vectorizable_reduction): Pick up single
2165         correct reduc_def_info.
2166         * tree-vect-slp.c (vect_analyze_slp_instance): Set
2167         STMT_VINFO_REDUC_DEF of the first stmt.
2169 2018-11-26  Andreas Krebbel  <krebbel@linux.ibm.com>
2171         * doc/invoke.texi: Document z14/arch12 -march option.
2173 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
2175         PR c/57166
2176         * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
2177         (Warning Options): Likewise.
2179 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
2181         PR web/79738
2182         * doc/extend.texi (Common Function Attributes): Clarify that
2183         functions with "const" attribute can read const global variables.
2185 2018-11-26  Alan Modra  <amodra@gmail.com>
2187         * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
2189 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
2191         PR other/54265
2192         * doc/extend.texi (Common Variable Attributes): Use preferred
2193         placement of type attributes in examples, plus whitespace fixes.
2194         (Type Attributes): Clarify why placement of attributes
2195         immediately after struct/union/enum keyword is preferred.
2196         (Common Type Attributes): Use preferred placement of type
2197         attributes in examples, plus more whitespace fixes.
2199 2018-11-25  Paul Koning  <ni1d@arrl.net>
2201         * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
2203 2018-11-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2205         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
2206         attributes): Document d_runtime.
2208 2018-11-25  Vladimir Makarov  <vmakarov@redhat.com>
2210         PR bootstrap/88157
2211         * ira-costs.c (record_operand_costs): Initiate register move cost
2212         for mode.
2214 2018-11-23  Jeff Law  <law@redhat.com>
2216         PR rtl-optimization/87468
2217         * tree-ssa-threadupdate.c (create_block_for_threading): Clear
2218         EDGE_IGNORE on all outgoing edges of the duplicate block.
2220 2018-11-23  Vladimir Makarov  <vmakarov@redhat.com>
2222         PR bootstrap/88157
2223         * ira-costs.c (record_operand_costs): Use bigger hard reg class if
2224         its mode does not fit to the original class.
2226 2018-11-23  Martin Sebor  <msebor@redhat.com>
2228         PR tree-optimization/87756
2229         * expr.c (string_constant): Synthesize a string literal from
2230         the address of a constant character.
2231         * tree.c (build_string_literal): Add an argument.
2232         * tree.h (build_string_literal): Same.
2234 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
2236         * config/aarch64/aarch64-cores.def: Define emag.
2237         * config/aarch64/aarch64-tune.md: Regenerated with emag.
2238         * config/aarch64/aarch64.c (emag_tunings): New struct.
2239         * doc/invoke.texi: Document mtune value.
2241 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
2242             Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
2244         * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
2245         for GCC 9.
2246         * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
2247         prefetch tunings.
2248         * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
2249         costs.
2250         * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
2251         for XGene1.
2253 2018-11-23  Richard Biener  <rguenther@suse.de>
2255         PR tree-optimization/88149
2256         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
2257         the case where there are two different def types for the
2258         same operand at different operand position in the same stmt.
2260 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
2262         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
2263         gen_atomic_compare_swap_1.
2264         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
2265         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
2266         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
2267         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
2268         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
2269         checks.
2270         * config/arm/arm.md (neon_vext<mode>),
2271         (neon_vrev64<mode>, neon_vrev32<mode>),
2272         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
2273         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
2274         before the pattern name.
2275         * config/arm/sync.md:
2276         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
2277         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
2279 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
2281         PR tree-optimization/86614
2282         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
2283         if TREE_NO_WARNING is set on ref.ref.
2285 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
2286             Alan Coopersmith  <alan.coopersmith@oracle.com>
2288         PR c/53608
2289         * doc/extend.texi (Designated Inits): Clarify handling of multiple
2290         initializers for unions.
2292 2018-11-22  Jan Hubicka  <jh@suse.cz>
2294         PR lto/88142
2295         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
2296         parameters; do not warn here.
2297         (odr_subtypes_equivalent_p): Likewise.
2298         (warn_odr): Fix typo.
2299         (warn_types_mismatch): Do not output confused warnings on integer types.
2300         (odr_types_equivalent_p): Update.
2302 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
2304         * config/i386/i386.c (ix86_option_override_internal,
2305         ix86_can_inline_p, classify_argument, construct_container,
2306         ix86_expand_prologue, ix86_expand_split_stack_prologue,
2307         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
2308         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
2309         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
2310         Formatting fixes.
2312         * config/i386/i386.c (ix86_option_override_internal): For
2313         stack_protector_guard related options, use opts_set->x_ instead
2314         of global_options_set. and prefix options with opts->x_ .  Move
2315         defaults for offset and reg into else block.
2317 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
2319         * config/i386/i386.c (ix86_check_avx_upper_register):
2320         Return true for all SSE registers with mode bitsize > 128.
2322 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
2324         PR rtl-optimization/87718
2325         * ira-costs.c: Remove trailing white-spaces.
2326         (record_operand_costs): Add a special treatment for moves
2327         involving a hard register.
2329 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
2331         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
2332         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
2334 2018-11-22  Martin Liska  <mliska@suse.cz>
2336         * common/config/i386/i386-common.c (processor_names): Add
2337         static assert and add missing "znver2".
2338         (ix86_get_valid_option_values): Add checking assert for null
2339         values and add "native" value if feasible.
2340         * config/i386/i386.h: Do not declare size of processor_names.
2341         * common/config/i386/i386-common.c:
2342         * config/i386/i386.c: Add static assert for size
2343         of processor_cost_table.
2345 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2347         * target-insns.def (stack_protect_combined_set): Define new standard
2348         pattern name.
2349         (stack_protect_combined_test): Likewise.
2350         * cfgexpand.c (stack_protect_prologue): Try new
2351         stack_protect_combined_set pattern first.
2352         * function.c (stack_protect_epilogue): Try new
2353         stack_protect_combined_test pattern first.
2354         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
2355         parameters to control which register to use as PIC register and force
2356         reloading PIC register respectively.  Insert in the stream of insns if
2357         possible.
2358         (legitimize_pic_address): Expose above new parameters in prototype and
2359         adapt recursive calls accordingly.  Use pic_reg if non null instead of
2360         cached one.
2361         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
2362         (arm_legitimize_address): Adapt to new legitimize_pic_address
2363         prototype.
2364         (thumb_legitimize_address): Likewise.
2365         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
2366         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
2367         (thumb1_expand_prologue): Likewise.
2368         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
2369         change.
2370         (arm_load_pic_register): Likewise.
2371         * config/arm/predicated.md (guard_addr_operand): New predicate.
2372         (guard_operand): New predicate.
2373         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
2374         prototype change.
2375         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
2376         prototype change.
2377         (stack_protect_combined_set): New expander..
2378         (stack_protect_combined_set_insn): New insn_and_split pattern.
2379         (stack_protect_set_insn): New insn pattern.
2380         (stack_protect_combined_test): New expander.
2381         (stack_protect_combined_test_insn): New insn_and_split pattern.
2382         (arm_stack_protect_test_insn): New insn pattern.
2383         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
2384         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
2385         (UNSPEC_SP_TEST): Likewise.
2386         * doc/md.texi (stack_protect_combined_set): Document new standard
2387         pattern name.
2388         (stack_protect_set): Clarify that the operand for guard's address is
2389         legal.
2390         (stack_protect_combined_test): Document new standard pattern name.
2391         (stack_protect_test): Clarify that the operand for guard's address is
2392         legal.
2394 2018-11-22  Richard Biener  <rguenther@suse.de>
2396         PR tree-optimization/88148
2397         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
2398         valueization callback parameter and handle it.
2399         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
2400         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
2402 2018-11-22  Richard Biener  <rguenther@suse.de>
2404         PR lto/87229
2405         PR lto/88112
2406         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
2407         which can appear in size expressions.
2408         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
2409         CALL_EXPR_BY_DESCRIPTOR.
2410         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
2411         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
2412         CALL_EXPR_BY_DESCRIPTOR.
2413         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
2415         Revert
2416         PR lto/87229
2417         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
2418         sizepos values.
2420 2018-11-22  Richard Biener  <rguenther@suse.de>
2422         PR tree-optimization/88069
2423         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
2424         apply to default defs.
2426 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
2428         PR target/85644
2429         PR target/86832
2430         * config/i386/i386.c (ix86_option_override_internal): Default
2431         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
2432         is defined.
2433         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
2434         stack_protect_test, stack_protect_test_<mode>): Use empty condition
2435         instead of TARGET_SSP_TLS_GUARD.
2437 2018-11-22  Martin Liska  <mliska@suse.cz>
2439         PR sanitizer/88017
2440         * config/i386/i386.c (ix86_option_override_internal): Error
2441         about usage -mabi=ms and -fsanitize=thread.
2443 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
2445         Revert the revert:
2446         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
2448         Revert:
2449         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
2451         * lra-spills.c (lra_final_code_change): Remove useless move insns.
2453 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
2455         PR middle-end/88129
2456         * function.c (expand_function_end): Do not emit extra blockage insn.
2458 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
2460         PR target/85667
2461         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
2462         of FIRST_SSE_REG for 4 or 8 byte modes.
2464 2018-11-21  Jan Hubicka  <jh@suse.cz>
2466         PR lto/87957
2467         * tree.c (fld_decl_context): Break out from ...
2468         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
2469         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
2470         (fld_incomplete_type_of): Build copy of TYP_DECL.
2471         * ipa-devirt.c (free_enum_values): Rename to ...
2472         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
2473         and TREE_TYPEs of TYPE_DECLs.
2474         (get_odr_type): Initialize odr_vtable_hash if needed.
2476 2018-11-21  Alexandre Oliva <oliva@adacore.com>
2478         * final.c (compute_discriminator): Declare.  Renamed from...
2479         (maybe_set_discriminator): ... this.  Set and return a local.
2480         (override_discriminator): New.
2481         (final_scan_insn_1): Set it.
2482         (notice_source_line): Adjust.  Always set discriminator.
2484 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
2486         PR target/87839
2487         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
2488         rIJ constraint for aarch64_plus_operand rather than rn.
2490 2018-11-21  Renlin Li  <renlin.li@arm.com>
2492         PR middle-end/84877
2493         * explow.h (get_dynamic_stack_size): Declare it as external.
2494         * explow.c (record_new_stack_level): Remove function static attribute.
2495         * function.c (assign_stack_local_1): Dynamically align the stack slot
2496         addr for parameter copy on the stack.
2498 2018-11-21  Richard Biener  <rguenther@suse.de>
2500         PR bootstrap/88133
2501         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
2502         * Makefile.in (bitmap.o-warn): Remove again.
2504 2018-11-20  Jeff Law  <law@redhat.com>
2506         PR tree-optimization/88069
2507         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
2508         degenerate virtual PHIs.
2510 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
2512         PR target/87317
2513         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
2514         nonimmediate_operand with register_operand.
2515         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
2516         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
2517         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
2518         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
2519         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
2520         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
2521         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
2522         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
2523         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
2524         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
2525         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
2526         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
2527         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
2528         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
2529         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
2530         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
2531         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
2532         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
2533         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
2534         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
2535         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
2536         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
2537         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
2538         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
2539         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
2540         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
2541         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
2543 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
2545         * read-rtl.c (apply_subst_iterator): Handle define_split and
2546         define_insn_and_split.
2548 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
2550         PR rtl-optimization/87817
2551         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
2552         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
2553         in the pattern to avoid triggering UB when operands[2] is zero.
2554         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
2555         (*tbm_bextri_<mode>): ... this.
2557 2018-11-21  Tom de Vries  <tdevries@suse.de>
2559         PR driver/79855
2560         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
2561         period.
2563 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
2565         * configure.ac: Add check for Binutils to determine whether vector
2566         load/store alignments hints are being supported.
2567         * config.in: Regenerate.
2568         * configure: Regenerate.
2569         * config/s390/s390.c (print_operand): Support new output
2570         modifier A.
2571         * config/s390/s390.md ("movti"): Append alignment hint output
2572         using the new output modifier 'A'.
2573         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
2574         ("*vec_ti_to_v1ti"): Likewise.
2576 2018-11-20  Martin Sebor  <msebor@redhat.com>
2578         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
2580 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
2582         PR lto/84044
2583         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
2584         compare ENUM values.
2586 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
2588         PR tree-optimization/87895
2589         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
2590         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
2591         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
2592         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
2593         has any preds and always use EDGE_PRED (, 0).
2595 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
2597         PR target/88070
2598         * mode-switching.c (create_pre_exit): After reload, always split the
2599         fallthrough edge to the exit block.
2601 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
2603         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
2604         * tree.c (build_array_type_1): Forward declare.
2605         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
2606         (fld_type_variant): Likewise.
2607         (fld_simplified_types): New hash.
2608         (fld_process_array_type): New function.
2609         (fld_incomplete_type_of): Handle array and enumeration types.
2610         (fld_simplified_type): Handle simplification of arrays.
2611         (free_lang_data): Allocate and free simplified types hash.
2613 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
2615         PR lto/87957
2616         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
2617         when sybtype already violates ODR.
2618         (get_odr_type): Do not ICE when insert is false and type duplicate
2619         is not registered yet.
2620         (register_odr_type): Be sure to register subtypes first.
2622 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
2624         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
2625         DImode parts of the target operand.
2627 2018-11-20  Nathan Sidwell  <nathan@acm.org>
2629         PR 87926
2630         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
2632 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
2634         PR lto/87997
2635         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
2636         type checking to...
2637         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
2639 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
2641         PR ipa/87706
2642         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
2643         * ipa.c (possible_inline_candidate_p): Break out from ..
2644         (process_references): ... here ; drop before_inlining_p;
2645         cleanup handling of alises.
2646         (walk_polymorphic_call_targets): Likewise.
2647         (symbol_table::remove_unreachable_nodes): Likewise.
2648         * passes.c (pass_data_ipa_remove_symbols): New structure.
2649         (pass_ipa_remove_symbols): New pass.
2650         (make_pass_ipa_remove_symbols): New function.
2651         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
2652         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
2654 2018-11-20  Richard Biener  <rguenther@suse.de>
2656         * tree-vect-stmts.c (vectorizable_condition): Do not get
2657         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
2658         pointless vect_is_simple_use calls.
2660 2018-11-20  Richard Biener  <rguenther@suse.de>
2662         PR tree-optimization/88074
2663         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
2664         backedges.
2666 2018-11-20  Richard Biener  <rguenther@suse.de>
2668         PR tree-optimization/88069
2669         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
2670         virtual PHI arguments.
2672 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
2674         PR target/88083
2675         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
2676         involved.
2677         * rtl.h (contains_constant_pool_address_p): New function.
2678         * rtlanal.c (contains_constant_pool_address_p): Likewise.
2680 2018-11-20  Richard Biener  <rguenther@suse.de>
2682         PR middle-end/83215
2683         * alias.c (component_uses_parent_alias_set_from): Remove
2684         alias-set zero and TYPE_TYPELESS_STORAGE case both already
2685         handled in other ways.
2687 2018-11-20  Richard Biener  <rguenther@suse.de>
2689         PR tree-optimization/88087
2690         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
2691         call fntype.
2692         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
2693         call fntype.
2695 2018-11-20  Richard Biener  <rguenther@suse.de>
2697         PR middle-end/88089
2698         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
2700 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
2702         PR lto/87957
2703         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
2705 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
2707         PR driver/50250
2708         * doc/invoke.texi (Link Options): Mention shared libraries
2709         in documentation for the -l option.  Simplify discussion and
2710         point to the system linker documentation for details.
2712 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
2714         PR rtl-optimization/88033
2715         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
2716         to itself.  Use HARD_REGISTER_NUM_P.
2718 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2720         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
2721         of the loop if vf is non-constant.
2723 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2725         PR tree-optimization/87025
2726         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
2727         immediately after creating the scope optinfo.
2728         (selftest::test_pr87025): New function.
2729         (selftest::dumpfile_c_tests): Call it.
2730         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
2731         that we're not popping the top-level records array.
2732         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
2733         not changing the kind of a "scope" optinfo.
2735 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2737         PR tree-optimization/87025
2738         * dump-context.h: Include "optinfo.h".
2739         (class optrecord_json_writer): New forward decl.
2740         (dump_context::forcibly_enable_optinfo_p): Delete.
2741         (dump_context::optinfo_enabled_p): New member function.
2742         (dump_context::optimization_records_enabled_p): New member
2743         function.
2744         (dump_context::set_json_writer): New member function.
2745         (dump_context::emit_optinfo): New member function.
2746         (dump_context::m_forcibly_enable_optinfo): Delete.
2747         (dump_context::m_json_writer): New member data.
2748         * dumpfile.c (dump_context::set_json_writer): New member function.
2749         (dump_context::finish_any_json_writer): New member function.
2750         (dump_context::end_scope): Replace call to
2751         optimization_records_maybe_pop_dump_scope with call to
2752         m_json_writer->pop_scope.
2753         (dump_context::optinfo_enabled_p): New member function.
2754         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
2755         to dump_context::emit_optinfo.
2756         (dump_context::emit_optinfo): New member function.
2757         (temp_dump_context::temp_dump_context): Replace
2758         m_forcibly_enable_optinfo with call to set_json_writer.
2759         (temp_dump_context::~temp_dump_context): Clean up any json writer.
2760         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
2761         optinfo-emit-json.h
2762         (the_json_writer): Delete.
2763         (optimization_records_start): Delete.
2764         (optimization_records_finish): Delete.
2765         (optimization_records_enabled_p): Delete, in favor of
2766         dump_context::optimization_records_enabled_p.
2767         (optimization_records_maybe_record_optinfo): Delete.
2768         (optimization_records_maybe_pop_dump_scope): Delete.
2769         * optinfo-emit-json.h: Include "json.h".  Delete forward
2770         decl of opt_pass.
2771         (optimization_records_start): Delete.
2772         (optimization_records_finish): Delete.
2773         (optimization_records_enabled_p): Delete.
2774         (optimization_records_maybe_record_optinfo): Delete.
2775         (optimization_records_maybe_pop_dump_scope): Delete.
2776         (class optrecord_json_writer): Move here from
2777         optinfo-emit-json.cc.
2778         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
2779         to optinfo::emit with call to dump_context::emit_optinfo.
2780         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2781         (optinfo_enabled_p): Delete, in favor of
2782         dump_context::optinfo_enabled_p.
2783         (optinfo_wants_inlining_info_p): Update for conversion o
2784         optimization_records_enabled_p to a member function of
2785         dump_context.
2786         * optinfo.h (optinfo_enabled_p): Delete, in favor of
2787         dump_context::optinfo_enabled_p.
2788         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2789         * toplev.c: Include "dump-context.h".
2790         (compile_file): Replace call to optimization_records_finish with
2791         dump_context::finish_any_json_writer.
2792         (do_compile): Replace call to optimization_records_start with
2793         conditionally creating a optrecord_json_writer for the
2794         dump_context.
2796 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
2798         PR tree-optimization/88071
2799         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
2800         of false as last argument to gsi_remove.
2801         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
2802         false as last argument to gsi_replace.
2804         PR debug/87039
2805         * omp-expand.c: Don't include debug.h.
2806         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
2807         from innermost outer parallel, task, teams or target that has a
2808         child_fn set, or, if there is no such outer region, use
2809         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
2810         whether a suitable BLOCK is found or not.
2811         (expand_parallel_call, expand_teams_call): Don't call
2812         adjust_context_and_scope here.
2813         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
2814         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
2815         adjust_context_and_scope.
2816         * dwarf2out.c (dwarf2out_early_global_decl): For
2817         decl_function_context recurse instead of calling dwarf2out_decl.
2819 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
2821         PR rtl-optimization/85925
2822         * rtl.h (word_register_operation_p): New predicate.
2823         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
2824         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
2825         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
2826         (num_sign_bit_copies1): Likewise.
2828 2018-11-19  Richard Biener  <rguenther@suse.de>
2830         PR lto/87229
2831         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
2832         sizepos values.
2834 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
2836         * compare-elim.c (struct comparison): Add not_in_a field.
2837         (is_not): New static function.
2838         (strip_not): Likewise.
2839         (conforming_compare): Handle a NOT in the first operand.
2840         (can_eliminate_compare): Likewise.
2841         (find_comparison_dom_walker::before_dom_children): Likewise.
2842         (try_eliminate_compare): Likewise.
2843         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
2844         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
2846 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
2848         * doc/extend.texi (Common Type Attributes): Fix typos.
2850 2018-11-19  Martin Liska  <mliska@suse.cz>
2852         PR gcov-profile/88045
2853         * coverage.c (coverage_begin_function): Add assert.
2855 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
2857         PR other/40498
2858         * doc/extend.texi (Common Function Attributes): Document that
2859         no_instrument_function applies to -p and -pg, too.
2860         * doc/invoke.texi (Instrumentation Options): Add cross-references
2861         to docs for -p, -pg, and -finstrument-functions.
2863 2018-11-18  Alan Modra  <amodra@gmail.com>
2865         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
2866         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
2867         (GNU_USER_TARGET_CRTN): Define.
2868         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
2869         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
2871 2018-11-18  Alan Modra  <amodra@gmail.com>
2873         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
2874         !HAVE_LD_PIE variant.
2875         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
2877 2018-11-17  Nick Clifton  <nickc@redhat.com>
2878             Sandra Loosemore  <sandra@codesourcery.com>
2880         PR driver/31357
2881         * doc/invoke.texi (Overall Options): Document that --help and
2882         --help=<value> options cannot be combined.
2884 2018-11-17  Jim Wilson  <jimw@sifive.com>
2886         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
2887         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
2888         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
2889         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
2890         to step2 if frame_pointer_needed and step1 is 0.
2892 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2894         PR c++/4225
2895         * doc/cpp.texi (System Headers): Add note about implicit
2896         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
2898 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2900         PR c++/4025
2901         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
2902         -fno-implicit-templates.
2904 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
2906         PR tree-optimization/87546
2907         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
2908         min_precision variable, initially set it to orig_precision, only does
2909         something if op_type's precision is <= min_precision and update
2910         min_precision whenever calling set_op.
2912 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
2914         PR ipa/87957
2915         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
2917 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2919         * doc/invoke.texi (Option Summary): Fix whitespace and line
2920         breaks in @gccoptlist environments.
2921         (Warning Options): Likewise.
2922         (Optimize Options): Likewise.
2923         (PowerPC SPE Options): Likewise.
2924         (RS/6000 and PowerPC Options): Likewise.
2926 2018-11-16  Jeff Law  <law@redhat.com>
2928         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
2929         gcc_assert.
2931 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2933         PR middle-end/23197
2934         * doc/invoke.texi (Optimize Options): Update options enabled by
2935         fprofile-generate, -fprofile-use, and -fauto-profile.
2937 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2939         PR target/87927
2940         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
2941         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
2942         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
2943         the byte_op hook.
2944         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
2945         * targhooks.c (default_print_patchable_function_entry): Assert
2946         asm_int_op does not return a NULL string.
2947         * varasm.c (integer_asm_op): Return the op for a partial int type
2948         when the requested size does not correspond to an integer type.
2949         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
2950         * doc/tm.texi: Regenerate.
2952 2018-11-12  Jason Merrill  <jason@redhat.com>
2954         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
2956 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
2958         * config/rs6000/constraints.md (wF constraint): Remove power9
2959         fusion documentation.  Just document wF's use for power8 fusion.
2960         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
2961         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
2962         fusion support to require the upper 12 bits to be all 0's or all
2963         1's.
2964         (fusion_gpr_mem_load): Add comment.
2965         (fusion_addis_mem_combo_load): Remove power9 fusion support.
2966         (fusion_addis_mem_combo_store): Delete.
2967         (fusion_offsettable_mem_operand): Delete.
2968         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
2969         power8 fusion here.
2970         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
2971         (POWERPC_MASKS): Delete power9 fusion.
2972         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
2973         (fusion_p9_p): Delete.
2974         (expand_fusion_p9_load): Delete.
2975         (expand_fusion_p9_store): Delete.
2976         (emit_fusion_p9_load): Delete.
2977         (emit_fusion_p9_store): Delete.
2978         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
2979         fusion support.
2980         (rs6000_option_override_internal): Set power8 fusion based on
2981         whether we are tuning for power8.  Delete power9 fusion support.
2982         (rs6000_opt_masks): Delete -mpower9-fusion switch.
2983         (emit_fusion_load): Rename emit_fusion_load_store to
2984         emit_fusion_load, and drop fusion store support.  Update callers.
2985         (emit_fusion_load_store): Likewise.
2986         (emit_fusion_gpr_load): Likewise.
2987         (fusion_p9_p): Delete.
2988         (expand_fusion_p9_load): Delete.
2989         (expand_fusion_p9_store): Delete.
2990         (emit_fusion_p9_load): Delete.
2991         (emit_fusion_p9_store): Delete.
2992         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
2993         (GPR_FUSION): Delete.
2994         (FPR_FUSION): Delete.
2995         (power9 fusion peephole2s): Delete.
2996         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
2997         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
2998         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
2999         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
3000         (fusion_p9_<mode>_constant): Delete.
3001         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
3002         power9 fusion switch.
3003         * doc/md.texi (PowerPC constraints): Update wF constraint
3004         documentation for power8 fusion only.
3006 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
3008         PR rtl-optimization/87475
3009         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
3010         CROSSING_JUMP_P insns.
3011         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
3013 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
3015         PR target/88051
3016         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
3017         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
3018         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
3020 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
3022         PR middle-end/88032
3023         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
3024         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
3026 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
3028         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
3029         location_t parameter.
3030         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3031         (switch_decision_tree::do_jump_if_equal): Likewise.
3032         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
3033         of switch statement to emit_case_nodes.
3034         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
3035         set it on the newly built GIMPLE comparison statement.
3036         (switch_decision_tree::do_jump_if_equal): Likewise.
3037         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
3038         in calls to do_jump_if_equal as well as recursive calls.
3040 2018-11-16  Andi Kleen  <ak@linux.intel.com>
3042         * config/i386/i386.md: Allow memory operands to ptwrite.
3044 2018-11-16  Michael Matz  <matz@suse.de>
3046         * system.h (PRsa): New macro.
3047         (SIZE_AMOUNT): Cast number to uint64_t.
3048         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
3049         (pool_usage::dump_footer): Likewise and also use PRIu64.
3050         * bitmap.h (bitmap_usage::dump): Likewise.
3051         * ggc-common.c (ggc_usage::dump): Likewise.
3052         * ggc-page.c (ggc_print_statistics): Likewise.
3053         * input.c (dump_line_table_statistics): Likewise.
3054         * mem-stats.h (mem_usage::dump): Likewise.
3055         (mem_usage::dump_footer): Likewise.
3056         * rtl.c (dump_rtx_statistics): Likewise.
3057         * tree-cfg.c (dump_cfg_stats): Likewise.
3058         * tree-dfa.c (dump_dfa_stats): Likewise.
3059         * tree-phinodes.c (phinodes_print_statistics): Likewise.
3060         * tree-ssanames (ssanames_print_statistics): Likewise.
3061         * vec.c (vec_usage::dump): Likewise.
3062         (vec_usage::dump_footer): Likewise.
3064 2018-11-16  Richard Biener  <rguenther@suse.de>
3066         PR tree-optimization/88011
3067         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
3068         replacing set_value_range_to_undefined and
3069         set_value_range_to_varying with method calls.
3071 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
3073         * config/s390/s390.md
3074         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
3075         assembler.
3076         (*r<noxa>sbg_<mode>_sll): Likewise.
3077         (*r<noxa>sbg_<mode>_srl): Likewise.
3078         (*r<noxa>sbg_sidi_srl): New pattern.
3080 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
3082         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
3083         Default to TARGET_VXWORKS_RTP.
3084         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3085         of TARGET_VXWORKS_RTP.
3086         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3087         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
3089 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
3091         PR c++/25759
3092         * doc/extend.texi (Common Type Attributes): Make it explicit
3093         that attribute "packed" can apply to C++ classes.
3095 2018-11-15  Martin Sebor  <msebor@redhat.com>
3097         PR c++/87541
3098         PR c++/87542
3099         * tree.c (type_argument_type): New function.
3100         * tree.h (type_argument_type): Declare it.
3101         * gcc/doc/extend.texi (alloc_align): Update and clarify.
3102         (alloc_size, nonnull, sentinel): Same.
3104 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
3105             Kwok Cheung Yeung  <kcy@codesourcery.com>
3107         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
3108         int_mode_for_size fails.
3109         (vectorizable_load): Likewise.
3111 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3113         * doc/ux.texi (Group logically-related diagnostics): Move
3114         discussion of auto_diagnostic_group into this new subsection.
3115         Give an example of where this grouping is used.
3117 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3119         PR other/19165
3120         * Makefile.in (OBJS): Move json.o to...
3121         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
3122         * common.opt (fdiagnostics-format=): New option.
3123         (diagnostics_output_format): New enum.
3124         * diagnostic-format-json.cc: New file.
3125         * diagnostic.c (default_diagnostic_final_cb): New function, taken
3126         from start of diagnostic_finish.
3127         (diagnostic_initialize): Initialize final_cb to
3128         default_diagnostic_final_cb.
3129         (diagnostic_finish): Move "being treated as errors" messages to
3130         default_diagnostic_final_cb.  Call any final_cb.
3131         (default_diagnostic_finalizer): Add diagnostic_t param.
3132         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
3133         diagnostic_finalizer callback.
3134         * diagnostic.h (enum diagnostics_output_format): New enum.
3135         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
3136         (struct diagnostic_context): Add "final_cb".
3137         (default_diagnostic_finalizer): Add diagnostic_t param.
3138         (diagnostic_output_format_init): New decl.
3139         * doc/invoke.texi (-fdiagnostics-format): New option.
3140         * dwarf2out.c (gen_producer_string): Ignore
3141         OPT_fdiagnostics_format_.
3142         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
3143         * lto-wrapper.c (append_diag_options): Ignore it.
3144         * opts.c (common_handle_option): Handle it.
3146 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
3148         PR tree-optimization/88015
3149         * graphite-isl-ast-to-gimple.c
3150         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
3151         for dump_enabled_p.
3152         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
3154 2018-11-15  Richard Biener  <rguenther@suse.de>
3156         PR middle-end/88029
3157         * gimple.c (gimple_call_flags): Union flags from decl, type
3158         and call fntype.
3159         * trans-mem.c (is_tm_pure_call): Simplify.
3161 2018-11-15  Richard Biener  <rguenther@suse.de>
3163         PR tree-optimization/88031
3164         * tree-vect-loop.c (vectorizable_reduction): Move check
3165         for multiple types earlier so we get the expected dump.
3166         Simplify calls to vectorizable_condition.
3167         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
3168         * tree-vect-stmts.c (vectorizable_condition): Instead of
3169         reduc_def and reduc_index take just a flag.  Simplify
3170         code-generation now that we can rely on the defs being set up.
3171         (vectorizable_comparison): Remove unused argument.
3173 2018-11-15  Richard Biener  <rguenther@suse.de>
3175         PR tree-optimization/88030
3176         * tree-complex.c (need_eh_cleanup): New global.
3177         (update_complex_assignment): Mark blocks that need EH update.
3178         (expand_complex_comparison): Likewise.
3179         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
3180         perform EH cleanup and schedule CFG cleanup if that did anything.
3182 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
3184         PR rtl-optimization/88018
3185         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
3186         on the presence of fallthru edge, rather than if it is a USE or not.
3188 2018-11-15  Richard Biener  <rguenther@suse.de>
3190         PR middle-end/87917
3191         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
3192         analyze_subscript_affine_affine properly.
3194 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3196         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
3197         (GLIBC_DYNAMIC_LINKER): Define.
3198         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
3200 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
3202         PR tree-optimization/84648
3203         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
3204         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
3205         by calling adjust_cond_for_loop_until_wrap.
3207 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
3209         PR other/56334
3210         * doc/extend.texi (Common Function Attributes): Clarify linker
3211         restrictions on "aligned" attribute.
3212         (Common Variable Attributes): Likewise.  Mention that linker
3213         restrictions don't apply to stack-allocated variables.
3215 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
3217         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
3218         as an lvalue.
3220 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
3222         PR lto/55102
3223         PR lto/56700
3224         * doc/invoke.texi (Optimize Options): Remove bad example about
3225         interaction between -flto and -O.  Replace it with a note that
3226         you need to compile with -O and not just link.  Copy-edit -flto
3227         discussion to reduce verbiage and improve flow.
3229 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
3231         PR middle-end/59658
3232         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
3233         also suppress many optimizations.  Alphabetize option lists for
3234         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
3235         correct documentation for those options to say that.
3236         * opts.c (default_options_table): Sort table by level and option
3237         name, to make it easier to correlate to the manual.
3239 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
3241         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
3242         printing of the names of segment registers.
3244 2018-11-14  Nathan Sidwell  <nathan@acm.org>
3246         PR debug/88006
3247         PR debug/87462
3248         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
3249         type list.
3251 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
3253         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
3255 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
3257         P1236R1 - Signed integers are two's complement
3258         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
3260         PR bootstrap/86739
3261         * hash-map.h (hash_map::iterator::reference_pair): New class.
3262         (hash_map::iterator::operator*): Return it rather than std::pair.
3264 2018-11-14  Jeff Law  <law@redhat.com>
3266         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
3267         iff the operand is a constant.
3269 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
3271         * gimple-ssa-evrp-analyze.c
3272         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
3273         ignore_equivs_equal_p to equal_p.
3274         * ipa-cp.c (meet_with_1): Use equal_p instead of
3275         ignore_equivs_equal_p.
3276         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
3277         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
3278         (value_range::operator==): Remove.
3279         (value_range::operator!=): Remove.
3280         (vrp_prop::visit_stmt): Use equal_p.
3281         * tree-vrp.h (value_range): Remove operator==, operator!=,
3282         ignore_equivs_equal_p.
3283         * vr-values.c (update_value_range): Use equal_p.
3285 2018-11-14  Michael Matz  <matz@suse.de>
3287         PR middle-end/86575
3288         * gimplify.c (collect_fallthrough_labels): Add new argument,
3289         return location via that, don't modify statements.
3290         (warn_implicit_fallthrough_r): Adjust call, don't use
3291         statement location directly.
3293 2018-11-14  Richard Biener  <rguenther@suse.de>
3295         PR middle-end/87985
3296         * tree-data-ref.c (split_constant_offset): Add wrapper
3297         allocating a cache hash-map.
3298         (split_constant_offset_1): Cache results of expanding
3299         expressions from SSA def stmts.
3301 2018-11-14  Richard Biener  <rguenther@suse.de>
3303         PR middle-end/88021
3304         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
3305         to lambda_int.
3306         (lambda_vector_mult_const): Likewise.
3307         (lambda_matrix_right_hermite): Use lambda_int temporaries.
3309 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
3310             Jackson Woodruff  <jackson.woodruff@arm.com>
3312         PR 71026/tree-optimization
3313         * match.pd: Simplify floating point comparisons.
3315 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
3317         PR rtl-optimization/87817
3318         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
3319         last argument has low 8 bits clear, fold to 0.
3321 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
3323         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
3324         MAX_OFILE_ALIGNMENT as needed.
3326 2018-11-14  Richard Biener  <rguenther@suse.de>
3328         PR tree-optimization/88019
3329         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
3330         COND_EXPR arguments to non-trapping overflow.
3332 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
3334         * config/arc/arc.h (reg_class): Reorder registers classes, remove
3335         unused register classes.
3336         (REG_CLASS_NAMES): Likewise.
3337         (REG_CLASS_CONTENTS): Likewise.
3338         (FIXED_REGISTERS): Make lp_count fixed.
3339         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
3340         (PROGRAM_COUNTER_REGNO): Remove.
3341         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
3342         register classes, use constants for register numbers, remove
3343         reg_class_contents references.
3344         (arc_process_double_reg_moves): Add asserts.
3345         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
3346         lra_in_progress predicate.
3347         (arc_init_reg_tables): Remove unused register classes.
3348         (arc_register_move_cost): Likewise.
3349         (arc_preferred_reload_class): Likewise.
3350         (hwloop_optimize): Update rtx patterns involving lp_count register.
3351         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
3352         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
3353         CHEAP_CORE_REGS.  Former one will be used for LRA.
3354         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
3355         one will be used for LRA.
3356         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
3357         one will be used for LRA.
3358         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
3359         Former one will be used for LRA.
3360         ("f"): Delete constraint.
3361         ("k"): Likewise.
3362         ("e"): Likewise.
3363         ("l"): Change it from register constraint to constraint.
3364         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
3365         (movhi_insn): Likewise.
3366         (movsi_insn): Update pattern.
3367         (arc_lp): Likewise.
3368         (dbnz): Likewise.
3369         (stack_tie): Remove 'b' constraint letter.
3370         (R4_REG): Define.
3371         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
3372         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
3373         (R57_REG, R59_REG, PCL_REG): Likewise.
3374         (ILINK1_REGNUM): Renamed to ILINK1_REG.
3375         (ILINK2_REGNUM): Renamed to ILINK2_REG.
3376         (Rgp): Remove.
3377         (SP_REGS): Likewise.
3378         (Rcw): Remove unused reg classes.
3379         * config/arc/predicates.md (dest_reg_operand): Just default on
3380         register_operand predicate.
3381         (mpy_dest_reg_operand): Likewise.
3382         (move_dest_operand): Use macros instead of constants.
3383         * config/arc/arc.opt (mlra): Switch to lra as default.
3385 2018-11-14  Richard Biener  <rguenther@suse.de>
3387         PR tree-optimization/87974
3388         * tree-vect-loop.c (vectorizable_reduction): When computing
3389         the vectorized reduction PHI vector type ignore constant
3390         and external defs.
3392 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
3394         PR tree-optimization/87977
3395         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
3396         stmt, build a new one and replace the old one with it.  Formatting fix.
3397         Call release_ssa_name (x) if !has_other_use and !delete_div.
3398         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
3399         verify lhs of stmt is still def.
3401 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
3403         PR rtl-optimization/87507
3404         * lower-subreg.c (operand_for_swap_move_operator): New function.
3405         (simple_move): Strip simple operators.
3406         (find_pseudo_copy): Likewise.
3407         (resolve_operand_for_swap_move_operator): New function.
3408         (resolve_simple_move): Strip simple operators and swap operands.
3410 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
3412         PR tree-optimization/87898
3413         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
3414         (ipa_simd_modify_function_body): Remove debug stmts where the first
3415         argument was changed into a non-decl.
3417 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
3419         PR rtl-optimization/87899
3420         * lra-lives.c (start_living): Update white space in comment.
3421         (enum point_type): New.
3422         (sparseset_contains_pseudos_p): New function.
3423         (update_pseudo_point): Likewise.
3424         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
3425         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
3426         handling.  Move early exit after adding conflicts.
3427         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
3428         if regno is already live.  Remove all handling of program points.
3429         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
3430         after adding conflicts.  Remove all handling of program points and
3431         ignore_reg_for_conflicts.
3432         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
3433         and do not guard call to mark_pseudo_live.
3434         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
3435         and do not guard call to mark_pseudo_dead.
3436         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
3437         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
3438         Use new function update_pseudo_point.  Handle register copies by
3439         removing the source register from the live set.  Handle INOUT operands.
3440         Update to the next program point using the unused_set, dead_set and
3441         start_dying sets.
3442         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
3444 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
3446         * builtins.c: Replace "source_location" with "location_t".
3447         * diagnostic-show-locus.c: Likewise.
3448         * diagnostic.c: Likewise.
3449         * dumpfile.c: Likewise.
3450         * gcc-rich-location.h: Likewise.
3451         * genmatch.c: Likewise.
3452         * gimple.h: Likewise.
3453         * gimplify.c: Likewise.
3454         * input.c: Likewise.
3455         * input.h: Likewise.  Eliminate the typedef.
3456         * omp-expand.c: Likewise.
3457         * selftest.h: Likewise.
3458         * substring-locations.h (get_source_location_for_substring):
3459         Rename to...
3460         (get_location_within_string): ...this.
3461         * tree-cfg.c: Replace "source_location" with "location_t".
3462         * tree-cfgcleanup.c: Likewise.
3463         * tree-diagnostic.c: Likewise.
3464         * tree-into-ssa.c: Likewise.
3465         * tree-outof-ssa.c: Likewise.
3466         * tree-parloops.c: Likewise.
3467         * tree-phinodes.c: Likewise.
3468         * tree-phinodes.h: Likewise.
3469         * tree-ssa-loop-ivopts.c: Likewise.
3470         * tree-ssa-loop-manip.c: Likewise.
3471         * tree-ssa-phiopt.c: Likewise.
3472         * tree-ssa-phiprop.c: Likewise.
3473         * tree-ssa-threadupdate.c: Likewise.
3474         * tree-ssa.c: Likewise.
3475         * tree-ssa.h: Likewise.
3476         * tree-vect-loop-manip.c: Likewise.
3478 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
3480         * doc/extend.texi: Fix typo in the weakref description.
3482 2018-11-13  Richard Biener  <rguenther@suse.de>
3484         PR tree-optimization/86991
3485         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
3486         group building until we have successfully detected the SLP
3487         reduction.
3488         (vect_is_simple_reduction): Remove fixup code here.
3490 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
3492         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
3493         (dump_gimple_stmt): Use it.
3494         (dump_gimple_stmt_loc): Likewise.
3495         (dump_gimple_expr): Likewise.
3496         (dump_gimple_expr_loc): Likewise.
3497         (dump_generic_expr): Likewise.
3498         (dump_generic_expr_loc): Likewise.
3499         (dump_printf): Likewise.
3500         (dump_printf_loc): Likewise.
3501         (dump_dec): Likewise.
3502         (dump_dec): Likewise.
3503         (dump_hex): Likewise.
3504         (dump_symtab_node): Likewise.
3505         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
3506         Guard dump call with dump_enabled_p.
3507         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
3508         * graphite-optimize-isl.c (optimize_isl): Likewise.
3509         * graphite.c (graphite_transform_loops): Likewise.
3510         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
3511         * tree-parloops.c (parallelize_loops): Likewise.
3512         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
3513         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
3514         (vect_prune_runtime_alias_test_list): Likewise.
3515         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
3516         (vect_estimate_min_profitable_iters): Likewise.
3517         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
3518         (vect_build_slp_tree_2): Likewise.
3519         (vect_supported_load_permutation_p): Likewise.
3520         (vect_slp_analyze_operations): Likewise.
3521         (vect_slp_analyze_bb_1): Likewise.
3522         (vect_slp_bb): Likewise.
3523         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
3524         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
3525         (pass_slp_vectorize::execute): Likewise.
3526         (increase_alignment): Likewise.
3528 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
3530         PR ipa/87955
3531         * ipa-inline.c (report_inline_failed_reason): Guard calls to
3532         cl_target_option_print_diff and cl_optimization_print_diff with
3533         if (dump_file).
3535 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
3537         * doc/invoke.texi (-fsave-optimization-record): Note that the
3538         output is compressed.
3539         * optinfo-emit-json.cc: Include <zlib.h>.
3540         (optrecord_json_writer::write): Compress the output.
3542 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
3544         * tree-vrp.c (value_range_base::dump): Dump type.
3545         Do not use INF nomenclature for 1-bit types.
3546         (dump_value_range): Group all variants to common dumping code.
3547         (debug): New overloaded functions for value_ranges.
3548         (value_range_base::dump): Remove no argument version.
3549         (value_range::dump): Same.
3551 2018-11-13  Richard Biener  <rguenther@suse.de>
3553         PR tree-optimization/87931
3554         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
3555         nested cycles we support to latch computations vectorizable_reduction
3556         handles.
3558 2018-11-13  Martin Liska  <mliska@suse.cz>
3560         PR tree-optimization/87885
3561         * cfghooks.c (account_profile_record): Rename
3562         to ...
3563         (profile_record_check_consistency): ... this.
3564         Calculate missing num_mismatched_freq_in.
3565         (profile_record_account_profile): New function
3566         that calculates time and size of a function.
3567         * cfghooks.h (struct profile_record): Remove
3568         all tuples.
3569         (struct cfg_hooks): Remove after_pass flag.
3570         (account_profile_record): Rename to ...
3571         (profile_record_check_consistency): ... this.
3572         (profile_record_account_profile): New.
3573         * cfgrtl.c (rtl_account_profile_record): Remove
3574         after_pass flag.
3575         * passes.c (check_profile_consistency): Do only
3576         checking.
3577         (account_profile): Calculate size and time of
3578         function only.
3579         (pass_manager::dump_profile_report): Reformat
3580         output.
3581         (execute_one_ipa_transform_pass): Call
3582         consistency check before clean upand call account_profile
3583         after a clean up is done.
3584         (execute_one_pass): Call check_profile_consistency and
3585         account_profile instead of using after_pass flag..
3586         * tree-cfg.c (gimple_account_profile_record): Likewise.
3588 2018-11-13  Martin Liska  <mliska@suse.cz>
3590         PR sanitizer/87930
3591         * config/i386/i386.c (ix86_option_override_internal): Error
3592         about usage -mabi=ms and -fsanitize={,kernel-}address.
3594 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3596         * config/aarch64/aarch64.c
3597         (aarch64_vectorize_preferred_vector_alignment): Change return type to
3598         poly_uint64.
3599         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
3600         alignment being a poly int.
3601         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
3602         return type to poly_uint64.
3603         * target.def (default_preferred_vector_alignment): Likewise.
3604         * targhooks.c (default_preferred_vector_alignment): Likewise.
3605         * targhooks.h (default_preferred_vector_alignment): Likewise.
3606         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
3607         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
3608         poly int.
3609         (vect_update_misalignment_for_peel): Likewise.
3610         (vect_enhance_data_refs_alignment): Likewise.
3611         (vect_find_same_alignment_drs): Likewise.
3612         (vect_duplicate_ssa_name_ptr_info): Likewise.
3613         (vect_setup_realignment): Likewise.
3614         (vect_can_force_dr_alignment_p): Change alignment parameter type to
3615         poly_uint64.
3616         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
3617         mask with a compile time variable vector alignment.
3618         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
3619         int.
3620         (vect_do_peeling): Exit early if vector alignment is not constant.
3621         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
3622         a poly int.
3623         (vectorizable_store): Likewise.
3624         (vectorizable_load): Likweise.
3625         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
3626         poly_uint64.
3627         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
3628         poly int.
3629         (vect_can_force_dr_alignment_p): Change alignment parameter type to
3630         poly_uint64.
3632 2018-11-13  Richard Biener  <rguenther@suse.de>
3634         PR tree-optimization/87962
3635         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
3636         detect outer reduction for disqualifying in-loop uses.
3638 2018-11-13  Richard Biener  <rguenther@suse.de>
3640         PR tree-optimization/87967
3641         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
3642         for constants for the scalar loop.
3644 2018-11-13  Alan Modra  <amodra@gmail.com>
3646         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
3647         integer constants when -mcmodel=medium.
3649 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3651         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
3652         * config/arc/elf.h (PROFILE_HOOK): Define.
3653         * config/arc/linux.h (PROFILE_HOOK): Likewise.
3655 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3657         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
3658         jump table data in the text section.
3660 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3662         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
3663         to fit the eh_return pattern.
3664         * config/arc/arc.md (eh_return): Define.
3665         (VUNSPEC_ARC_EH_RETURN): Likewise.
3666         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
3667         new implementation.
3668         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
3671 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3673         * common/config/arc/arc-common.c (arc_option_optimization_table):
3674         Millicode optimization is default on for size optimizations.
3675         * config/arc/arc-protos.h (arc_check_multi): New function.
3676         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
3677         (ENTER_LEAVE_START_REG): Define.
3678         (ENTER_LEAVE_END_REG): Likewise.
3679         (arc_override_options): Disable millicode when long calls option
3680         is on.
3681         (arc_frame_info): Change it from int to bool.
3682         (arc_compute_frame_size): Clean up.
3683         (arc_save_restore): Remove.
3684         (frame_save_reg): New function.
3685         (frame_restore_reg): Likewise.
3686         (arc_enter_leave_p): Likewise.
3687         (arc_save_callee_saves): Likewise.
3688         (arc_restore_callee_saves): Likewise.
3689         (arc_save_callee_enter): Likewise.
3690         (arc_restore_callee_leave): Likewise.
3691         (arc_save_callee_milli): Likewise.
3692         (arc_restore_callee_milli): Likewise.
3693         (arc_expand_prologue): Reimplement to emit enter/leave
3694         instructions.
3695         (arc_expand_epilogue): Likewise.
3696         (arc_check_multi): New function.
3697         * config/arc/arc.md (push_multi_fp): New pattern.
3698         (push_multi_fp_blink): Likewise.
3699         (pop_multi_fp): Likewise.
3700         (pop_multi_fp_blink): Likewise.
3701         (pop_multi_fp_ret): Likewise.
3702         (pop_multi_fp_blink_ret): Likewise.
3703         * config/arc/arc.opt (mmillicode): Update option.
3704         (mcode-density-frame): New option.
3705         * config/arc/predicates.md (push_multi_operand): New predicate.
3706         (pop_multi_operand): Likewise.
3707         * doc/invoke.texi (ARC): Update ARC options information.
3710 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3712         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
3713         * config/arc/arc.c (operands_ok_ldd_std): New function.
3714         (mem_ok_for_ldd_std): Likewise.
3715         (gen_operands_ldd_std): Likewise.
3716         * config/arc/arc.md: Add peephole2 rules for std/ldd.
3718 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
3720         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
3721         test on stack_usage_file.
3722         (lang_dependent_init): Do not open the .su file if generating LTO.
3724 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
3726         PR rtl-optimization/87918
3727         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
3728         simplify_gen_relational rather than simplify_gen_binary.
3730 2018-11-13  Richard Biener  <rguenther@suse.de>
3732         * tree-ssanames.h (set_range_info): Use value_range_base.
3733         (get_range_info): Likewise.
3734         * tree-ssanames.c (set_range_info): Likewise.
3735         (get_range_info): Likewise.
3736         * tree-vrp.c (value_range_base::union_helper): Split
3737         out common parts of value_range[_base]::union_.
3738         (value_range_base::union_): Update.
3739         (value_range::union_): Likewise.
3740         (determine_value_range_1): Use value_range_base.
3741         (determine_value_range): Likewise.
3742         * tree-vrp.h (value_range_base::union_helper): Move ...
3743         (value_range::union_helper): ... from here.
3745 2018-11-13  Alan Modra  <amodra@gmail.com>
3747         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
3748         offset for PRE_DEC.
3749         (rs6000_secondary_reload_gpr): Don't call find_replacement.
3751 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
3753         PR middle-end/59634
3754         * doc/invoke.texi (Optimize Options): Clarify that the
3755         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
3756         apply to data cache size.
3758 2018-11-13  Alan Modra  <amodra@gmail.com>
3760         * config/rs6000/predicates.md (logical_const_operand),
3761         (logical_operand): Correct comment.
3762         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
3764 2018-11-13  Alan Modra  <amodra@gmail.com>
3766         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
3768 2018-11-13  Alan Modra  <amodra@gmail.com>
3770         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
3771         description.
3772         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
3773         covered by alternative.
3774         (movcc_internal1): Ignore h for register preference.
3775         (mov<mode>_hardfloat64): Likewise.
3776         (mov<mode>_softfloat): Ignore c, l, h for register preference.
3778 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3780         PR preprocessor/47823
3781         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
3782         documentation to...
3783         (Pragmas): ...here.
3784         * doc/extend.texi (Pragmas): Note additional pragmas documented
3785         in the CPP manual.
3787 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
3789         PR middle-end/86677
3790         PR middle-end/87528
3791         * tree-scalar-evolution.c (expression_expensive_p): Make
3792         BUILTIN POPCOUNT as expensive when backend does not define it.
3794 2018-11-12  Fredrik Noring  <noring@nocrew.org>
3796         * config.gcc: Update with-llsc defaults for MIPS r5900.
3798 2018-11-12  Martin Liska  <mliska@suse.cz>
3800         PR gcov-profile/87442
3801         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
3802         options.
3803         * doc/invoke.texi: Document them.
3804         * tree-profile.c (parse_profile_filter): New.
3805         (parse_profile_file_filtering): Likewise.
3806         (release_profile_file_filtering): Likewise.
3807         (include_source_file_for_profile): Likewise.
3808         (tree_profiling): Filter source files based on the
3809         newly added options.
3811 2018-11-12  Martin Liska  <mliska@suse.cz>
3813         PR target/87903
3814         * doc/extend.texi: Add missing values for __builtin_cpu_is and
3815         __builtin_cpu_supports for x86 target.
3817 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3819         PR middle-end/21110
3820         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
3821         have to be Pmode.
3822         (Arithmetic): Likewise for "lo_sum".
3824 2018-11-12  Renlin Li  <renlin.li@arm.com>
3826         PR target/87815
3827         * dse.c (get_stored_val): Add check for compile-time constantness
3828         of gap.
3830 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
3832         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
3833         (ARMv8_5a): New fgroup.
3834         (armv8.5-a): New arch.
3835         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
3836         sb and predres.
3837         * config/arm/arm-tables.opt: Regenerate.
3838         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
3839         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
3840         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
3841         Add matching rules for -march=armv8.5-a and extensions.
3842         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
3843         Add sb and predres to all armv8-a except armv8.5-a.
3845 2018-11-12  Richard Biener  <rguenther@suse.de>
3847         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
3848         overload for single value.
3849         (value_range[_base]::set_nonnull): New.
3850         (value_range[_base]::set_null): Likewise.
3851         (value_range): Document bitmap copying behavior, mark
3852         copy constructor and assignment operator deleted.
3853         (value_range::move): New.
3854         (value_range::set_and_canonicalize): Default bitmap to zero.
3855         (set_value_range_to_nonnull): Remove.
3856         (set_value_range_to_null): Likewise.
3857         (set_value_range): Likewise.
3858         (set_value_range_to_value): Likewise.
3859         (extract_range_from_unary_expr): Work on value_range_base.
3860         (extract_range_from_binary_expr_1): Likewise.  Rename to...
3861         (extract_range_from_binary_expr): ... this.
3862         * tree-vrp.c (value_range::update): Clear equiv bitmap
3863         if required.
3864         (value_range::move): New, move equiv bitmap.
3865         (value_range_base::set_undefined): Avoid assignment.
3866         (value_range::set_undefined): Likewise.
3867         (value_range_base::set_varying): Likewise.
3868         (value_range::set_varying): Likewise.
3869         (set_value_range): Remove.
3870         (value_range_base::set): New overload for value.
3871         (value_range::set): Likewise.
3872         (set_value_range_to_nonnull): Remove.
3873         (value_range_base::set_nonnull): New.
3874         (value_range::set_nonnull): Likewise.
3875         (set_value_range_to_null): Remove.
3876         (value_range_base::set_null): New.
3877         (value_range::set_null): Likewise.
3878         (range_is_null): Work on value_range_base.
3879         (range_is_nonnull): Likewise.
3880         (ranges_from_anti_range): Likewise.
3881         (extract_range_into_wide_ints): Likewise.
3882         (extract_range_from_multiplicative_op): Likewise.
3883         (extract_range_from_binary_expr): Likewise.  Update for API changes.
3884         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
3885         handling.
3886         (value_range::intersect_helper): Avoid copy and assignment.
3887         (value_range::union_helper): Likewise.
3888         (determine_value_range_1): Adjust.
3889         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
3890         Avoid assignment by using move.
3891         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
3892         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3893         Likewise.
3894         * tree-ssanames.c (get_range_info): Likewise.
3895         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
3896         * vr-values.c (vr_values::get_value_range): Adjust.
3897         (vr_values::update_value_range): Likewise.
3898         (symbolic_range_based_on_p): Work on value_range_base.
3899         (vr_values::extract_range_from_binary_expr): Use value_range_base.
3900         (vr_values::extract_range_from_unary_expr): Likewise.
3901         (vr_values::extract_range_from_cond_expr): Avoid assignment.
3902         (vr_values::extract_range_from_comparison): Adjust.
3903         (vr_values::check_for_binary_op_overflow): Use value_range_base.
3904         (vr_values::extract_range_basic): Adjust.
3905         (vr_values::adjust_range_with_scev): Likewise.
3906         (vr_values::vrp_visit_assignment_or_call): Likewise.
3907         (vr_values::get_vr_for_comparison): Change API to avoid
3908         assignment and copy construction.
3909         (vr_values::compare_name_with_value): Adjust accordingly.
3910         (vr_values::compare_names): Likewise.
3911         (vr_values::extract_range_from_phi_node): Avoid assignment and
3912         bogus in-place modify of equiv bitmap.
3913         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
3914         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
3915         for extract_range_from_unary_expr API change.
3916         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
3918 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
3920         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
3922 2018-11-12  Richard Biener  <rguenther@suse.de>
3924         * tree-vrp.h (value_range_base::symbolic_p,
3925         value_range_base::constant_p, value_range_base::zero_p,
3926         value_range_base::singleton_p): Move from value_range.
3927         (value_range::dump): Add.
3928         * gimple-ssa-evrp-analyze.c
3929         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
3930         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
3931         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3932         Use set_varying.
3933         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
3934         (value_range::constant_p): Likewise.
3935         (value_range::singleton_p): Likewise.
3936         (value_range_base::dump): Add.
3937         (set_value_range_to_undefined): Remove.
3938         (set_value_range_to_varying): Likewise.
3939         (range_int_cst_p): Take value_range_base argument.
3940         (range_int_cst_singleton_p): Likewise.
3941         (value_range_constant_singleton): Likewise.
3942         (vrp_set_zero_nonzero_bits): Likewise.
3943         (extract_range_from_multiplicative_op): Use set_varying.
3944         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
3945         (extract_range_from_unary_expr): Likewise.
3946         (dump_value_range_base): Change to overload of dump_value_range.
3947         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
3948         (vrp_prop::visit_stmt): Likewise.
3949         (value_range::intersect_helper): Likewise.
3950         (value_range::union_helper): Likewise.
3951         (determine_value_range_1): Likewise.
3953 2018-11-12  Richard Biener  <rguenther@suse.de>
3955         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
3956         (set_value_range_to_null): Likewise.
3957         * vr-values.c (vr_values::extract_range_from_comparison):
3958         Clear equiv for constant singleton ranges.
3960 2018-11-12  Wei Xiao  <wei3.xiao@intel.com>
3962         * config/i386/sse.md: Combine VFIXUPIMM* patterns
3963         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3964         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3965         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
3966         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3967         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3968         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
3970 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3972         PR c/69502
3973         * doc/extend.texi (Common Type Attributes): For the align type
3974         attribute, copy language about decreasing alignment from the
3975         corresponding variable attribute.
3977 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3979         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
3980         -frounding-math.
3982 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3984         PR c++/43105
3985         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
3986         -frtti and -fno-rtti code.
3988 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3990         PR c/26366
3991         * doc/extend.texi (Other Builtins): Document probability associated
3992         with __builtin_expect.
3994 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
3996         PR target/87928
3997         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
3998         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
3999         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
4000         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
4002 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4004         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
4006 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4008         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
4010 2018-11-11  Richard Biener  <rguenther@suse.de>
4012         * tree-vrp.h (class value_range_base): New base class for
4013         value_range containing all but the m_equiv member.
4014         (dump_value_range_base): Add.
4015         (range_includes_zero_p): Work on value_range_base.
4016         * tree-vrp.c (value_range_base::set): Split out base handling
4017         from...
4018         (value_range::set): this.
4019         (value_range::set_equiv): New.
4020         (value_range_base::value_range_base): New constructors.
4021         (value_range_base::check): Split out base handling from...
4022         (value_range::check): this.
4023         (value_range::equal_p): Refactor in terms of
4024         ignore_equivs_equal_p which is now member of the base.
4025         (value_range_base::set_undefined): New.
4026         (value_range_base::set_varying): Likewise.
4027         (value_range_base::dump):Split out base handling from...
4028         (value_range::dump): this.
4029         (value_range_base::set_and_canonicalize): Split out base handling
4030         from...
4031         (value_range::set_and_canonicalize): this.
4032         (value_range_base::union_): New.
4033         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
4034         for m_vr.
4035         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
4036         instead of value_range everywhere.
4037         (ipcp_vr_lattice::print): Use dump_value_range_base.
4038         (ipcp_vr_lattice::meet_with): Adjust.
4039         (ipcp_vr_lattice::meet_with_1): Likewise.
4040         (ipa_vr_operation_and_type_effects): Likewise.
4041         (propagate_vr_across_jump_function): Likewise.
4042         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
4043         (ipa_get_value_range): Likewise.
4044         (ipa_set_jfunc_vr): Likewise.
4045         (ipa_compute_jump_functions_for_edge): Likewise.
4047 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
4049         PR middle-end/65703
4050         * doc/invoke.texi (Optimize Options): Add @opindex entries
4051         for the positive forms of -fno-xxx and -mno-xxx options
4052         that were lacking them.
4054 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
4056         * combine.c (make_more_copies): Only make an intermediate copy if the
4057         dest of a move is a pseudo.
4059 2018-11-09  Maya Rashish  <coypu@sdf.org>
4061         PR target/87221
4062         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
4063         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
4065 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
4067         PR driver/41179
4068         PR middle-end/65703
4069         * doc/invoke.texi (Optimize Options): Clarify default behavior
4070         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
4072 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4074         PR tree-optimization/87940
4075         * expr.c (string_constant): Don't strip NOPS in subexpressions.
4076         Fold PLUS_EXPR correctly.
4078 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
4080         PR target/87762
4081         * config/s390/s390.c (s390_safe_relative_long_p): New function.
4082         (annotate_constant_pool_refs): Skip insns which support
4083         relative addressing.
4084         (annotate_constant_pool_refs_1): New helper function.
4085         (find_constant_pool_ref): Skip insns which support relative
4086         addression.
4087         (find_constant_pool_ref_1): New helper function.
4088         (replace_constant_pool_ref): Skip insns which support
4089         relative addressing.
4090         (replace_constant_pool_ref_1): New helper function.
4091         (s390_mainpool_start): Adapt to the new signature.
4092         (s390_mainpool_finish): Likewise.
4093         (s390_chunkify_start): Likewise.
4094         (s390_chunkify_finish): Likewise.
4095         (pass_s390_early_mach::execute): Likewise.
4096         (s390_prologue_plus_offset): Likewise.
4097         (s390_emit_prologue): Likewise.
4098         (s390_emit_epilogue): Likewise.
4100 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
4102         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
4103         but unsupported lastprivate with conditional modifier.
4105 2018-11-09  Jeff Law  <law@redhat.com>
4107         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
4108         unused argument better.  Add gcc_unreachable to silence warning.
4110 2018-11-09  Martin Sebor  <msebor@redhat.com>
4112         PR middle-end/81824
4113         * attribs.c (has_attribute): New helper function.
4114         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
4115         * attribs.h (decls_mismatched_attributes): Declare.
4116         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
4117         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
4118         * common.opt (-Wattribute-alias): Take an argument.
4119         (-Wno-attribute-alias): New option.
4120         * doc/extend.texi (Common Function Attributes): Document copy.
4121         (Common Variable Attributes): Same.
4122         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
4123         (-Wattribute-alias): Document new option argument.
4125 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
4127         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
4128         matching criteria.  Remove unused array initializer.
4130 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
4131             Jinsong Ji  <jji@us.ibm.com>
4133         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
4134         constraints by introducing a new temporary.
4135         (_mm_cvtss_si64): Likewise.
4137 2018-11-09  Martin Liska  <mliska@suse.cz>
4139         * common.opt: Add -fipa-stack-alignment flag.
4140         * doc/invoke.texi: Document it.
4141         * final.c (rest_of_clean_state): Guard stack
4142         shrinking with flag.
4144 2018-11-09  Martin Liska  <mliska@suse.cz>
4146         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
4147         to ...
4148         (ipa_discover_variable_flags): ... this.
4149         * common.opt: Come up with new flag -fipa-reference-addressable.
4150         * doc/invoke.texi: Document it.
4151         * ipa-reference.c (propagate): Call the renamed fn.
4152         * ipa-visibility.c (whole_program_function_and_variable_visibility):
4153         Likewise.
4154         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
4155         ...
4156         (ipa_discover_variable_flags): ... this.  Discover
4157         non-addressable variables only with the newly added flag.
4158         * opts.c: Enable the newly added flag with -O1 and higher
4159         optimization level.
4161 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
4163         * json.cc (selftest::test_writing_literals): Fix comment.
4165 2018-11-09  Martin Liska  <mliska@suse.cz>
4167         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
4168         string to a stack buffer.
4169         (aarch64_parse_cpu): Likewise.
4170         (aarch64_parse_tune): Likewise.
4172 2018-11-09  Richard Biener  <rguenther@suse.de>
4174         PR tree-optimization/87953
4175         * tree-vect-loop.c (vectorizable_reduction): For analysis
4176         always pass ops[0] to vectorizable_condition.
4178 2018-11-09  Stafford Horne  <shorne@gmail.com>
4179             Richard Henderson  <rth@twiddle.net>
4180             Joel Sherrill  <joel@rtems.org>
4182         * common/config/or1k/or1k-common.c: New file.
4183         * config/or1k/*: New.
4184         * config.gcc (or1k*-*-*): New.
4185         * configure.ac (or1k*-*-*): New test for openrisc tls.
4186         * configure: Regenerated.
4187         * doc/install.texi: Document OpenRISC triplets.
4188         * doc/invoke.texi: Document OpenRISC arguments.
4189         * doc/md.texi: Document OpenRISC.
4191 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
4193         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
4194         (arm7tdmi-s): Delete CPU.
4195         (arm710t): Add aliases for arm720t and arm740t.
4196         (arm720t, arm740t): Delete CPUs.
4197         (arm920t): Add aliases for arm920, arm922t and arm940t.
4198         (arm920, arm922t, arm940t): Delete CPUs.
4199         (arm10tdmi): Add alias for arm1020t.
4200         (arm1020t): Delete CPU.
4201         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
4202         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
4203         (arm10e): Add aliases for arm1020e and arm1022e.
4204         (arm1020e, arm1022e): Delete CPU.
4205         * config/arm/arm.md (generic_sched): Remove entries that are now
4206         handled by aliases.
4207         (generic_vfp): Likewise.
4208         * config/arm/arm1020e.md: Simplify tuning selection based on alias
4209         changes.
4210         * config/arm/arm-tune.md: Regenerated.
4211         * config/arm/arm-tables.opt: Regenerated.
4213 2018-11-09  Richard Biener  <rguenther@suse.de>
4215         PR tree-optimization/87621
4216         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
4217         op with only phi inputs.
4218         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
4219         (ch_base::copy_headers): Run CSE on copied loop headers.
4220         (pass_ch_vect::process_loop_p): Simplify.
4222 2018-11-09  Alexandre Oliva <oliva@adacore.com>
4224         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
4225         for non-w64 x86_64 biarch.
4227 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
4229         PR rtl-optimization/86438
4230         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
4231         of in_b for the compare if in_b is SET_DEST.
4233         PR target/87793
4234         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
4235         non-toplevel UNSPEC.
4237 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
4239         * tree-vrp.c (value_range::check): Do not access internals
4240         directly.
4241         (value_range::singleton_p): Same.
4242         (value_range::type): Same.
4243         (vrp_finalize): Use value_range API.
4245 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
4247         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
4249 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
4251         * vr-values.c (vr_values::get_value_range): Use value_range API
4252         instead of piecing together ranges.
4253         (vr_values::update_value_range): Same.
4255 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
4257         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
4258         of performing ad-hoc calculations.
4259         * tree-ssanames.c (set_range_info): New overloaded function
4260         accepting value_range &.
4261         (get_range_info): Same.
4262         * tree-ssanames.h (set_range_info_raw): Remove.
4263         (set_range_info): New prototype.
4264         (get_range_info): Same.
4265         * tree-vrp.h (value_range::null_p): Rename to zero_p.
4266         * tree-vrp.c (value_range::null_p): Same.
4268 2018-11-09  Jan Hubicka  <jh@suse.cz>
4270         * tree.c (fld_type_variant_equal_p): Test user align flag.
4271         (flt_type_variant): Copy user align flag.
4272         (fld_incomplete_type_of): Clear it.
4274 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4276         * config/arm/neon.md (div<mode>3): New pattern.
4278 2018-11-08  Andi Kleen  <ak@linux.intel.com>
4280         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
4281         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
4282         (ix86_handle_option): Handle OPT_mptwrite.
4283         * config/i386/cpuid.h (bit_PTWRITE): Add.
4284         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
4285         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
4286         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
4287         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
4288         (ix86_option_override_internal): Handle PTA_PTWRITE.
4289         (ix86_valid_target_attribute_inner_p): Define ptwrite.
4290         (def_builtin2): Force UINT64 to be 64bit only.
4291         * config/i386/i386.h (TARGET_PTWRITE): Add.
4292         (TARGET_PTWRITE_P): Add.
4293         (PTA_PTWRITE): Add.
4294         * config/i386/i386.md: Define ptwrite.
4295         * config/i386/i386.opt: Add -mptwrite.
4296         * config/i386/immintrin.h (_ptwrite64): Add.
4297         (_ptwrite32): Add
4298         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
4299         * doc/invoke.texi: Document -mptwrite.
4301 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
4303         PR rtl-optimization/87600
4304         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
4305         * lra-constraints.c (process_alt_operands): Skip illegal hard
4306         register usage.  Prefer reloading non hard register operands.
4308 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
4310         PR other/36572
4311         * doc/invoke.texi (Optimize Options): Clarify default behavior
4312         for -fno-sched-interblock and -fno-sched-spec.
4314 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
4316         * collect2.c (linker_select):  Add USE_LLD_LD.
4317         (ld_suffixes): Add ld.lld.
4318         (main): Handle -fuse-ld=lld.
4319         * common.opt (-fuse-ld=lld): New option.
4320         * doc/invoke.texi (-fuse-ld=lld): Document.
4321         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
4323 2018-11-08  Paul Koning  <ni1d@arrl.net>
4325         * config/pdp11/constraints.md: Add "Z" series constraints for use
4326         with pre-dec and post-inc addressing.
4327         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
4328         (pdp11_expand_operands): Add int argument (word count).
4329         (pdp11_sp_frame_offset): Delete.
4330         (pdp11_cmp_length): New function.
4331         (pushpop_regeq): New function.
4332         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
4333         Add hook.
4334         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
4335         frame layout.
4336         (pdp11_initial_elimination_offset): Ditto.
4337         (pdp11_expand_operands): Add word count argument.  Bugfixes.
4338         (output_move_multiple): Change how pointer adjustment is done.
4339         (pdp11_gen_int_label): Correct format.
4340         (output_ascii): Ditto.
4341         (pdp11_asm_output_var): Add code for DEC assembler case.
4342         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
4343         value.
4344         (legitimate_const_double_p): Ditto.
4345         (pdp11_register_move_cost): Adjust for new register classes.
4346         (pdp11_regno_reg_class): Ditto.
4347         (expand_block_move): Delete.
4348         (pushpop_regeq): New function.
4349         (pdp11_legitimate_address_p): Bugfix in check for constant
4350         offset.
4351         (pdp11_sp_frame_offset): Delete.
4352         (pdp11_reg_save_size): New helper function for new frame layout.
4353         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
4354         (pdp11_expand_shift): Bugfix in check for constant shift count.
4355         (pdp11_shift_length): Ditto.
4356         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
4357         (pdp11_cmp_length): New function.
4358         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
4359         some compile options.
4360         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
4361         (CALL_USED_REGISTERS): Ditto.
4362         (ELIMINABLE_REGS): Ditto.
4363         (REGISTER_NAMES): Ditto.
4364         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
4365         constraints.
4366         (REG_CLASS_NAMES): Ditto.
4367         (REG_CLASS_CONTENTS): Ditto.  Also remove
4368         HARD_FRAME_POINTER_REGNUM.
4369         (CPU_REG_CLASS): New macro.
4370         (CLASS_MAX_NREGS): Adjust for new register classes.
4371         (FUNCTION_PROFILER): Make no-op.
4372         (may_call_alloca): Remove unused declaration.
4373         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
4374         (ASM_OUTPUT_SKIP): Fix format.
4375         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
4376         (HARD_FRAME_POINTER_REGNUM): Remove.
4377         (return): Delete.
4378         (*rts): Rename.  Remove epilogue related checks.
4379         (cmpsi, cmpdi): New insn.
4380         (cbranch<mode>4): Change to apply to SI and DI modes as well.
4381         (mov<mode>): Change constraints to enforce that push/pop
4382         destination cannot use the same register as source.
4383         (*mov<mode><cc_cc>): Ditto.
4384         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
4385         at assembly output rather than as RTL expander.
4386         (zero_extendqihi2): Bugfix in check for same registers.
4387         (adddi3_nocc): Bugfix in check for constant operand.
4388         (addsi3_nocc): Ditto.
4389         (subdi3_nocc): Ditto.
4390         (subsi3_nocc): Ditto.
4391         (negdi2_nocc): Copy input to pdp11_expand_operands.
4392         (negsi2_nocc): Ditto.
4393         (bswap2_nocc): Ditto.
4394         * config/pdp11/pdp11.opt (mlra): Fix documentation.
4395         * config/pdp11/t-pdp11: Use -Os.
4397 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
4399         * config/arm/parsecpu.awk (/alias/): New parsing rule.
4400         (/begin cpu/): Check that the cpu name hasn't been previously defined.
4401         (gen_comm_data): Print out CPU alias tables.
4402         (check_cpu): Match aliases when checking the CPU name.
4403         * config/arm/arm-protos.h (cpu_alias): New structure.
4404         (cpu_option): Add entry for aliases.
4405         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
4406         strongarm1100 and strongarm1110.
4407         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
4408         (config/arm/arm-generic.md): Remove redundant references to
4409         strongarm110, strongarm1100 and strongarm1110.
4410         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4411         Scan aliases for additional hints.
4412         (arm_parse_cpu_option_name): Also match a cpu name against the list
4413         of aliases.
4414         * config/arm/arm-tables.opt: Regenerated.
4415         * config/arm/arm-tune.md: Regenerated.
4417 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
4419         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
4420         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
4421         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
4422         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
4423         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
4424         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
4425         * gengtype.c (open_base_files): Add omp-general.h.
4426         * gimple.c (gimple_build_omp_critical):
4427         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
4428         gimple_omp_taskgroup_set_clauses.
4429         (gimple_build_omp_atomic_load): Add mo argument, call
4430         gimple_omp_atomic_set_memory_order.
4431         (gimple_build_omp_atomic_store): Likewise.
4432         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
4433         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
4434         instead of GSS_OMP.
4435         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
4436         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
4437         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
4438         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
4439         different value for GF_OMP_ATOMIC_NEED_VALUE.
4440         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
4441         comments.
4442         (struct gimple_statement_omp_single_layout): And remove here.
4443         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
4444         than gimple_statement_omp_single_layout.
4445         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
4446         GIMPLE_OMP_TEAMS.
4447         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
4448         (gimple_omp_subcode): Formatting fix.
4449         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
4450         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
4451         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
4452         gimple_omp_teams_host, gimple_omp_teams_set_host,
4453         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
4454         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
4455         gimple_omp_taskgroup_set_clauses): New inline functions.
4456         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
4457         (gimple_build_omp_atomic_store): Likewise.
4458         (gimple_omp_atomic_seq_cst_p): Remove.
4459         (gimple_omp_atomic_memory_order): New function.
4460         (gimple_omp_atomic_set_seq_cst): Remove.
4461         (gimple_omp_atomic_set_memory_order): New function.
4462         (gimple_build_omp_taskgroup): Add clauses argument.
4463         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
4464         (dump_gimple_omp_task): Print taskwait with depend clauses.
4465         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
4466         dump_omp_atomic_memory_order.
4467         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
4468         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
4469         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
4470         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
4471         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
4472         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
4473         ORT_UNTIED_TASKLOOP enumerators.
4474         (enum gimplify_defaultmap_kind): New.
4475         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
4476         target_map_pointers_as_0len_arrays members, add defaultmap.
4477         (new_omp_context): Initialize defaultmap member.
4478         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4479         (maybe_fold_stmt): Don't fold even in host teams regions.
4480         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
4481         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
4482         ctx->omp_firstprivatize_variable.
4483         (omp_add_variable): Don't add private/firstprivate for VLAs in
4484         ORT_TASKGROUP.
4485         (omp_default_clause): Print "taskloop" rather than "task" if
4486         ORT_*TASKLOOP.
4487         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4488         Handle new defaultmap clause kinds.
4489         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
4490         iterator to be lastprivate or private.  Fix up diagnostics if linear
4491         is used on collapse>1 simd iterator.
4492         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4493         (gimplify_omp_depend): New function.
4494         (gimplify_scan_omp_clauses): Add shared clause on parallel for
4495         combined parallel master taskloop{, simd} if taskloop has
4496         firstprivate, lastprivate or reduction clause.  Handle
4497         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
4498         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
4499         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
4500         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
4501         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
4502         lastprivate.
4503         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
4504         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
4505         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
4506         OMP_CLAUSE_{TASK,IN}_REDUCTION.
4507         (gimplify_omp_task): Handle taskwait with depend clauses.
4508         (gimplify_omp_for): Add shared clause on parallel for combined
4509         parallel master taskloop{, simd} if taskloop has firstprivate,
4510         lastprivate or reduction clause.  Use ORT_TASKLOOP or
4511         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
4512         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
4513         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
4514         __for_end and __for_range temporaries on OMP_PARALLEL for
4515         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
4516         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
4517         sandwiched in between two taskloops.
4518         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
4519         instead of ctx->omp_firstprivatize_variable.
4520         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
4521         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
4522         host teams, use gimplify_and_return_first etc. for body like
4523         for target or target data constructs, and at the end call
4524         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
4525         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
4526         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
4527         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
4528         gimple_omp_atomic_set_seq_cst calls.
4529         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
4530         case, handle taskgroup clauses.
4531         * lto-streamer-out.c (hash_tree): Handle
4532         OMP_CLAUSE_{TASK,IN}_REDUCTION.
4533         * Makefile.in (GTFILES): Add omp-general.h.
4534         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
4535         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
4536         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
4537         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
4538         BUILT_IN_GOMP_LOOP_DOACROSS_START,
4539         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
4540         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
4541         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
4542         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
4543         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
4544         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
4545         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
4546         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
4547         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
4548         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
4549         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
4550         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
4551         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
4552         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
4553         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
4554         * omp-expand.c (workshare_safe_to_combine_p): Return false for
4555         non-worksharing loops.
4556         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
4557         (determine_parallel_type): Don't combine parallel with worksharing
4558         which has _reductemp_ clause.
4559         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
4560         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
4561         if there is nonmonotonic modifier or if there is no modifier and no
4562         ordered clause.  For dynamic and guided schedule without monotonic
4563         and nonmonotonic modifier, default to nonmonotonic.
4564         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
4565         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
4566         task reductions.
4567         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
4568         there are any reduction clauses.
4569         (expand_taskwait_call): New function.
4570         (expand_teams_call): New function.
4571         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
4572         expand_teams_call for it.  Formatting fix.  Handle taskwait with
4573         depend clauses.
4574         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
4575         of worksharing loops with task reductions.
4576         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
4577         expansion of worksharing loops with task reductions.
4578         (expand_omp_sections): Handle expansion of sections with task
4579         reductions.
4580         (expand_omp_synch): For host teams call expand_omp_taskreg.
4581         (omp_memory_order_to_memmodel): New function.
4582         (expand_omp_atomic_load, expand_omp_atomic_store,
4583         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
4584         instead of gimple_omp_atomic_seq_cst_p.
4585         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
4586         depend clauses as a standalone directive.
4587         * omp-general.c (enum omp_requires): New variable.
4588         (omp_extract_for_data): Initialize have_reductemp member.  Allow
4589         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
4590         GT_EXPR loops depending on incr sign.  Formatting fixes.
4591         * omp-general.h (struct omp_for_data): Add have_reductemp member.
4592         (enum omp_requires): New enum.
4593         (omp_requires_mask): Declare.
4594         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
4595         Fix comment typos.
4596         * omp-low.c (struct omp_context): Add task_reductions and
4597         task_reduction_map fields.
4598         (is_host_teams_ctx): New function.
4599         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
4600         (use_pointer_for_field): Use is_global_var instead of
4601         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
4602         in outer contexts.
4603         (build_outer_var_ref): Ignore taskgroup outer contexts.
4604         (delete_omp_context): Release task_reductions and task_reduction_map.
4605         (scan_sharing_clauses): Don't add any fields for reduction clause on
4606         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
4607         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
4608         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
4609         Handle OMP_CLAUSE_NONTEMPORAL.
4610         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
4611         needed.
4612         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
4613         clauses with task modifier.
4614         (scan_omp_task): Handle taskwait with depend clauses.
4615         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
4616         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
4617         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
4618         (scan_omp_for): Fix comment formatting.
4619         (scan_omp_teams): Handle host teams constructs.
4620         (check_omp_nesting_restrictions): Allow teams with no outer
4621         OpenMP context.  Adjust diagnostics for teams strictly nested into
4622         some explicit OpenMP construct other than target.  Allow OpenMP atomics
4623         inside of simd regions.
4624         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
4625         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
4626         taskreg_nesting_level while scanning host teams construct.
4627         (task_reduction_read): New function.
4628         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
4629         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
4630         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
4631         second argument create_tmp_var if it is NULL.  Don't ignore shared
4632         clauses in is_host_teams_ctx contexts.  Handle
4633         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
4634         clauses.
4635         (lower_reduction_clauses): Ignore reduction clauses with task
4636         modifier.  Remove second argument create_tmp_var if it is NULL.
4637         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
4638         (lower_send_clauses): Ignore reduction clauses with task modifier.
4639         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
4640         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
4641         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
4642         rely that it is the last stmt in body so far.  Ignore outer taskgroup
4643         contexts.
4644         (omp_task_reductions_find_first, omp_task_reduction_iterate,
4645         lower_omp_task_reductions): New functions.
4646         (lower_omp_sections): Handle reduction clauses with taskgroup
4647         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
4648         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
4649         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
4650         modifiers.
4651         (lower_omp_taskgroup): Handle taskgroup reductions.
4652         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
4653         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
4654         (lower_depend_clauses): If there are any
4655         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
4656         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
4657         seen, assume lowering is done already and return early.  Set kind
4658         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
4659         (lower_omp_taskreg): Handle reduction clauses with task modifier on
4660         parallel construct.  Handle reduction clause on taskloop construct.
4661         Handle taskwait with depend clauses.
4662         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
4663         for host teams constructs.
4664         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
4665         nontemporal and _reductemp_ clause entries.
4666         (omp_clause_code_name): Likewise.
4667         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
4668         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
4669         * tree-core.h (enum omp_clause_code): Add
4670         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
4671         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
4672         (struct tree_base): Add omp_atomic_memory_order field into union.
4673         Remove OMP_ATOMIC_SEQ_CST comment.
4674         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
4675         and OMP_CLAUSE_DEPEND_DEPOBJ.
4676         (struct tree_omp_clause): Add subcode.defaultmap_kind.
4677         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
4678         OpenMP constructs with body and clauses operands.
4679         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
4680         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
4681         (OMP_TASKGROUP_CLAUSES): Define.
4682         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
4683         OMP_CLAUSE__LOOPTEMP_.
4684         (OMP_ATOMIC_SEQ_CST): Remove.
4685         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
4686         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
4687         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
4688         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
4689         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
4690         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
4691         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
4692         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
4693         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
4694         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
4695         Define.
4696         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
4697         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4698         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
4699         (convert_local_omp_clauses): Likewise.  Remove useless test.
4700         * tree-parloops.c (create_call_for_reduction_1): Pass
4701         OMP_MEMORY_ORDER_RELAXED as new argument to
4702         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
4703         * tree-pretty-print.c (dump_omp_iterators): New function.
4704         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
4705         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
4706         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
4707         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
4708         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
4709         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
4710         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
4711         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
4712         (dump_omp_atomic_memory_order): New function.
4713         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
4714         taskwait with depend clauses.
4715         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
4716         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
4717         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
4718         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
4719         write_ts_omp_clause_tree_pointers): Likewise.
4721 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4723         PR ipa/86395
4724         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
4725         "-missed", "-note", and "-all" sub-options.
4726         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
4727         API.
4728         (can_early_inline_edge_p): Likewise.
4729         (want_early_inline_function_p): Likewise.
4730         (want_inline_self_recursive_call_p): Likewise.
4731         (recursive_inlining): Likewise.
4732         (inline_small_functions): Likewise.
4733         (flatten_function): Likewise.
4734         (ipa_inline): Likewise.
4735         (inline_always_inline_functions): Likewise.
4736         (early_inline_small_functions): Likewise.
4737         (early_inliner): Likewise.
4738         * tree-inline.c (expand_call_inline): Likewise.
4740 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4742         * pretty-print.c (pp_format): Handle %f.
4743         (selftest::test_pp_format): Add test of %f.
4744         * pretty-print.h (pp_double): New macro.
4746 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4748         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
4749         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
4750         (dump_pretty_printer::decode_format): Implement "%C" for
4751         cgraph_node *.
4752         (selftest::test_capture_of_dump_calls): Rename "where" to
4753         "stmt_loc".  Convert test_decl to a function decl and set its
4754         location.  Add a symbol_table_test RAII instance and a
4755         cgraph_node, using it to test "%C" and dump_symtab_node.
4757 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
4759         PR middle-end/87916
4760         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
4762 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4764         * cgraph.c: Include "selftest.h".
4765         (saved_symtab): New variable.
4766         (selftest::symbol_table_test::symbol_table_test): New ctor.
4767         (selftest::symbol_table_test::~symbol_table_test): New dtor.
4768         (selftest::test_symbol_table_test): New test.
4769         (selftest::cgraph_c_tests): New.
4770         * cgraph.h (saved_symtab): New decl.
4771         (selftest::symbol_table_test): New class.
4772         * selftest-run-tests.c (selftest::run_tests): Call
4773         selftest::cgraph_c_tests.
4774         * selftest.h (selftest::cgraph_c_tests): New decl.
4776 2018-11-08  Richard Biener  <rguenther@suse.de>
4778         * tree-data-ref.h (lambda_int): New typedef.
4779         (lambda_vector_gcd): Adjust.
4780         (lambda_vector_new): Likewise.
4781         (lambda_matrix_new): Likewise.
4782         * tree-data-ref.c  (print_lambda_vector): Adjust.
4784 2018-11-08  Richard Biener  <rguenther@suse.de>
4786         PR tree-optimization/87929
4787         * tree-complex.c (expand_complex_comparison): Clean EH.
4789 2018-11-08  Martin Liska  <mliska@suse.cz>
4791         * doc/extend.texi: Reword.
4792         * predict.c (expr_expected_value_1): Likewise.
4794 2018-11-08  Richard Biener  <rguenther@suse.de>
4796         PR tree-optimization/87913
4797         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
4798         of extreme values to ordered comparisons.
4800 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4802         PR middle-end/42726
4803         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
4805 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4807         * doc/invoke.texi: Remove leading dash from @opindex entries
4808         throughout the file.
4810 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4812         PR driver/80828
4813         * doc/invoke.texi (Option Summary): Add -e and --entry.
4814         (Link Options): Likewise.
4816 2018-11-07  Nathan Sidwell  <nathan@acm.org>
4818         PR 87926
4819         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
4820         --disable-checking bootstrap.
4822 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
4824         * configure: Regenerated.
4826 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4828         PR c/87691
4829         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
4830         to the mode of the widest field iff the widest field has mode class
4831         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
4832         reference.
4834 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
4836         * common.opt: Add -Wattribute-warning.
4837         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
4838         * expr.c (expand_expr_real_1): Add new attribute to warning_at
4839         call to allow user configure behavior of "warning" attribute.
4841 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
4843         * target.def: Put @: after every vs., e.g., and i.e. where it is
4844         followed by whitespace.
4845         * doc/extend.texi: Ditto.
4846         * doc/fragments.texi: Ditto.
4847         * doc/gimple.texi: Ditto.
4848         * doc/implement-c.texi: Ditto.
4849         * doc/install.texi: Ditto.
4850         * doc/invoke.texi: Ditto.
4851         * doc/md.texi: Ditto.
4852         * doc/plugins.texi: Ditto.
4853         * doc/rtl.texi: Ditto.
4854         * doc/sourcebuild.texi: Ditto.
4855         * doc/tm.texi.in: Ditto.
4856         * doc/ux.texi: Ditto.
4857         * doc/tm.texi: Regenerate.
4859 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4861         * config/arm/arm-cpus.in (ares): New entry.
4862         * config/arm/arm-tables.opt: Regenerate.
4863         * config/arm/arm-tune.md: Likewise.
4864         * doc/invoke.texi (ARM Options): Document ares.
4866 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4868         * config/aarch64/aarch64-cores.def (ares): Define.
4869         * config/aarch64/aarch64-tune.md: Regenerate.
4870         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
4872 2018-11-07  Jan Hubicka  <jh@suse.cz>
4874         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
4875         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
4876         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
4877         TYPE_DECL.
4879 2018-11-07  Richard Biener  <rguenther@suse.de>
4881         PR tree-optimization/87914
4882         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
4883         of nested cycles.
4884         (vectorizable_reduction): Handle shifts and rotates by dispatching
4885         to vectorizable_shift.
4886         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
4887         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
4888         def cases.
4889         (vectorizable_shift): Export and handle being called as
4890         vect_nested_cycle.
4891         (vect_analyze_stmt): Call vectorizable_shift after
4892         vectorizable_reduction.
4893         * tree-vectorizer.h (vectorizable_shift): Declare.
4895 2018-11-07  Jan Hubicka  <jh@suse.cz>
4897         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
4898         than const decls in TREE_VALUE of enum.
4899         (dump_type_inheritance_graph): Improve duplicate dumping.
4900         (free_enum_values): New.
4901         (build_type_inheritance_graph): Use it.
4902         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
4903         which are not main variants or not ODR types.
4904         (verify_type_variant): Expect variants to have no TYPE_VALUES.
4906 2018-11-07  Richard Biener  <rguenther@suse.de>
4908         * ipa-inline.c (want_inline_small_function_p): Compute
4909         big_speedup_p lazily and last.
4911 2018-11-07  Jan Hubicka  <jh@suse.cz>
4913         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
4914         building incomplete variant of complete type.
4915         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
4916         variant of complete type.
4918 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4920         * config/mips/mips.c: Fix typo in documentation of
4921         mips_loongson_ext2_prefetch_cookie.
4922         (mips_option_override): fix brain twister logical.
4923         * config/mips/mips.h: Fix typo in documentation of
4924         ISA_HAS_CTZ_CTO and define pattern.
4925         * config/mips/mips.md (prefetch): Hoist EXT2 above
4926         the 2EF/EXT block.
4927         (prefetch_indexed): Hoist EXT2 above the EXT block.
4929 2018-11-07  Jan Hubicka  <jh@suse.cz>
4931         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
4932         return and parameter types of function and method types.
4933         (free_lang_data_in_cgraph): Update.
4935 2018-11-07  Martin Liska  <mliska@suse.cz>
4937         PR rtl-optimization/87868
4938         * postreload-gcse.c (eliminate_partially_redundant_load): Set
4939         threshold to max_count if we would overflow.
4940         * profile-count.h: Make max_count a public constant.
4942 2018-11-07  Martin Liska  <mliska@suse.cz>
4944         * mem-stats.h: Fix GNU coding style.
4946 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4948         * config/mips/gs264e.md: New.
4949         * config/mips/mips-cpus.def: Define gs264e.
4950         * config/mips/mips-tables.opt: Regenerate.
4951         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4952         gs264e.
4953         (mips_issue_rate): Add support for gs264e.
4954         (mips_multipass_dfa_lookahead): Likewise.
4955         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
4956         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
4957         (MIPS_ASE_MSA_SPEC): New.
4958         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
4959         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
4960         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
4961         * config/mips/mips.md: Include gs264e.md.
4962         (processor): Add gs264e.
4963         * config/mips/mips.opt (MSA): Use Mask instead of Var.
4964         * doc/invoke.texi: Add gs264e to supported architectures.
4966 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4968         * config/mips/gs464e.md: New.
4969         * config/mips/mips-cpus.def: Define gs464e.
4970         * config/mips/mips-tables.opt: Regenerate.
4971         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4972         gs464e.
4973         (mips_issue_rate): Add support for gs464e.
4974         (mips_multipass_dfa_lookahead): Likewise.
4975         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
4976         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
4977         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
4978         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
4979         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
4980         * config/mips/mips.md: Include gs464e.md.
4981         (processor): Add gs464e.
4982         * doc/invoke.texi: Add gs464e to supported architectures.
4984 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4986         * config/mips/loongson3a.md: Rename to ...
4987         * config/mips/gs464.md: ... here.
4988         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
4989         as an alias of gs464 processor.
4990         * config/mips/mips-tables.opt: Regenerate.
4991         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
4992         instead of PROCESSOR_LOONGSON_3A.
4993         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
4994         TUNE_LOONGSON_3A.
4995         (mips_option_override): Enable MMI and EXT for gs464.
4996         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
4997         Rename TUNE_LOONGSON_3A to TUNE_GS464.
4998         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
4999         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
5000         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
5001         TARGET_LOONGSON_3A.
5002         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
5003         (processor): Add gs464;
5004         * doc/invoke.texi: Add gs464 to supported architectures.
5006 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5008         * config/mips/mips-protos.h
5009         (mips_loongson_ext2_prefetch_cookie): New prototype.
5010         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
5011         (mips_option_override): Enable TARGET_LOONGSON_EXT when
5012         TARGET_LOONGSON_EXT2 is true.
5013         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
5014         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
5015         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
5016         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
5017         TARGET_LOONGSON_EXT2.
5018         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
5019         (define_insn "ctz<mode>2"): New insn pattern.
5020         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
5021         (define_insn "prefetch_indexed_<mode>"): Include
5022         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
5023         * config/mips/mips.opt (-mloongson-ext2): Add option.
5024         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
5026 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5028         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
5029         __mips_loongson_ext.
5030         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
5031         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
5032         -mloongson-ext.
5033         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
5034         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
5035         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
5036         instead of TARGET_LOONGSON_3A.
5037         * config/mips/mips.opt (-mloongson-ext): Add option.
5038         * gcc/doc/invoke.texi (-mloongson-ext): Document.
5040 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
5042         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
5043         * config/mips/loongson.md: Move to ...
5044         * config/mips/loongson-mmi.md: here; Adjustment.
5045         * config/mips/loongson.h: Move to ...
5046         State as deprecated. Include loongson-mmiintrin.h for back
5047         compatibility and warning.
5048         * config/mips/loongson-mmiintrin.h: ... here.
5049         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
5050         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
5051         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5052         (mips_option_override): Make sure MMI use hard float;
5053         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
5054         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
5055         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
5056         TARGET_LOONGSON_VECTORS.
5057         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
5058         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
5059         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
5060         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
5061         -mloongson-mmi.
5062         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
5063         TARGET_LOONGSON_VECTORS.
5064         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
5065         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5066         (Loongson MMI patterns): Include loongson-mmi.md instead of
5067         loongson.md.
5068         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
5069         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
5071 2018-11-07  Richard Biener  <rguenther@suse.de>
5073         PR lto/87906
5074         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
5075         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
5077 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
5079         PR rtl-optimization/87874
5080         * lra.c (lra_substitute_pseudo): Do not create a subreg for
5081         const wide ints.
5083 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
5085         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
5086         if not in indexed or indirect form.
5087         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
5088         (bswap<mode>2_store): Ditto.
5090 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
5092         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
5093         the UNSPEC.
5095 2018-11-06  Richard Biener  <rguenther@suse.de>
5097         PR tree-optimization/86850
5098         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
5099         instead of src.m_vec.
5101 2018-11-06  Jan Hubicka  <jh@suse.cz>
5103         * tree.c (fld_simplified_type_name): Break out form ...
5104         (free_lang_data_in_type): ... here.
5105         (fld_type_variant_equal_p): Use it.
5107 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5109         * config/default-d.c: Include memmodel.h.
5111         * config/sol2-d.c: New file.
5112         * config/t-sol2 (sol2-d.o): New rule.
5113         * config.gcc <*-*-solaris2*>: Set d_target_objs,
5114         target_has_targetdm.
5116 2018-11-06  Jan Hubicka  <jh@suse.cz>
5118         * tree.c (fld_type_variant): Also copy alignment; be sure that
5119         new variant is equal.
5121 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5123         PR target/87762
5124         * config/s390/s390.md: Add relative_long attribute.
5126 2018-11-06  Jan Hubicka  <jh@suse.cz>
5128         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
5129         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
5130         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
5131         stream TYPE_NEEDS_CONSTRUCTING.
5132         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
5133         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
5135 2018-11-06  Richard Biener  <rguenther@suse.de>
5137         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
5138         dump-scope ...
5139         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
5141 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5143         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
5144         REG_CLASS_CONTENTS[GEN_REGS].
5145         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
5147 2018-11-06  Jan Hubicka  <jh@suse.cz>
5149          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
5151 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5153         PR sanitizer/80953
5154         * config/sol2.h (ASAN_CC1_SPEC): Define.
5155         (LD_WHOLE_ARCHIVE_OPTION): Define.
5156         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
5157         (ASAN_REJECT_SPEC): Provide default.
5158         (LIBASAN_EARLY_SPEC): Define.
5159         (LIBTSAN_EARLY_SPEC): Define.
5160         (LIBLSAN_EARLY_SPEC): Define.
5161         * config/i386/sol2.h (CC1_SPEC): Redefine.
5162         (ASAN_REJECT_SPEC): Define.
5164         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
5165         (TARGET_ASAN_SHADOW_OFFSET): Define.
5166         (sparc_asan_shadow_offset): New function.
5167         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
5168         (ASAN_REJECT_SPEC): Define.
5170 2018-11-06  Jan Hubicka  <jh@suse.cz>
5172         * tree.c (fld_type_variant): Copy canonical type.
5173         (fld_incomplete_type_of): Check that canonical types looks sane;
5174         copy canonical type.
5175         (verify_type): Accept when incomplete type has complete canonical type.
5177 2018-11-06  Jan Hubicka  <jh@suse.cz>
5179         * tree.c (free_lang_data): Reset overwite_assembler_name,
5180         print_xnode, print_decl, print_type and print_identifier of
5181         langhooks.
5183 2018-11-06  Richard Biener  <rguenther@suse.de>
5185         PR tree-optimization/87889
5186         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5187         Do nothing if old and new arg are the same
5189 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
5191         PR target/87723
5192         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
5193         attributes for operands 3 and 4.
5195 2018-11-06  Richard Biener  <rguenther@suse.de>
5197         PR middle-end/18041
5198         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
5199         matching bitfield insertion.
5201 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
5203         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
5204         (reg_next_debug_use): New.
5205         (attempt_change): Propagate adjusted expression into affected
5206         debug insns.
5207         (merge_in_block): Track uses in debug insns.
5208         (pass_inc_dec::execute): Allocate and release
5209         reg_next_debug_use.
5211 2018-11-06  Wei Xiao  <wei3.xiao@intel.com>
5213         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
5214         (_mm512_fixupimm_round_pd): Update parameters and builtin.
5215         (_mm512_maskz_fixupimm_round_pd): Ditto.
5216         (_mm512_fixupimm_round_ps): Ditto.
5217         (_mm512_maskz_fixupimm_round_ps): Ditto.
5218         (_mm_fixupimm_round_sd): Ditto.
5219         (_mm_maskz_fixupimm_round_sd): Ditto.
5220         (_mm_fixupimm_round_ss): Ditto.
5221         (_mm_maskz_fixupimm_round_ss): Ditto.
5222         (_mm512_fixupimm_pd): Ditto.
5223         (_mm512_maskz_fixupimm_pd): Ditto.
5224         (_mm512_fixupimm_ps): Ditto.
5225         (_mm512_maskz_fixupimm_ps): Ditto.
5226         (_mm_fixupimm_sd): Ditto.
5227         (_mm_maskz_fixupimm_sd): Ditto.
5228         (_mm_fixupimm_ss): Ditto.
5229         (_mm_maskz_fixupimm_ss): Ditto.
5230         (_mm512_mask_fixupimm_round_pd): Update builtin.
5231         (_mm512_mask_fixupimm_round_ps): Ditto.
5232         (_mm_mask_fixupimm_round_sd): Ditto.
5233         (_mm_mask_fixupimm_round_ss): Ditto.
5234         (_mm512_mask_fixupimm_pd): Ditto.
5235         (_mm512_mask_fixupimm_ps): Ditto.
5236         (_mm_mask_fixupimm_sd): Ditto.
5237         (_mm_mask_fixupimm_ss): Ditto.
5238         * config/i386/avx512vlintrin.h:
5239         (_mm256_fixupimm_pd): Update parameters and builtin.
5240         (_mm256_maskz_fixupimm_pd): Ditto.
5241         (_mm256_fixupimm_ps): Ditto.
5242         (_mm256_maskz_fixupimm_ps): Ditto.
5243         (_mm_fixupimm_pd): Ditto.
5244         (_mm_maskz_fixupimm_pd): Ditto.
5245         (_mm_fixupimm_ps): Ditto.
5246         (_mm_maskz_fixupimm_ps): Ditto.
5247         (_mm256_mask_fixupimm_pd): Update builtin.
5248         (_mm256_mask_fixupimm_ps): Ditto.
5249         (_mm_mask_fixupimm_pd): Ditto.
5250         (_mm_mask_fixupimm_ps): Ditto.
5251         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
5252         * config/i386/i386-builtin.def: Update builtin definitions.
5253         * config/i386/i386.c: Handle new builtin types and remove useless ones.
5254         * config/i386/sse.md: Update VFIXUPIMM* patterns.
5255         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5256         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5257         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
5258         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5259         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5260         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
5261         * config/i386/subst.md:
5262         (round_saeonly_sd_mask_operand4): Add new subst_attr.
5263         (round_saeonly_sd_mask_op4): Ditto.
5264         (round_saeonly_expand_operand5): Ditto.
5265         (round_saeonly_expand): Update.
5267 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
5269         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
5271 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
5273         PR rtl-optimization/87871
5274         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
5276 2018-11-05  Paul Koning  <ni1d@arrl.net>
5278         * doc/sourcebuild.texi (target attributes): Document new "inf"
5279         effective target keyword.
5281 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
5283         * config/s390/s390.c (s390_register_move_cost): Increase costs for
5284         moves involving the CC reg.
5286 2018-11-05  Richard Biener  <rguenther@suse.de>
5288         PR tree-optimization/87873
5289         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
5290         argument.
5291         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
5292         * tree-vect-loop.c (vect_transform_loop): When splitting the
5293         loop exit also create forwarder PHIs for constants.
5294         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5295         Handle constant to_arg, add extra checking we match up the correct
5296         PHIs.
5298 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
5300         * config/s390/s390.md: QImode and HImode for load on condition.
5302 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
5304         * config/s390/predicates.md: Fix typo.
5305         * config/s390/s390.md: Allow immediates for load on condition.
5307 2018-11-05  Martin Liska  <mliska@suse.cz>
5309         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
5310         * bitmap.h (struct bitmap_usage): Likewise.
5311         * ggc-common.c (SCALE): Remove.
5312         (LABEL): Likewise.
5313         (struct ggc_usage): Use SIZE_AMOUNT. And update
5314         compare method.
5315         * ggc-page.c (SCALE): Remove.
5316         (STAT_LABEL): Remove.
5317         (ggc_print_statistics): Use SIZE_AMOUNT.
5318         * gimple.h (SCALE): Remove.
5319         (LABEL): Likewise.
5320         * input.c (ONE_K): Remove.
5321         (ONE_M): Likewise.
5322         (SCALE): Likewise.
5323         (STAT_LABEL): Likewise.
5324         (FORMAT_AMOUNT): Likewise.
5325         (dump_line_table_statistics): Use SIZE_AMOUNT.
5326         * mem-stats.h (struct mem_usage): Likewise.
5327         * rtl.c (dump_rtx_statistics): Likewise.
5328         (rtx_alloc_counts): Change type to size_t.
5329         (rtx_alloc_sizes): Likewise.
5330         (rtx_count_cmp): New.
5331         (dump_rtx_statistics): Sort first based on counts.
5332         * tree.c (tree_nodes_cmp): New.
5333         (tree_codes_cmp): New.
5334         (dump_tree_statistics): Sort first based on counts.
5335         * system.h (ONE_K): New.
5336         (ONE_M): Likewise.
5337         (SIZE_SCALE): Likewise.
5338         (SIZE_LABEL): Likewise.
5339         (SIZE_AMOUNT): Likewise.
5340         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
5341         * tree-dfa.c (dump_dfa_stats): Likewise.
5342         * tree-phinodes.c (phinodes_print_statistics): Likewise.
5343         * tree-ssanames.c (ssanames_print_statistics): Likewise.
5344         * tree.c (dump_tree_statistics): Likewise.
5345         * vec.c (struct vec_usage): Likewise.
5346         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
5347         trigger a -Werror=format-overflow with
5348         --enable-gather-detailed-stats.
5350 2018-11-05  Martin Liska  <mliska@suse.cz>
5352         * mem-stats.h (mem_alloc_description::release_instance_overhead):
5353         Return T *.
5354         * vec.c (struct vec_usage): Register m_element_size.
5355         (vec_prefix::register_overhead): New arguments: elements and
5356         element_size.
5357         (vec_prefix::release_overhead): Subtract elements.
5358         * vec.h (struct vec_prefix): Change signature.
5359         (va_heap::reserve): Pass proper arguments.
5360         (va_heap::release): Likewise.
5362 2018-11-05  Martin Liska  <mliska@suse.cz>
5364         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
5365         style.
5366         * vec.c: Likewise.
5368 2018-11-05  Richard Biener  <rguenther@suse.de>
5370         * tree-scalar-evolution.h (final_value_replacement_loop): Update
5371         prototype.
5372         * tree-scalar-evolution.c (final_value_replacement_loop): Return
5373         whether anything was done.
5374         (scev_const_prop): Remove constant propagation part, fold
5375         remains into ...
5376         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
5377         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
5378         conditionally.
5380 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
5382         PR tree-optimization/87859
5383         * gimple-ssa-store-merging.c (struct merged_store_group): Add
5384         only_constants and first_nonmergeable_order members.
5385         (merged_store_group::merged_store_group): Initialize them.
5386         (merged_store_group::do_merge): Clear only_constants member if
5387         adding something other than INTEGER_CST store.
5388         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
5389         stores with order >= first_nonmergeable_order.  Use
5390         merged_store->only_constants instead of always recomputing it.
5391         Set merged_store->first_nonmergeable_order if we've skipped any
5392         stores.  Attempt to merge overlapping INTEGER_CST stores that
5393         we would otherwise skip.
5395         PR sanitizer/87837
5396         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
5398 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
5400         PR target/87853
5401         * config/i386/emmintrin.h (__v16qs): New to cope with option
5402         -funsigned-char.
5403         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
5404         (_mm_cmplt_epi8): Likewise.
5405         (_mm_cmpgt_epi8): Likewise.
5407 2018-11-05  Richard Biener  <rguenther@suse.de>
5409         PR rtl-optimization/87852
5410         * fwprop.c (use_killed_between): Only consider single-defs of the
5411         use whose definition statement dominates the use.
5413 2018-11-05  Martin Liska  <mliska@suse.cz>
5415         PR web/87829
5416         * doc/invoke.texi: Remove options that are
5417         not disabled with -Os.
5419 2018-11-05  Martin Liska  <mliska@suse.cz>
5421         PR c/87811
5422         * doc/extend.texi: Update constrain about the last argument
5423         of __builtin_expect_with_probability.
5425 2018-11-05  Martin Liska  <mliska@suse.cz>
5427         PR c/87811
5428         * predict.c (expr_expected_value_1): Verify
5429         that last argument is a real constants and emit
5430         error.
5432 2018-11-05  Martin Liska  <mliska@suse.cz>
5434         PR gcov-profile/77698
5435         * ipa-profile.c (ipa_profile): Adjust hotness threshold
5436         only in LTO mode.
5438 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5440         PR tree-optimization/86572
5441         * builtins.c (c_strlen): Handle negative offsets in a safe way.
5443 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5445         PR tree-optimization/87672
5446         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
5447         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
5449 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
5451         PR middle-end/58372
5452         * cfgexpand.c (pass_expand::execute): Move the call to
5453         finish_eh_generation in front of the call to expand_stack_alignment.
5455 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
5457         * common/config/i386/i386-common.c (processor_alias_table): Add
5458         znver2 entry.
5459         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
5460         (case ${target}): Add znver2.
5461         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
5462         -march=native recognize znver2 processors.
5463         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
5464         * config/i386/i386.c (m_znver2): New definition.
5465         (m_ZNVER): New definition.
5466         (m_AMD_MULTIPLE): Includes m_znver2.
5467         (processor_cost_table): Add znver2 entry.
5468         (processor_target_table): Add znver2 entry.
5469         (get_builtin_code_for_version): Set priority for
5470         PROCESSOR_ZNVER2.
5471         (processor_model): Add M_AMDFAM17H_ZNVER2.
5472         (arch_names_table): Ditto.
5473         (ix86_reassociation_width): Include znver2.
5474         * config/i386/i386.h (TARGET_znver2): New definition.
5475         (struct ix86_size_cost): Add TARGET_ZNVER2.
5476         (enum processor_type): Add PROCESSOR_ZNVER2.
5477         * config/i386/i386.md (define_attr "cpu"): Add znver2.
5478         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
5479         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
5480         (ix86_adjust_cost): Add znver2.
5481         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
5482         * gcc/doc/extend.texi: Add details about znver2.
5483         * gcc/doc/invoke.texi: Add details about znver2.
5485 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
5487         PR target/87079
5489         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
5490         pattern.
5492 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5494         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
5495         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
5497 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
5499         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
5500         shrink-wrapping when -mtrack-speculation.
5502 2018-11-02  Richard Biener  <rguenther@suse.de>
5504         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
5505         (pop_cost_one_pair): Do not free pair.
5506         (pop_best_coalesce): Likewise.
5507         (create_coalesce_list): Initialize obstack.
5508         (delete_coalesce_list): Free obstack.
5509         (find_coalesce_pair): Obstack-allocate coalesce pairs.
5510         (add_cost_one_coalesce): Likewise.
5511         (struct live_track): Remove bitmap pointer indirections.
5512         (new_live_track): Adjust.
5513         (delete_live_track): Likewise.
5514         (live_track_remove_partition): Likewise.
5515         (live_track_add_partition): Likewise.
5516         (live_track_live_p): Likewise.
5517         (live_track_process_def): Likewise.
5518         (live_track_clear_base_vars): Likewise.
5520 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5522         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
5523         * configure: Regenerate.
5524         * config.in: Regenerate.
5525         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
5526         HAVE_AS_REGISTER_PSEUDO_OP guard.
5527         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
5528         !HAVE_AS_REGISTER_PSEUDO_OP support.
5529         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
5530         guard.
5532 2018-11-02  Richard Biener  <rguenther@suse.de>
5534         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
5535         with is_type_die.
5537 2018-11-02  Richard Biener  <rguenther@suse.de>
5539         PR tree-optimization/87776
5540         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
5541         executable when iterating but running into rpo-vn-max-loop-depth
5542         and not eliding the iteration.
5544 2018-11-30  Jan Hubicka  <jh@suse.cz>
5546         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
5547         TYPE_DECL.
5549 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
5551         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
5552         prototype.
5553         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
5554         Combine with rs6000_address_for_fpconvert.
5555         (rs6000_address_for_fpconvert): Combine with
5556         rs6000_force_indexed_or_indirect_mem.
5557         (rs6000_expand_vector_init): Change function call from
5558         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
5559         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
5560         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
5561         (floatsi<mode>2_lfiwax_mem): Ditto.
5562         (floatunssi<mode>2_lfiwzx): Ditto.
5563         (floatunssi<mode>2_lfiwzx_mem): Ditto.
5564         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
5565         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
5566         (fix_trunc<mode>si2_stfiwx): Ditto.
5567         (fixuns_trunc<mode>si2_stfiwx): Ditto.
5568         (float_<mode>si2_hw): Ditto.
5569         (floatuns_<mode>si2_hw): Ditto.
5570         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
5571         (vsx_splat_<mode>): Ditto.
5573 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
5575         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
5576         at end of assembler input text.
5577         * configure: Regenerate.
5579 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
5581         PR tree-optimization/87826
5582         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
5583         negative or larger or equal to type's precision.
5585 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
5587         * opts.c (default_options_table): Do not enable
5588         OPT_fdelayed_branch at -Og.
5589         * doc/invoke.texi (-fdelayed-branch): Document it.
5591 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
5593         * optabs-libfuncs.c (build_libfunc_function_visibility):
5594         New, split out from...
5595         (build_libfunc_function): ... here.
5596         (init_one_libfunc_visibility): New, split out from ...
5597         (init_one_libfunc): ... here.
5599         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
5600         scratch register need not be early-clobber.  Document the reason
5601         why we cannot use ST<OP>.
5603 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
5605         PR bootstrap/82856
5606         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
5607         line for second argument of AC_DEFINE_UNQUOTED.
5608         * doc/install.texi (Tools/packages necessary for modifying GCC):
5609         Update to autoconf 2.69 and automake 1.15.1.
5610         * aclocal.m4, config.in, configure: Regenerate.
5612 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
5614         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
5615         initialization.
5617 2018-10-31  Martin Liska  <mliska@suse.cz>
5619         PR driver/83193
5620         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
5621         Add new argument invalid_extension.
5622         (aarch64_get_all_extension_candidates): New function.
5623         (aarch64_rewrite_selected_cpu): Add NULL to function call.
5624         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
5625         new argument.
5626         (aarch64_get_all_extension_candidates): New function.
5627         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
5628         argument invalid_extension.
5629         (aarch64_parse_cpu): Likewise.
5630         (aarch64_print_hint_for_extensions): New function.
5631         (aarch64_validate_mcpu): Provide hint about invalid extension.
5632         (aarch64_validate_march): Likewise.
5633         (aarch64_handle_attr_arch): Pass new argument.
5634         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
5635         (aarch64_handle_attr_isa_flags): Likewise.
5637 2018-10-31  Richard Biener  <rguenther@suse.de>
5639         PR middle-end/70359
5640         PR middle-end/86270
5641         * tree-outof-ssa.c (insert_backedge_copies): Restrict
5642         copy generation to useful cases.  Place the copy before
5643         the definition of the backedge value when possible.
5645 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5647         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
5648         * config/arc/arc.c (arc_active_insn): New function.
5649         (check_store_cacheline_hazard): Likewise.
5650         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
5651         (arc_override_options): Disable delay slot scheduler for older
5652         A7.
5653         (arc_store_addr_hazard_p): New implementation, old one renamed to
5654         ...
5655         (arc_store_addr_hazard_internal_p): Renamed.
5656         (arc_reorg): Don't combine into brcc instructions which are part
5657         of hardware hazard solution.
5658         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
5659         (tune_arc700): Likewise.
5660         * config/arc/arc.opt (arc7xx): New tune value.
5661         * config/arc/arc700.md: Improve A7 scheduler.
5663 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5665         * config/arc/arc.c (arc_override_options): Remove
5666         TARGET_COMPACT_CASESI.
5667         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
5668         (CASE_VECTOR_MODE): Likewise.
5669         (CASE_VECTOR_PC_RELATIVE): Likewise.
5670         (CASE_VECTOR_SHORTEN_MODE): Likewise.
5671         (CASE_VECTOR_SHORTEN_MODE1): Delete.
5672         (ADDR_VEC_ALIGN): Update.
5673         (ASM_OUTPUT_CASE_LABEL): Undefine.
5674         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
5675         (TARGET_BI_BIH): Define.
5676         (DEFAULT_BRANCH_INDEX): Likewise.
5677         * config/arc/arc.md (casesi): Rework to accept BI/BIH
5678         instructions, remove compact_casesi use case.
5679         (casesi_compact_jump): Remove.
5680         (casesi_dispatch): New pattern.
5681         * config/arc/arc.opt: Add mbranch-index option. Deprecate
5682         compact_casesi option.
5683         * doc/invoke.texi: Document mbranch-index option.
5685 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5687         * config/arc/arc.c (arc_get_tp): Remove function.
5688         (arc_emit_call_tls_get_addr): Likewise.
5689         (arc_call_tls_get_addr): New function.
5690         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
5691         * config/arc/arc.md (tls_load_tp_soft): Remove.
5692         (tls_gd_get_addr): Likewise.
5694 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5696         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
5697         (mulsi3_600_lib): Remove pattern.
5698         (umulsi3_highpart_600_lib_le): Likewise.
5699         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
5700         (umulsidi3): Remove call to umulsidi3_600_lib.
5701         (umulsidi3_600_lib): Remove pattern.
5702         (peephole2): Remove peephole using the above deprecated patterns.
5704 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5706         PR target/87374
5707         * config/arm/arm.c (arm_option_check_internal): Disable the combined
5708         use of -mslow-flash-data and -mword-relocations.
5709         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
5710         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
5711         flag_pic.
5712         * doc/invoke.texi (-mword-relocations): Mention conflict with
5713         -mslow-flash-data.
5714         (-mslow-flash-data): Reciprocally.
5716 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
5718         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
5719         16-byte modes held in GP registers to use an even regno.
5721         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
5722         (aarch64_atomic_ldop_supported_p): Remove.
5723         (aarch64_gen_atomic_ldop): Remove.
5724         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
5725         Fully expand LSE operations here.
5726         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
5727         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
5728         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
5729         and use ATOMIC_LDOP instead; use register_operand for the input;
5730         drop the split and emit insns directly.
5731         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
5732         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
5733         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
5735         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
5736         (aarch64_gen_atomic_ldop): Don't call it.
5737         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
5738         Use aarch64_reg_or_zero.
5739         (aarch64_atomic_exchange<ALLI>): Likewise.
5740         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
5741         operand 0; use aarch64_reg_or_zero for input; merge ...
5742         (@aarch64_atomic_swp<ALLI>): ... this and remove.
5744         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
5745         (aarch64_split_compare_and_swap): Use it.
5746         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
5747         test oldval against the proper predicate.
5748         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
5749         Use nonmemory_operand for expected.
5750         (cas_short_expected_pred): New.
5751         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
5752         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
5753         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
5754         (aarch64_plushi_operand): New.
5756         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
5757         Force oldval into the rval register for TARGET_LSE; emit the compare
5758         during initial expansion so that it may be deleted if unused.
5759         (aarch64_gen_atomic_cas): Remove.
5760         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
5761         Change =&r to +r for operand 0; use match_dup for operand 2;
5762         remove is_weak and mod_f operands as unused.  Drop the split
5763         and merge with...
5764         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
5765         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
5766         (@aarch64_atomic_cas<GPI>): Similarly.
5768 2018-10-31  Richard Biener  <rguenther@suse.de>
5770         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
5771         using ABSU_EXPR.
5773 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5775         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
5776         * config/aarch64/aarch64.md: Include saphira.md
5777         * config/aarch64/saphira.md: New file for pipeline description.
5779 2018-10-30  Martin Sebor  <msebor@redhat.com>
5781         PR middle-end/87041
5782         * gimple-ssa-sprintf.c (format_directive): Use %G to include
5783         inlining context.
5784         (sprintf_dom_walker::compute_format_length):
5785         Avoid setting POSUNDER4K here.
5786         (get_destination_size): Handle null argument values.
5787         (get_user_idx_format): New function.
5788         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
5789         functions, including user-defined with attribute format printf.
5790         Use %G to include inlining context.
5791         Set POSUNDER4K here.
5793 2018-10-30  Jan Hubicka  <jh@suse.cz>
5795         * params.def (lto-partitions): Bump from 32 to 128.
5797 2018-10-30  Jan Hubicka  <jh@suse.cz>
5799         * tree.c
5800         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
5801         head in file.
5802         (free_lang_data_in_type): Forward declare.
5803         (fld_type_variant_equal_p): New function.
5804         (fld_type_variant): New function
5805         (fld_incomplete_types): New hash.
5806         (fld_incomplete_type_of): New function
5807         (fld_simplfied-type): New function.
5808         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
5809         (free_lang_data): Allocate and free fld_incomplete_type; update call
5810         of free_lang_data_in_decl.
5812 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
5814         * gcov.c (output_lines): Remove duplicate line.
5816 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5818         * config/rs6000/rs6000.md (bswapdi2): Force address into register
5819         if not in indexed or indirect form.
5820         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
5821         (bswapdi2_store): Ditto.
5822         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
5823         helper function.
5824         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
5825         Prototype for helper function.
5827 2018-10-30  Martin Sebor  <msebor@redhat.com>
5829         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
5830         (target, pragma GCC optimize, pragma GCC target): Ditto.
5832 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
5834         * doc/extend.texi: Fix prototype and description of
5835         __builtin_expect_with_probability.
5837 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
5839         * cgraph.h (clone_function_name_1): Replaced by new
5840           clone_function_name_numbered that takes name as string; for
5841           privatize_symbol_name_1 use only.
5842           (clone_function_name): Renamed to
5843           clone_function_name_numbered to be explicit about numbering.
5844           (clone_function_name): New two-argument function that does
5845           not number its output.
5846           (clone_function_name): New three-argument function that
5847           takes a number to append to its output.
5848         * cgraphclones.c (duplicate_thunk_for_node):
5849           (clone_function_name_1): Renamed.
5850           (clone_function_name_numbered): Two new functions.
5851           (clone_function_name): Improved documentation.
5852           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
5853         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
5854         * final.c (final_scan_insn_1): Use the new clone_function_name
5855           without numbering.
5856         * multiple_target.c (create_dispatcher_calls): Ditto.
5857           (create_target_clone): Ditto.
5858         * omp-expand.c (grid_expand_target_grid_body): Ditto.
5859         * omp-low.c (create_omp_child_function_name): Ditto.
5860         * omp-simd-clone.c (simd_clone_create): Ditto.
5861         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
5862           new clone_function_name without numbering.
5864 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
5866         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
5867         Assert that the allocation size is not zero.
5869 2018-10-30  Richard Biener  <rguenther@suse.de>
5871         PR tree-optimization/87800
5872         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
5873         non-induction or reduction PHIs.
5875 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
5877         * config/aarch64/falkor-tag-collision-avoidance.c
5878         (execute_tag_collision_avoidance): Call df_note_add_problem.
5880 2018-10-30  Martin Liska  <mliska@suse.cz>
5882         * doc/extend.texi: Fix typo in documentation
5883         of __builtin_expect_with_probability.
5885 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5887         PR c++/87721
5888         * input.c (get_substring_ranges_for_loc): Detect if
5889         linemap_resolve_location gives us a NULL map, and reject
5890         this case.
5892 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
5894         * config.gcc (xstormy16-*-elf): Set tm_d_file.
5896 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5897             Martin Sebor  <msebor@redhat.com>
5898             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
5900         * doc/ux.texi (Quoting): New subsection, adapted from material at
5901         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
5902         MartinSebor and ManuelLopezIbanez.
5903         (Fix-it hints): Note that fix-it hints shouldn't be marked for
5904         translation.
5906 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
5908         PR middle-end/87469
5909         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
5910         max value.
5912 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5914         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
5916 2018-10-29  Paul Koning  <ni1d@arrl.net>
5918         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
5920 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5922         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
5923         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
5924         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
5925         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
5926         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
5927         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
5928         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
5929         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
5930         Change 'vector' to '__vector'.
5931         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
5932         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
5933         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
5934         _mm_avg_pu16): Likewise.  And, whitespace corrections.
5936 2018-10-29  Richard Biener  <rguenther@suse.de>
5938         PR tree-optimization/87785
5939         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
5940         internal defs.
5942 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5944         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
5946 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5948         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
5949         ports configuration macro, defaults to "gnu".
5950         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
5951         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
5953 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5955         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
5957 2018-10-29  Richard Biener  <rguenther@suse.de>
5959         PR tree-optimization/87790
5960         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
5961         (vect_make_slp_decision): Adjust.
5962         (vect_slp_analyze_bb_1): Likewise.
5963         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
5964         edges.
5966 2018-10-29  Richard Biener  <rguenther@suse.de>
5968         PR tree-optimization/87785
5969         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
5970         and processing.
5971         (vect_build_slp_tree): Likewise.
5972         (vect_gather_slp_loads): New function.
5973         (vect_analyze_slp_instance): Gather loads separately from the
5974         SLP tree build.
5976 2018-10-29  Martin Liska  <mliska@suse.cz>
5978         * Makefile.in: Make dependency to json.o.
5979         * doc/gcov.texi: Document new JSON format, remove
5980         old intermediate format documentation.
5981         * gcov.c (struct function_info): Come up with m_name and
5982         m_demangled_name.
5983         (function_info::function_info): Initialize it.
5984         (function_info::~function_info): Release it.
5985         (main): Rename flag_intermediate_format to flag_json_format.
5986         (print_usage): Describe --json-format.
5987         (process_args): Set flag_json_format.
5988         (output_intermediate_line): Remove.
5989         (output_intermediate_json_line): Likewise.
5990         (get_gcov_intermediate_filename): Return new extension
5991         ".gcov.json.gz".
5992         (output_intermediate_file): Implement JSON emission.
5993         (output_json_intermediate_file): Implement JSON emission.
5994         (generate_results): Use ::get_name for function name.
5995         Handle JSON output file.
5996         (read_graph_file): Use ::get_name instead of cplus_demangle.
5997         (read_count_file): Likewise.
5998         (solve_flow_graph): Likewise.
5999         (add_line_counts): Likewise.
6000         (accumulate_line_counts): Use new flag_json_format.
6001         (output_function_details): Use ::get_name instead of cplus_demangle.
6002         (output_lines): Likewise.
6003         * json.cc (test_writing_literals): Add new tests.
6004         * json.h (class literal): Add new boolean constructor.
6006 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
6008         PR rtl-optimization/87701
6009         PR rtl-optimization/87780
6010         * combine.c (make_more_copies): Rewrite.
6012 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
6014         * doc/generic.texi (ABSU_EXPR): Document.
6015         * match.pd (absu(x)*absu(x) -> x*x): Handle.
6016         (absu(absu(X)) -> absu(X)): Likewise.
6017         (absu(-X) -> absu(X)): Likewise.
6018         (absu(X)  where X is nonnegative -> X): Likewise.
6020 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
6022         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
6023         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
6024         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
6025         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
6026         (s-tm-texi): Also check timestamp on d-target.def.
6027         (generated_files): Add TM_D_H and d-target-hooks-def.h.
6028         (build/genhooks.o): Also depend on D_TARGET_DEF.
6029         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
6030         variables.
6031         * config/aarch64/aarch64-d.c: New file.
6032         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
6033         Define.
6034         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
6035         prototype.
6036         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
6037         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
6038         * config/arm/arm-d.c: New file.
6039         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
6040         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
6041         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6042         * config/arm/t-arm (arm-d.o): New rule.
6043         * config/default-d.c: New file.
6044         * config/glibc-d.c: New file.
6045         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6046         * config/i386/i386-d.c: New file.
6047         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
6048         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
6049         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6050         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
6051         * config/i386/t-i386 (i386-d.o): New rule.
6052         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6053         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6054         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
6055         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6056         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6057         * config/mips/mips-d.c: New file.
6058         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
6059         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
6060         * config/mips/t-mips (mips-d.o): New rule.
6061         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6062         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6063         * config/powerpcspe/powerpcspe-d.c: New file.
6064         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
6065         New prototype.
6066         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6067         Support GNU D by using 0 as the language type.
6068         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
6069         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
6070         * config/riscv/riscv-d.c: New file.
6071         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
6072         prototype.
6073         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
6074         * config/riscv/t-riscv (riscv-d.o): New rule.
6075         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6076         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6077         * config/rs6000/rs6000-d.c: New file.
6078         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
6079         prototype.
6080         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6081         Support GNU D by using 0 as the language type.
6082         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
6083         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
6084         * config/s390/s390-d.c: New file.
6085         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
6086         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
6087         * config/s390/t-s390 (s390-d.o): New rule.
6088         * config/sparc/sparc-d.c: New file.
6089         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
6090         prototype.
6091         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
6092         * config/sparc/t-sparc (sparc-d.o): New rule.
6093         * config/t-glibc (glibc-d.o): New rule.
6094         * configure: Regenerated.
6095         * configure.ac (tm_d_file): New variable.
6096         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
6097         * doc/contrib.texi (Contributors): Add self for the D frontend.
6098         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
6099         * doc/install.texi (Configuration): Mention libphobos as an option for
6100         --enable-shared.  Mention d as an option for --enable-languages.
6101         (Testing): Mention check-d as a target.
6102         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
6103         name suffixes.  Mention d as a -x option.
6104         * doc/sourcebuild.texi (Top Level): Mention libphobos.
6105         * doc/standards.texi (Standards): Add section on D language.
6106         * doc/tm.texi: Regenerated.
6107         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
6108         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
6109         * dwarf2out.c (is_dlang): New function.
6110         (gen_compile_unit_die): Use DW_LANG_D for D.
6111         (declare_in_namespace): Return module die for D, instead of adding
6112         extra declarations into the namespace.
6113         (gen_namespace_die): Generate DW_TAG_module for D.
6114         (gen_decl_die): Handle CONST_DECLSs for D.
6115         (dwarf2out_decl): Likewise.
6116         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
6117         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
6118         kinds of aggregates.
6119         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
6120         * genhooks.c: Include d/d-target.def.
6122 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
6124         PR target/85669
6125         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
6126         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
6127         (STACK_DYNAMIC_OFFSET): Likewise.
6129 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
6131         PR target/80024
6132         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
6133         error message.
6135 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
6137         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
6138         return type and other typos.
6140 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
6142         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
6143         a shorter sequence with fewer branches.
6144         (emit_final_str_compare_gpr): Ditto.
6146 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6148         * config/rs6000/tmmintrin.h: New file.
6149         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
6151 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6153         * config/rs6000/mmintrin.h: Enable 32bit compilation.
6154         * config/rs6000/xmmintrin.h: Likewise.
6156 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
6158         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
6160 2018-10-26  Richard Biener  <rguenther@suse.de>
6162         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
6163         and wrapper.
6164         (vect_mark_slp_stmts_relevant): Likewise.
6165         (vect_detect_hybrid_slp_stmts): Likewise.
6166         (vect_bb_slp_scalar_cost): Likewise.
6167         (vect_remove_slp_scalar_calls): Likewise.
6169 2018-10-26  Jan Hubicka  <jh@suse.cz>
6171         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
6172         (warn_types_mismatch): Fix walk of DECL_NAME.
6173         (odr_types_equivalent_p): Fix overactive assert.
6175 2018-10-26  Richard Biener  <rguenther@suse.de>
6177         PR tree-optimization/87105
6178         * tree-vectorizer.h (_slp_tree::refcnt): New member.
6179         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
6180         refcnt.
6181         (vect_create_new_slp_node): Initialize refcnt to one.
6182         (bst_traits): Move.
6183         (scalar_stmts_set_t, bst_fail): Remove.
6184         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
6185         (vect_build_slp_tree): Add bst_map argument and lookup
6186         already created SLP nodes.
6187         (vect_print_slp_tree): Handle a SLP graph, print SLP node
6188         addresses.
6189         (vect_slp_rearrange_stmts): Handle a SLP graph.
6190         (vect_analyze_slp_instance): Adjust and free SLP nodes from
6191         the CSE map.  Fix indenting.
6192         (vect_schedule_slp_instance): Add short-cut.
6194 2018-10-26  Martin Liska  <mliska@suse.cz>
6196         PR testsuite/86158
6197         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
6198         addr_expr and not with pointers.
6200 2018-10-26  Jan Hubicka  <jh@suse.cz>
6202         * tree.c (free_lang_data_in_type): Only check main variants.
6203         * ipa-devirt.c (warn_odr): Make static.
6204         (types_same_for_odr): Drop strict variant.
6205         (types_odr_comparable): Likewise.
6206         (odr_or_derived_type_p): Look for main variants.
6207         (odr_name_hasher::equal): Cleanup comment.
6208         (odr_subtypes_equivalent): Add warn and warned arguments; check main
6209         variants.
6210         (type_variants_equivalent_p): break out from ...
6211         (odr_types_equivalent): ... here; go for main variants where needed.
6212         (warn_odr): ... here; turn static.
6213         (warn_types_mismatch): Compare mangled names of main variants.
6214         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
6215         (type_with_linkage_p): Sanity check that we look at main variant.
6216         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
6217         * tree.h (types_same_for_odr): Drop strict argument.
6219 2018-10-26  Richard Biener  <rguenther@suse.de>
6221         PR tree-optimization/87746
6222         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
6223         Simplify and fix WRT strided store groups with size not
6224         equal to step in element count.
6225         (vect_analyze_group_access_1): Dump the whole group.
6227 2018-10-25  Carl Love  <cel@us.ibm.com>
6229         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
6230         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
6231         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
6232         precicion entry for each overloaded builtin.
6233         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
6234         VSCEDPUO): Rename overloaded name.
6235         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
6236         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
6237         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
6238         define_expand for xscmpexqp instruction.
6239         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
6241 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
6242             Jinsong Ji <jji@us.ibm.com>
6244         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
6245         function with vec_sl.
6246         (_mm_slli_epi32): Likewise.
6247         (_mm_slli_epi64): Likewise.
6248         (_mm_srai_epi16): Replace deprecated function with vec_sra.
6249         (_mm_srai_epi32): Likewise.
6250         (_mm_srli_epi16): Replace deprecated function with vec_sr.
6251         (_mm_srli_epi32): Likewise.
6252         (_mm_srli_epi64): Likewise.
6253         (_mm_sll_epi16): Replace deprecated function with vec_sl.
6254         (_mm_sll_epi32): Likewise.
6255         (_mm_sll_epi64): Likewise.
6256         (_mm_sra_epi16): Replace deprecated function with vec_sra.
6257         (_mm_sra_epi32): Likewise.
6258         (_mm_srl_epi16): Replace deprecated function with vec_sr.
6259         (_mm_srl_epi32): Likewise.
6260         (_mm_srl_epi64): Likewise.
6262 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
6263             Jinsong Ji <jji@us.ibm.com>
6265         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
6266         comparison operators with vec_cmp* for compatibility due to
6267         unfortunate history; clean up formatting and use types more
6268         appropriately.
6269         (_mm_sll_epi32): Likewise.
6270         (_mm_sll_epi64): Likewise.
6271         (_mm_srl_epi16): Likewise.
6272         (_mm_srl_epi32): Likewise.
6273         (_mm_srl_epi64): Likewise.
6275 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
6276             Jinsong Ji <jji@us.ibm.com>
6278         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
6279         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
6280         __vector __bool int.  Use vec_cmpgt in preference to deprecated
6281         function vec_vcmpgtfp.
6282         (_mm_max_ps): Likewise.
6284 2018-10-25  Jeff Law  <law@redhat.com>
6286         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
6287         if returning false.
6289 2018-10-25  Martin Sebor  <msebor@redhat.com>
6291         * doc/extend.texi (aligned): Expand attribute description.
6292         (Alignment): Rename section.  Discuss function arguments.
6294 2018-10-25  Jan Hubicka  <jh@suse.cz>
6296         * ipa-devirt.c (main_odr_variant): Remove.
6297         (hash_odr_name, types_same_for_odr, types_odr_comparable,
6298         odr_name_hasher::equal, odr_subtypes_equivalent_p):
6299         Drop use of main_odr_variant.
6300         (add_type_duplicate): Silence confused warnings on integer types.
6301         (get_odr_type): Always look for main variant.
6302         (register_odr_type): Simplify.
6304 2018-10-25  Richard Biener  <rguenther@suse.de>
6306         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
6307         Initialize ng to silence error with release checking bootstrap.
6309 2018-10-25  Richard Biener  <rguenther@suse.de>
6311         * tree-if-conv.c: Include tree-ssa-sccvn.h.
6312         (tree_if_conversion): Run CSE on the if-converted loop body.
6314 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6316         * config/s390/constraints.md (ZL): New constraint.
6317         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
6318         operands.
6319         * config/s390/s390.md (movdi_larl): Remove.
6320         (movdi_64): Add the LARL alternative.
6322 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6324         PR bootstrap/87747
6325         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
6326         (RTX_CODE_HWINT_P): New macro.
6327         (rtx_code_size): Use RTX_CODE_HWINT_P ().
6329 2018-10-25  Jan Hubicka  <jh@suse.cz>
6331         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
6332         is anonymous.
6334 2018-10-25  Richard Biener  <rguenther@suse.de>
6336         PR tree-optimization/87665
6337         PR tree-optimization/87745
6338         * tree-vectorizer.h (get_earlier_stmt): Remove.
6339         (get_later_stmt): Pick up UID from the original non-pattern stmt.
6341 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
6343         * options.texi (Deprecated): Move list to Var section.
6345 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
6346             Jinsong Ji <jji@us.ibm.com>
6348         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
6349         __vector long to __vector long long.
6350         (_mm_cvtpd_ps): Likewise.
6351         (_mm_cvttpd_epi32): Likewise.
6352         (_mm_cvtpi32_pd): Likewise.
6353         (_mm_unpackhi_epi64): Likewise.
6354         (_mm_unpacklo_epi64): Likewise.
6356 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
6358         PR rtl-optimization/87720
6359         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
6361 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
6363         * gimple-ssa-isolate-paths.c
6364         (find_implicit_erroneous_behavior): Do not change code if the
6365         pass is running for warnings only.
6366         (find_explicit_erroneous_behavior): Likewise.
6368 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
6370         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
6371         Define as rs6000_mangle_decl_assembler_name.
6372         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
6373         long double to IEEE long double, switch the names of the long
6374         double built-in functions to be <func>f128 instead of <func>l.
6376 2018-10-24  Martin Sebor  <msebor@redhat.com>
6378         * doc/extend.texi (nonnull): List no-argument form.  Reference
6379         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
6381 2018-10-24  Richard Biener  <rguenther@suse.de>
6383         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
6385 2018-10-24  Martin Liska  <mliska@suse.cz>
6387         PR tree-optimization/84436
6388         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
6389         Remove.
6390         (switch_conversion::contains_linear_function_p): New.
6391         (switch_conversion::build_one_array): Support linear
6392         transformation on input.
6393         * tree-switch-conversion.h (struct switch_conversion): Add
6394         contains_linear_function_p declaration.
6396 2018-10-24  Richard Biener  <rguenther@suse.de>
6398         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
6399         if its argument is CONSTANT_CLASS_P.
6401 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
6403         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
6404         it is wrong for forward declarations.
6406 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6408         * config/s390/s390.c (s390_check_qrst_address): Add the missing
6409         SYMBOL_REF_P () check.
6411 2018-10-24  Richard Biener  <rguenther@suse.de>
6413         PR tree-optimization/87105
6414         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
6415         dump classification.
6416         (vect_analyze_data_ref_accesses): Handle duplicate loads and
6417         stores by splitting the affected group after the fact.
6418         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
6419         fail the SLP build because of size constraints.
6421 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6423         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
6424         * configure: Regenerate.
6425         * config.in: Regenerate.
6426         * varasm.c (mergeable_string_section): Use readonly_data_section
6427         if linker doesn't support SHF_MERGE with alignment > 8.
6428         (mergeable_constant_section): Likewise.
6430 2018-10-24  Richard Biener  <rguenther@suse.de>
6432         PR tree-optimization/84013
6433         * tree-ssa-structalias.c (struct msdi_data): New struct for
6434         marshalling data to walk_stmt_load_store_ops.
6435         (maybe_set_dependence_info): Refactor as callback for
6436         walk_stmt_load_store_ops.
6437         (compute_dependence_clique): Set restrict info on all stmt kinds.
6439 2018-10-24  Martin Liska  <mliska@suse.cz>
6441         * cgraph.c (cgraph_node::dump):
6442         Remove reduntant dumps and make tp_first_run dump more compact.
6444 2018-10-24  Richard Biener  <rguenther@suse.de>
6446         PR tree-optimization/87665
6447         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
6448         to reflect reality.
6450 2018-10-12  Jeff Law  <law@redhat.com>
6452         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
6453         for H8/S.
6455 2018-10-23  Richard Biener  <rguenther@suse.de>
6457         * tree-vrp.c (add_assert_info): Guard dump_printf with
6458         dump_enabled_p.
6459         * gimple-ssa-evrp-analyze.c
6460         (evrp_range_analyzer::record_ranges_from_incoming_edge):
6461         Use value_range::ignore_equivs_equal_p.
6463 2018-10-23  Richard Biener  <rguenther@suse.de>
6465         PR tree-optimization/87105
6466         PR tree-optimization/87608
6467         * passes.def (pass_all_early_optimizations): Add early phi-opt
6468         after dce.
6469         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
6470         addition to debug stmts.
6471         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
6472         and abs replacement early.
6473         * tree-cfg.c (gimple_empty_block_p): Likewise.
6475 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
6477         PR target/86383
6478         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
6479         specified to configure.
6480         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
6482 2018-10-23  Richard Biener  <rguenther@suse.de>
6484         PR tree-optimization/87700
6485         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
6487 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
6489         PR target/87674
6490         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
6491         second argument from __mmask16 to __mmask8.
6492         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
6493         _mm_mask_packs_epi32): Likewise.
6494         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
6495         Likewise.
6496         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
6498 2018-10-23  Richard Biener  <rguenther@suse.de>
6500         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
6502 2018-10-23  Richard Biener  <rguenther@suse.de>
6504         PR tree-optimization/86144
6505         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
6506         over simd attribute.
6508 2018-10-23  Richard Biener  <rguenther@suse.de>
6510         PR tree-optimization/87693
6511         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
6512         the case we do not find the taken edge.
6514 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
6515             Jinsong Ji  <jji@us.ibm.com>
6517         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
6518         (_mm_store_pd): Use unaligned vector type for pointer cast.
6519         (_mm_maskmoveu_si128): Likewise.
6520         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
6521         (_mm_store_ps): Use unaligned vector type for pointer cast.
6523 2018-10-22  Paul Koning  <ni1d@arrl.net>
6525         * symtab.c (symtab_node::increase_alignment): Correct max
6526         alignment check.
6528 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
6530         PR tree-optimization/87633
6531         * match.pd: Do not generate unordered integer comparisons.
6533 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
6535         PR rtl-optimization/87600
6536         * combine.c: Add include of expr.h.
6537         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
6538         register to a pseudo.
6539         (make_more_copies): New function, add a copy to a new pseudo after
6540         the moves from hard registers into pseudos.
6541         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
6542         later.  Call make_more_copies.
6544 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
6546         * lra-constraints.c (process_alt_operands): New local array,
6547         matching_early_clobber.  Check matching_early_clobber before
6548         decrementing reject, and set matching_early_clobber after.
6550 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
6552         PR target/87598
6553         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
6554         call output_addr_const and hope for the best.
6556 2018-10-22  Richard Biener  <rguenther@suse.de>
6558         * gimple-ssa-evrp-analyze.c
6559         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
6560         smarter about what ranges to use.
6561         * tree-vrp.c (add_assert_info): Dump here.
6562         (register_edge_assert_for_2): Instead of here at multiple but
6563         not all places.
6565         * gcc.dg/tree-ssa/evrp12.c: New testcase.
6566         * gcc.dg/predict-6.c: Adjust.
6567         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
6568         * gcc.dg/tree-ssa/vrp02.c: Likewise.
6569         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
6571 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
6572         Richard Biener  <rguenther@suse.de>
6574         * bitmap.h: Update data structure documentation, including a
6575         description of bitmap views as either linked-lists or splay trees.
6576         (struct bitmap_element_def): Update comments for splay tree bitmaps.
6577         (struct bitmap_head_def): Likewise.
6578         (bitmap_list_view, bitmap_tree_view): New prototypes.
6579         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
6580         tree_form fields.
6581         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
6582         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
6583         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
6584         released bitmap element here.
6585         (bitmap_element_free): Remove.
6586         (bitmap_elt_clear_from): Work on splay tree bitmaps.
6587         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
6588         this function similar ones such that linked-list bitmap implementation
6589         functions are grouped.
6590         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
6591         and moved for grouping.
6592         (bitmap_list_insert_element_after): Renamed from
6593         bitmap_elt_insert_after, and moved for grouping.
6594         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
6595         (bitmap_tree_link_left, bitmap_tree_link_right,
6596         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
6597         bitmap_tree_link_element, bitmap_tree_unlink_element,
6598         bitmap_tree_find_element): New functions for splay-tree bitmap
6599         implementation.
6600         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
6601         Renamed and moved, see above entries.
6602         (bitmap_tree_listify_from): New function to convert part of a splay
6603         tree bitmap to a linked-list bitmap.
6604         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
6605         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
6606         (bitmap_find_bit): Remove.
6607         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
6608         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
6609         Handle splay tree bitmaps.
6610         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
6611         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
6612         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
6613         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
6614         bitmap_intersect_compl_p, bitmap_ior_and_compl,
6615         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
6616         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
6617         corresponding changes to use linked-list specific bitmap_element
6618         manipulation functions as applicable for efficiency.
6619         (bitmap_tree_to_vec): New function.
6620         (debug_bitmap_elt_file): New function split out from ...
6621         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
6622         (bitmap_print): Likewise.
6624         PR tree-optimization/63155
6625         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
6626         SSA edge worklists.
6627         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
6628         in tree-view.
6630 2018-10-22  Martin Liska  <mliska@suse.cz>
6632         PR tree-optimization/87686
6633         Revert
6634         2018-08-29  Martin Liska  <mliska@suse.cz>
6636         * tree-switch-conversion.c (switch_conversion::expand):
6637         Strenghten assumption about gswitch statements.
6639 2018-10-22  Martin Liska  <mliska@suse.cz>
6641         * ipa-icf.c (sem_item::compare_attributes): Remove.
6642         (sem_item::compare_referenced_symbol_properties): Use
6643         attribute_list_equal instead.
6644         (sem_function::equals_wpa): Likewise.
6645         * ipa-icf.h: Remove compare_attributes.
6647 2018-10-22  Richard Biener  <rguenther@suse.de>
6649         PR middle-end/87682
6650         * mem-stats.h (mem_usage::operator==): Fix pasto.
6652 2018-10-22  Richard Biener  <rguenther@suse.de>
6654         PR tree-optimization/87640
6655         * tree-vrp.c (set_value_range_with_overflow): Decompose
6656         incomplete result.
6657         (extract_range_from_binary_expr_1): Adjust.
6659 2018-10-22  Martin Jambor  <mjambor@suse.cz>
6661         * tree-eh.h (stmt_could_throw_p): Add function parameter.
6662         (stmt_can_throw_external): Likewise.
6663         (stmt_can_throw_internal): Likewise.
6664         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
6665         (lower_eh_constructs_2): Likewise.
6666         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
6667         (stmt_can_throw_external): Likewise.
6668         (stmt_can_throw_internal): Likewise.
6669         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
6670         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
6671         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
6672         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
6673         (pass_lower_eh_dispatch::execute): Pass cfun to
6674         stmt_can_throw_external.
6675         (cleanup_empty_eh): Likewise.
6676         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
6677         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
6678         stmt_can_throw_external instead of pushing it to cfun.
6679         (symbol_table::create_edge): Likewise.
6680         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
6681         stmt_can_throw_internal.
6682         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
6683         to stmt_could_throw_p.
6684         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
6685         stmt_can_throw_internal.
6686         (pass_store_merging::execute): Likewise.
6687         * gimple-ssa-strength-reduction.c
6688         (find_candidates_dom_walker::before_dom_children): Pass cfun to
6689         stmt_could_throw_p.
6690         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
6691         stmt_can_throw_internal.
6692         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
6693         to stmt_can_throw_external.
6694         (check_stmt): Pass cfun to stmt_could_throw_p.
6695         (check_stmt): Pass cfun to stmt_can_throw_external.
6696         (pass_nothrow::execute): Likewise.
6697         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
6698         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
6699         stmt_can_throw_internal.
6700         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
6701         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
6702         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
6703         * tree-complex.c (expand_complex_libcall): Pass cfun to
6704         stmt_could_throw_p and to stmt_can_throw_internal.
6705         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
6706         * tree-inline.c (copy_edges_for_bb): Likewise.
6707         (maybe_move_debug_stmts_to_successors): Likewise.
6708         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
6709         stmt_could_throw_p.
6710         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
6711         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
6712         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
6713         stmt_can_throw_internal.
6714         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
6715         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
6716         stmt_could_throw_p.
6717         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
6718         stmt_can_throw_internal.
6719         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
6720         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
6721         stmt_could_throw_p.
6722         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
6723         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
6724         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6725         (convert_mult_to_fma_1): Likewise.
6726         (convert_to_divmod): Likewise.
6727         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
6728         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
6729         * tree-ssa-propagate.c
6730         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
6731         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
6732         (maybe_optimize_range_tests): Likewise.
6733         (linearize_expr_tree): Likewise.
6734         (reassociate_bb): Likewise.
6735         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6736         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
6737         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
6738         (handle_char_store): Likewise.
6739         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
6740         stmt_can_throw_internal.
6741         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
6742         stmt_could_throw_p.
6743         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6744         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
6745         (vectorizable_simd_clone_call): Likewise.
6746         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
6747         (gimple_stringop_fixed_value): Likewise.
6749 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
6751         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
6752         literal pool references.
6753         (s390_check_qrst_address): Adapt to the new behavior of
6754         s390_loadrelative_operand_p ().
6756 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6758         PR target/72782
6759         * config/i386/sse.md (*andnot<mode>3_bcst): New.
6761 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6763         PR target/72782
6764         * config/i386/sse.md (*<code><mode>3_bcst): New.
6766 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6768         PR target/72782
6769         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
6770         V4DI, V16SI and V8DI.
6771         (*sub<mode>3<mask_name>_bcst): New.
6772         (*add<mode>3<mask_name>_bcst): Likewise.
6774 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
6775             Jinsong Ji  <jji@us.ibm.com>
6777         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
6778         __m64 with __vector unsigned long long for compatibility.
6779         (_mm_movemask_epi8): Likewise.
6780         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
6781         (_mm_cvttps_pi32): Likewise.
6782         (_mm_cvtpi32_ps): Likewise.
6783         (_mm_cvtps_pi16): Likewise.
6784         (_mm_loadh_pi): Likewise.
6785         (_mm_storeh_pi): Likewise.
6786         (_mm_movehl_ps): Likewise.
6787         (_mm_movelh_ps): Likewise.
6788         (_mm_loadl_pi): Likewise.
6789         (_mm_storel_pi): Likewise.
6790         (_mm_movemask_ps): Likewise.
6791         (_mm_shuffle_pi16): Likewise.
6793 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6795         PR target/72782
6796         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
6797         __builtin_ia32_vfnmsubpd512_mask.
6798         (_mm512_mask_fnmsub_round_pd): Likewise.
6799         (_mm512_fnmsub_pd): Likewise.
6800         (_mm512_mask_fnmsub_pd): Likewise.
6801         (_mm512_maskz_fnmsub_round_pd): Use
6802         __builtin_ia32_vfnmsubpd512_maskz.
6803         (_mm512_maskz_fnmsub_pd): Likewise.
6804         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
6805         (_mm512_mask_fnmsub_round_ps): Likewise.
6806         (_mm512_fnmsub_ps): Likewise.
6807         (_mm512_mask_fnmsub_ps): Likewise.
6808         (_mm512_maskz_fnmsub_round_ps): Use
6809         __builtin_ia32_vfnmsubps512_maskz.
6810         (_mm512_maskz_fnmsub_ps): Likewise.
6811         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
6812         __builtin_ia32_vfnmsubpd256_mask.
6813         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
6814         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6815         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
6816         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6817         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6818         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
6819         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
6820         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
6821         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
6822         __builtin_ia32_vfnmsubpd.
6823         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
6824         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
6825         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
6826         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
6827         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
6828         * config/i386/i386-builtin.def: Add
6829         __builtin_ia32_vfnmsubpd256_mask,
6830         __builtin_ia32_vfnmsubpd256_maskz,
6831         __builtin_ia32_vfnmsubpd128_mask,
6832         __builtin_ia32_vfnmsubpd128_maskz,
6833         __builtin_ia32_vfnmsubps256_mask,
6834         __builtin_ia32_vfnmsubps256_maskz,
6835         __builtin_ia32_vfnmsubps128_mask,
6836         __builtin_ia32_vfnmsubps128_maskz,
6837         __builtin_ia32_vfnmsubpd512_mask,
6838         __builtin_ia32_vfnmsubpd512_maskz,
6839         __builtin_ia32_vfnmsubps512_mask,
6840         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
6841         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
6842         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
6843         __builtin_ia32_vfnmsubpd256.
6844         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
6845         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
6846         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
6847         Likewise.
6848         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
6849         Likewise.
6850         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
6851         Likewise.
6852         (fmai_vmfnmsub_<mode><round_name>): Likewise.
6854 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6856         PR target/72782
6857         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
6858         __builtin_ia32_vfnmaddpd512_mask.
6859         (_mm512_mask_fnmadd_round_pd): Likewise.
6860         (_mm512_fnmadd_pd): Likewise.
6861         (_mm512_mask_fnmadd_pd): Likewise.
6862         (_mm512_maskz_fnmadd_round_pd): Use
6863         __builtin_ia32_vfnmaddpd512_maskz.
6864         (_mm512_maskz_fnmadd_pd): Likewise.
6865         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
6866         (_mm512_mask_fnmadd_round_ps): Likewise.
6867         (_mm512_fnmadd_ps): Likewise.
6868         (_mm512_mask_fnmadd_ps): Likewise.
6869         (_mm512_maskz_fnmadd_round_ps): Use
6870         __builtin_ia32_vfnmaddps512_maskz.
6871         (_mm512_maskz_fnmadd_ps): Likewise.
6872         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
6873         __builtin_ia32_vfnmaddpd256_mask.
6874         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
6875         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
6876         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
6877         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6878         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6879         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
6880         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
6881         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
6882         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
6883         __builtin_ia32_vfnmaddpd.
6884         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
6885         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
6886         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
6887         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
6888         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
6889         * config/i386/i386-builtin.def: Add
6890         __builtin_ia32_vfnmaddpd256_mask,
6891         __builtin_ia32_vfnmaddpd256_maskz,
6892         __builtin_ia32_vfnmaddpd128_mask,
6893         __builtin_ia32_vfnmaddpd128_maskz,
6894         __builtin_ia32_vfnmaddps256_mask,
6895         __builtin_ia32_vfnmaddps256_maskz,
6896         __builtin_ia32_vfnmaddps128_mask,
6897         __builtin_ia32_vfnmaddps128_maskz,
6898         __builtin_ia32_vfnmaddpd512_mask,
6899         __builtin_ia32_vfnmaddpd512_maskz,
6900         __builtin_ia32_vfnmaddps512_mask,
6901         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
6902         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
6903         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
6904         __builtin_ia32_vfnmaddpd256.
6905         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
6906         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
6907         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
6908         Likewise.
6909         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
6910         Likewise.
6911         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
6912         Likewise.
6913         (fmai_vmfnmadd_<mode><round_name>): Likewise.
6915 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6917         PR target/72782
6918         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
6919         __builtin_ia32_vfmsubpd512_mask.
6920         (_mm512_mask_fmsub_round_pd): Likewise.
6921         (_mm512_fmsub_pd): Likewise.
6922         (_mm512_mask_fmsub_pd): Likewise.
6923         (_mm512_maskz_fmsub_round_pd): Use
6924         __builtin_ia32_vfmsubpd512_maskz.
6925         (_mm512_maskz_fmsub_pd): Likewise.
6926         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
6927         (_mm512_mask_fmsub_round_ps): Likewise.
6928         (_mm512_fmsub_ps): Likewise.
6929         (_mm512_mask_fmsub_ps): Likewise.
6930         (_mm512_maskz_fmsub_round_ps): Use
6931         __builtin_ia32_vfmsubps512_maskz.
6932         (_mm512_maskz_fmsub_ps): Likewise.
6933         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
6934         __builtin_ia32_vfmsubpd256_mask.
6935         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
6936         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6937         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
6938         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6939         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6940         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
6941         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
6942         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
6943         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
6944         __builtin_ia32_vfmsubpd.
6945         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
6946         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
6947         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
6948         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
6949         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
6950         * config/i386/i386-builtin.def: Add
6951         __builtin_ia32_vfmsubpd256_mask,
6952         __builtin_ia32_vfmsubpd256_maskz,
6953         __builtin_ia32_vfmsubpd128_mask,
6954         __builtin_ia32_vfmsubpd128_maskz,
6955         __builtin_ia32_vfmsubps256_mask,
6956         __builtin_ia32_vfmsubps256_maskz,
6957         __builtin_ia32_vfmsubps128_mask,
6958         __builtin_ia32_vfmsubps128_maskz,
6959         __builtin_ia32_vfmsubpd512_mask,
6960         __builtin_ia32_vfmsubpd512_maskz,
6961         __builtin_ia32_vfmsubps512_mask,
6962         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
6963         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
6964         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
6965         __builtin_ia32_vfmsubpd256.
6966         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
6967         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
6968         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
6969         Likewise.
6970         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
6971         Likewise.
6972         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
6973         Likewise.
6974         (fmai_vmfmsub_<mode><round_name>): Likewise.
6976 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6978         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
6979         Remove plus.  Renamed to ...
6980         (*sub<mode>3<mask_name>_bcst): This.
6981         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
6982         (*add<mode>3<mask_name>_bcst): This.
6984 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6986         PR target/72782
6987         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
6989 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6991         PR target/87662
6992         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
6993         (_mm_or_epi32): Likewise.
6994         (_mm256_xor_epi32): Likewise.
6995         (_mm_xor_epi32): Likewise.
6996         (_mm256_or_epi64): Likewise.
6997         (_mm_or_epi64): Likewise.
6998         (_mm256_xor_epi64): Likewise.
6999         (_mm_xor_epi64): Likewise.
7001 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
7003         PR target/72782
7004         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
7006 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
7008         PR middle-end/87647
7009         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
7011 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
7013         * doc/ux.texi: Move @section directly after @node.
7015 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
7017         PR middle-end/85488
7018         PR middle-end/87649
7019         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
7020         depend closely nested inside of loop with ordered clause with
7021         a parameter.
7023 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
7025         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
7026         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
7027         * doc/ux.texi: New file.
7029 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
7031         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
7032         be the first CR field allocated.
7034 2018-10-19  Richard Biener  <rguenther@suse.de>
7036         PR target/87657
7037         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
7038         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
7040 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
7042         PR target/72782
7043         * config/i386/sse.md
7044         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
7045         (*add<mode>3<mask_name>_bcst_2): Likewise.
7047 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
7049         * config/i386/sse.md
7050         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7051         Replace nonimmediate_operand with register_operand.
7052         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7053         Likewise.
7054         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7055         Likewise.
7057 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7059         PR rtl-optimization/87596
7060         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
7061         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
7062         for instructions in FROM..TO range.
7064 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
7066         * cfgexpand.c (expand_one_var): Use specific wording in error message
7067         for non-local frame variables.
7068         * stor-layout.c (layout_decl): Do not issue a warning for them.
7070 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
7072         * haifa-sched.c (priority): Add force_recompute parameter.
7073         (apply_replacement): Call priority () with force_recompute = true.
7074         (restore_pattern): Likewise.
7076 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7078         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
7079         HOST_BITS_PER_WIDE_INT.
7080         (test_vector_ops_duplicate): Likewise.
7082 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7084         PR target/72782
7085         * config/i386/sse.md (VF_AVX512): New.
7086         (avx512bcst): Likewise.
7087         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7088         Likewise.
7089         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7090         Likewise.
7091         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7092         Likewise.
7094 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
7096         * doc/invoke.texi (-dumpversion): Improve grammar.
7097         (-dumpfullversion): Make more consistent with -dumpversion.
7099 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
7101         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
7102         Set JUMP_LABEL to the jump insn.
7103         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
7104         Predict emitted jump and add label to jump insn.
7106 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
7108         PR tree-optimization/87562
7109         * input.c (get_substring_ranges_for_loc): Use
7110         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
7111         getting the linemap for the endpoint.  Verify that it's either
7112         in the same linemap as the start point's spelling location, or
7113         at least in the same file.
7115 2018-10-18  Richard Biener  <rguenther@suse.de>
7117         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
7118         feed width-specific load/store costs through ix86_vec_cost.
7119         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
7120         (k8_cost): Likewise.
7121         (bdver_cost): Likewise.
7122         (znver1_cost): Likewise.
7123         (btver1_cost): Likewise.
7124         (btver2_cost): Likewise.
7126 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7128         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
7129         to simplify subreg of vec_merge.
7131 2018-10-18  Richard Biener  <rguenther@suse.de>
7133         * config/i386/i386.c: Fix costing of vector FMA.
7135 2018-10-18  Richard Biener  <rguenther@suse.de>
7137         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
7138         and argument.
7139         (ix86_builtin_vectorization_cost): For vec_construct properly
7140         cost insertion into SSE regs.
7141         (...): Adjust calls to ix86_vec_cost.
7143 2018-10-18  Richard Biener  <rguenther@suse.de>
7145         PR middle-end/87087
7146         Revert
7147         2018-02-07  Richard Biener  <rguenther@suse.de>
7149         PR tree-optimization/84204
7150         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7151         this place.
7153 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
7155         PR target/87537
7156         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
7157         of vec_duplicate.
7158         (test_vector_ops_duplicate): Add test for a scalar subreg of a
7159         VEC_MERGE of a VEC_DUPLICATE.
7161 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
7163         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
7164         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
7165         * doc/standards.texi (C Language): Document C2X.
7166         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
7167         (rl78_option_override): Handle "GNU C2X" language name.
7169 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
7171         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
7172         Document C17 as published in 2018.
7174 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
7176         PR middle-end/87623
7177         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
7178         bail out if both sides do not have the same storage order.
7180 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
7182         * bitmap.c (bitmap_head::dump): New.
7183         * bitmap.h (bitmap_head): Add dump().
7184         * gimple-ssa-evrp-analyze.c
7185         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
7186         (evrp_range_analyzer::set_ssa_range_info): Same.
7187         (evrp_range_analyzer::record_ranges_from_phis): Same.
7188         (evrp_range_analyzer::record_ranges_from_stmt): Same.
7189         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
7190         * gimple-ssa-sprintf.c (get_int_range): Same.
7191         (format_integer): Same.
7192         (sprintf_dom_walker::handle_gimple_call): Same.
7193         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
7194         (ipcp_vr_lattice::top_p): Same.
7195         (ipcp_vr_lattice::bottom_p): Same.
7196         (ipcp_vr_lattice::set_to_bottom): Same.
7197         (ipa_vr_operation_and_type_effects): Same.
7198         (propagate_vr_across_jump_function): Same.
7199         (ipcp_store_vr_results): Same.
7200         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
7201         (ipa_print_node_jump_functions_for_edge): Same.
7202         (ipa_get_value_range): Same.
7203         (ipa_compute_jump_functions_for_edge): Same.
7204         (ipa_write_jump_function): Same.
7205         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
7206         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
7207         Same.
7208         * vr-values.c (set_value_range_to_nonnegative): Same.
7209         (set_value_range_to_truthvalue): Same.
7210         (vr_values::get_value_range): Same.
7211         (vr_values::set_defs_to_varying): Same.
7212         (vr_values::update_value_range): Same.
7213         (symbolic_range_based_on_p): Same.
7214         (vr_values::op_with_boolean_value_range_p): Same.
7215         (vr_values::extract_range_for_var_from_comparison_expr): Same.
7216         (vr_values::extract_range_from_ssa_name): Same.
7217         (vr_values::extract_range_from_binary_expr): Same.
7218         (vr_values::extract_range_from_unary_expr): Same.
7219         (vr_values::extract_range_from_cond_expr): Same.
7220         (vr_values::extract_range_from_comparison): Same.
7221         (vr_values::check_for_binary_op_overflow): Same.
7222         (vr_values::extract_range_basic): Same.
7223         (vr_values::extract_range_from_assignment): Same.
7224         (compare_ranges): Same.
7225         (compare_range_with_value): Same.
7226         (vr_values::adjust_range_with_scev): Same.
7227         (vrp_valueize): Same.
7228         (vrp_valueize_1): Same.
7229         (vr_values::get_vr_for_comparison): Same.
7230         (vr_values::compare_name_with_value): Same.
7231         (vr_values::compare_names): Same.
7232         (vr_values::vrp_evaluate_conditional): Same.
7233         (find_case_label_ranges): Same.
7234         (vr_values::vrp_visit_switch_stmt): Same.
7235         (vr_values::extract_range_from_phi_node): Same.
7236         (vr_values::simplify_div_or_mod_using_ranges): Same.
7237         (vr_values::simplify_bit_ops_using_ranges): Same.
7238         (test_for_singularity): Same.
7239         (range_fits_type_p): Same.
7240         (vr_values::simplify_cond_using_ranges_1): Same.
7241         (vr_values::simplify_switch_using_ranges): Same.
7242         (vr_values::simplify_float_conversion_using_ranges): Same.
7243         (vr_values::two_valued_val_range_p): Same.
7244         (vr_values::add_equivalence): Move to value_range::equiv_add.
7245         * vr-values.h (vr_values::add_equivalence): Remove.
7246         (VR_INITIALIZER): Remove.
7247         * tree-vrp.c (value_range::set): New.
7248         (value_range::equiv_add): New.
7249         (value_range::value_range): New.
7250         (value_range::deep_copy): New.
7251         (value_range::check): New.
7252         (value_range::equal_p): New.
7253         (value_range::ignore_equivs_equal_p): New.
7254         (value_range::operator==): New.
7255         (value_range::operator!=): New.
7256         (value_range::symbolic_p): New.
7257         (value_range::numeric_p): New.
7258         (value_range::set_undefined): New.
7259         (value_range::set_varying): New.
7260         (value_range::may_contain_p): New.
7261         (value_range::equiv_clear): New.
7262         (value_range::singleton_p): New.
7263         (value_range::intersect): New.
7264         (value_range::dump): New.
7265         (value_range::set_and_canonicalize): New.
7266         (set_value_range): Adjust for value_range API.
7267         (set_value_range_to_undefined): Same.
7268         (set_value_range_to_varying): Same.
7269         (set_and_canonicalize_value_range): Same.
7270         (set_value_range_to_nonnull): Same.
7271         (set_value_range_to_null): Same.
7272         (range_is_null): Same.
7273         (range_is_nonnull): Same.
7274         (range_int_cst_p): Same.
7275         (range_int_cst_singleton_p): Same.
7276         (symbolic_range_p): Same.
7277         (range_includes_zero_p): Same.
7278         (value_range_constant_singleton): Same.
7279         (vrp_set_zero_nonzero_bits): Same.
7280         (ranges_from_anti_range): Same.
7281         (extract_range_into_wide_ints): Same.
7282         (extract_range_from_multiplicative_op): Same.
7283         (set_value_range_with_overflow): Same.
7284         (extract_range_from_binary_expr_1): Same.
7285         (extract_range_from_unary_expr): Same.
7286         (dump_value_range): Same.
7287         (debug_value_range): Same.
7288         (vrp_prop::check_array_ref): Same.
7289         (vrp_prop::check_mem_ref): Same.
7290         (vrp_prop::vrp_initialize): Same.
7291         (vrp_prop::visit_stmt): Same.
7292         (intersect_ranges): Same.
7293         (vrp_prop::visit_phi): Same.
7294         (vrp_prop::vrp_finalize): Same.
7295         (determine_value_range_1): Same.
7296         (determine_value_range): Same.
7297         (vrp_intersect_ranges_1): Rename to...
7298         (vrp_intersect_1): this.
7299         (vrp_intersect_ranges): Rename to...
7300         (value_range::intersect_helper): ...this.
7301         (vrp_meet_1): Rename to...
7302         (value_range::union_helper): ...this.
7303         (vrp_meet): Rename to...
7304         (value_range::union_): ...this.
7305         (copy_value_range): Remove.
7306         * tree-vrp.h (struct value_range): Rewrite into a proper class.
7307         (value_range::vrtype): New.
7308         (value_range::type): New.
7309         (value_range::equiv): New.
7310         (value_range::min): New.
7311         (value_range::max): New.
7312         (value_range::varying_p): New.
7313         (value_range::undefined_p): New.
7314         (value_range::null_p): New.
7315         (value_range::equiv_add): New.
7316         (copy_value_range): Remove.
7318 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
7320         * Makefile.in (SELFTEST_TARGETS): New.
7321         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
7322         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
7323         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
7324         c/Make-lang.in.
7325         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
7326         (selftest-c++-gdb, selftest-c++-valgrind): Move to
7327         cp/Make-lang.in.
7328         * configure: Regenerate.
7329         * configure.ac (selftest_languages): New.
7331 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
7333         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
7334         overflow wraps argument.
7335         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
7336         wide_int_range_multiplicative_op.
7337         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
7338         overflow wraps argument.
7339         (wide_int_range_multiplicative_op): Same.
7340         (wide_int_range_lshift): Same.
7341         (wide_int_range_div): Same.
7342         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
7343         (wide_int_range_lshift): Same.
7344         (wide_int_range_div): Same.
7346 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
7348         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
7349         use sign as argument.
7350         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
7351         wide_int_range_shift_undefined_p.
7353 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7355         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
7356         Rename to...
7357         (@despeculate_copy<ALLI_TI:mode>): ... This.
7358         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
7359         switch statement.
7361 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7363         * config.gcc: Obsolete *-*-solaris2.10*.
7364         * doc/install.texi (Specific, *-*-solaris2*): Document it.
7366 2018-10-12  Jeff Law  <law@redhat.com>
7368         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
7369         reg + sym +- const_int addressing modes.
7371 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
7373         * common.opt (fdiagnostics-minimum-margin-width=): New option.
7374         * diagnostic-show-locus.c (layout::layout): Apply the minimum
7375         margin width.
7376         (layout::start_annotation_line): Only print up to 3 of the
7377         margin character, to avoid touching the left-hand side.
7378         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
7379         minimum margin width, as set by test_diagnostic_context's ctor.
7380         (selftest::test_fixit_insert_containing_newline): Likewise.
7381         (selftest::test_fixit_insert_containing_newline_2): Likewise.
7382         (selftest::test_line_numbers_multiline_range): Clear
7383         dc.min_margin_width.
7384         * diagnostic.c (diagnostic_initialize): Initialize
7385         min_margin_width.
7386         * diagnostic.h (struct diagnostic_context): Add field
7387         "min_margin_width".
7388         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
7389         * opts.c (common_handle_option): Handle
7390         OPT_fdiagnostics_minimum_margin_width_.
7391         * selftest-diagnostic.c
7392         (selftest::test_diagnostic_context::test_diagnostic_context):
7393         Initialize min_margin_width to 6.
7394         * toplev.c (general_init): Initialize global_dc->min_margin_width.
7396 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
7398         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
7399         Fix usage of "error_at_rich_loc" in the comment.
7401 2018-10-15  Renlin Li  <renlin.li@arm.com>
7403         PR target/87563
7404         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
7405         if-conversioned loop when it contains ifn with types not
7406         supported by backend.
7407         * internal-fn.c (expand_direct_optab_fn): Add an assert.
7408         (direct_internal_fn_supported_p): New helper function.
7409         * internal-fn.h (direct_internal_fn_supported_p): Declare.
7411 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
7413         PR target/87572
7414         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
7415         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
7416         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
7418 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
7420         PR tree-optimization/87022
7421         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
7422         bits in dist vector rather than the first one.
7424 2018-10-15  Richard Biener  <rguenther@suse.de>
7426         PR middle-end/87610
7427         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
7428         (visit_loadstore): When a used restrict tag escaped verify that
7429         the points-to solution of "other" pointers do not include
7430         escaped.
7431         (compute_dependence_clique): If a used restrict tag escaped
7432         communicated that down to visit_loadstore.
7434 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
7436         * config/s390/s390.c (s390_expand_vec_init): Force vector element
7437         into reg if it isn't a general operand.
7439 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
7441         PR target/87599
7442         * config/i386/sse.md (*vec_dupv2di): Add register source to
7443         movddup.
7445 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
7447         PR target/87572
7448         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
7449         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
7450         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
7451         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
7453 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
7455         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
7456         (notice_args_size): Set it in the current trace if no insn that can
7457         throw internally has been seen yet.
7458         (connect_traces): When connecting args_size between traces, allow the
7459         incoming values not to match if there is an insn setting it before the
7460         first insn that can throw internally; in that case, force the creation
7461         of a CFI note on this latter insn.
7463 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
7465         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
7466         the base class, because using the injected-class-name was not clearly
7467         specified until DR 176.
7469 2018-10-12  Paul Koning  <ni1d@arrl.net>
7471         * config/pdp11/pdp11.md (doloop_end): New expander.
7472         (doloop_end_insn): renamed from "doloop_end".
7473         (addqi3): New pattern.
7474         (subqi3): New pattern.
7475         * config/pdp11/predicates.md (incdec_operand): New predicate.
7477 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
7479         PR middle-end/81376
7480         * real.c (format_helper::can_represent_integral_type_p): New function
7481         * real.h (format_helper::can_represent_integral_type_p): Ditto.
7482         * match.pd: New pattern.
7484 2018-10-12  Alexandre Oliva <oliva@adacore.com>
7486         * configure.ac: Introduce --enable-large-address-aware
7487         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
7488         * doc/install.texi: Document it.
7489         * configure, config.in: Rebuilt.
7490         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
7491         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
7492         (LINK_SPEC): Insert it.
7493         * config/i386/mingw-w64.h: Likewise.
7495         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
7497 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
7499         PR rtl-optimization/87600
7500         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
7502 2018-10-12  Paul Koning  <ni1d@arrl.net>
7504         * doc/md.texi (doloop_end): Document that the pattern code may
7505         need to check operand mode.
7507 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
7509         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
7510         to zero-extend between int and floating-point registers.
7511         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
7512         ldp into floating-point registers.  Add type and arch attributes.
7513         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
7514         Use f_loads for type attribute.
7516 2018-10-11  Martin Sebor  <msebor@redhat.com>
7518         * doc/extend.texi (attribute packed): Correct typos.
7520 2018-10-11  Martin Sebor  <msebor@redhat.com>
7522         * doc/extend.texi (attribute flatten): Mention interaction with
7523         noinline.
7525 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
7527         PR target/87156
7528         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7529         Set new_decl virtual flag to zero.
7531 2018-10-11  Martin Sebor  <msebor@redhat.com>
7533         PR middle-end/87593
7534         * doc/extend.texi (attribute format_arg): Discuss using multiple
7535         attributes on a single function.
7537 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
7539         PR tree-optimization/86829
7540         * match.pd (sin (atan (x))): New simplification rules.
7541         (cos (atan (x))): Likewise.
7542         * real.c (build_sinatan_real): New function.
7543         * real.h (build_sinatan_real): Prototype.
7545 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
7547         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
7548         function.
7549         (fold_mergeeo_helper): New helper function.
7550         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
7551         intrinsics.  Correct some whitespace indentation issues.
7553 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
7555         PR target/87511
7556         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
7557         Use HOST_WIDE_INT_1U for shift.
7559 2018-10-11  Doug Rupp  <rupp@adacore.com>
7560             Olivier Hainque  <hainque@adacore.com>
7562         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
7563         Pass --relax to the linker for RTPs.
7564         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
7566 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
7567             Jan Hubicka  <jh@suse.cz>
7568             Martin Jambor  <mjambor@suse.cz>
7570         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
7571         the same elements are repeated rather than printing all of them.
7572         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
7573         "repeated" elements.
7574         * read-rtl-function.c (test_loading_repeat): New function.
7575         (read_rtl_function_c_tests): Call test_loading_repeat.
7576         * rtl-tests.c (test_dumping_repeat): New function.
7577         (rtl_tests_c_tests): Call test_dumping_repeat.
7579 2018-10-11  Richard Biener  <rguenther@suse.de>
7581         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
7582         bdver?_cost): Unify to ...
7583         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
7584         * config/i386/i386.c (processor_cost_table): Adjust.
7586 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
7588         PR middle-end/87574
7589         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
7590         the thunk when expanding to GIMPLE.
7592 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7594         PR bootstrap/87551
7595         * varasm.c (mergeable_string_section): Don't try to move zero-length
7596         strings to the merge section.
7598 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
7600         PR target/87573
7601         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
7603 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
7605         PR target/87550
7606         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
7607         to special_args set.
7609 2018-10-10  Richard Biener  <rguenther@suse.de>
7611         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
7612         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
7613         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
7614         and recursing and pattern terminating the recursion on SSE
7615         vector width using ix86_expand_reduc.
7616         (reduc_sminmax_scal_<mode>): Split into part reducing to half
7617         width and recursing and SSE2 vector variant doing the final
7618         reduction with ix86_expand_reduc.
7619         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
7620         with terminating the recursion at AVX level, splitting that
7621         to SSE there.
7623 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
7625         * genmatch.c (error_cb): Rename to...
7626         (diagnostic_cb): ...this, converting int params to enums.
7627         (fatal_at): Update for renaming.
7628         (warning_at): Likewise.
7629         (main): Likewise.
7630         * input.c (selftest::ebcdic_execution_charset::apply):
7631         Update for renaming of...
7632         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
7633         to...
7634         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
7635         converting level and reason to enums.
7636         (class selftest::lexer_error_sink): Rename to...
7637         (class selftest::lexer_test_options): ...this, renaming field
7638         "m_errors" to "m_diagnostics".
7639         (selftest::lexer_test_options::apply): Update for renaming of...
7640         (selftest::lexer_test_options::on_error): ...this, renaming to...
7641         (selftest::lexer_test_options::on_diagnostic): ...this
7642         converting level and reason to enums.
7643         (selftest::test_lexer_string_locations_raw_string_unterminated):
7644         Update for renamings.
7645         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
7646         "reason".
7648 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
7650         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
7651         * config/rs6000/pmmintrin.h: New file.
7653 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
7655         PR tree-optimization/86659
7656         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
7658 2018-10-09  Richard Biener  <rguenther@suse.de>
7660         PR tree-optimization/63155
7661         * tree-ssa-structalias.c: Include tree-ssa.h.
7662         (get_constraint_for_ssa_var): For undefs return nothing_id.
7663         (find_func_aliases): Cleanup PHI handling.
7665 2018-10-09  Richard Biener  <rguenther@suse.de>
7667         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
7668         replacements.
7670 2018-10-09  Martin Liska  <mliska@suse.cz>
7672         * asan.c (asan_emit_stack_protection): If a stack variable
7673         is located in a same file as current function, then emit
7674         line info into variable definition string.
7676 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7678         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
7679         information.
7681 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7683         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
7684         on the thunk.
7686 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7688         PR tree-optimization/86659
7689         * gimple-match.h (struct gimple_match_op): Add reverse field.
7690         (gimple_match_op::set_op): New overloaded method.
7691         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
7692         the REF_REVERSE_STORAGE_ORDER flag on the value.
7693         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
7694         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
7696 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
7698         PR middle-end/63155
7699         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
7700         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
7702 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
7704         PR target/87517
7705         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
7706         Defined with __builtin_ia32_vfmaddsubpd512_mask.
7708 2018-10-08  Richard Biener  <rguenther@suse.de>
7710         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
7711         cost the same as AVX128 ones.
7713 2018-10-08  Paul Koning  <ni1d@arrl.net>
7715         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
7716         (expand_block_move): New function.
7717         * config/pdp11/pdp11.c (output_block_move): Remove.
7718         (expand_block_move): New function.
7719         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
7720         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
7721         (*movmemhi1): Remove.
7723 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
7725         * config/s390/2827.md: Increase latencies for some FP instructions.
7727 2018-10-08  Richard Biener  <rguenther@suse.de>
7729         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7730         Open a dump scope.
7731         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
7732         * tree-vectorizer.h (dump_stmt_cost): Adjust.
7733         (add_stmt_cost): Dump return value of the hook.
7735 2018-10-08  Richard Biener  <rguenther@suse.de>
7737         PR tree-optimization/63155
7738         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
7739         (ssa_propagation_engine::ssa_propagate): Remove redundant
7740         bitmap bit clearing.
7742 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
7744         PR rtl-optimization/86939
7745         PR rtl-optimization/87479
7746         * ira.h (non_conflicting_reg_copy_p): New prototype.
7747         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
7748         (make_hard_regno_dead): Don't add conflicts for register
7749         ignore_reg_for_conflicts.
7750         (make_object_dead): Likewise.
7751         (non_conflicting_reg_copy_p): New function.
7752         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
7753         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
7754         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
7755         (make_hard_regno_dead): Don't add conflicts for register
7756         ignore_reg_for_conflicts.  Remove special conflict handling of
7757         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
7758         check_pic_pseudo_p and update callers.
7759         (mark_pseudo_dead): Don't add conflicts for register
7760         ignore_reg_for_conflicts.
7761         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
7763 2018-10-05  Andrew Waterman  <andrew@sifive.com>
7764             Jim Wilson  <jimw@sifive.com>
7766         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
7767         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
7768         new pattern using HONOR_SNANS that emits one extra instruction.
7770 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7772         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
7773         patterns): Merge SI and DI patterns to a GPR pattern.
7774         (unnamed define_insn and define_split for record form of that): Merge
7775         to a single define_insn_and_split pattern.
7777 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
7779         PR c++/56856
7780         * input.c (expand_location_to_spelling_point): Add param "aspect"
7781         and use rather than hardcoding LOCATION_ASPECT_CARET.
7782         (get_substring_ranges_for_loc): Handle the case of a single token
7783         within a macro expansion.
7784         * input.h (expand_location_to_spelling_point): Add "aspect" param,
7785         defaulting to LOCATION_ASPECT_CARET.
7787 2018-10-05  Paul Koning  <ni1d@arrl.net>
7789         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
7790         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
7791         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
7792         (pdp11_guard_type): New function.
7794 2018-10-05  Paul Koning  <ni1d@arrl.net>
7796         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
7797         * config/pdp11/pdp11.opt (mfloat32): Remove.
7798         (mfloat64): Remove.
7799         * doc/invoke.texi (pdp11 -mfloat32): Remove:
7800         (pdp11 -mfloat64): Remove.
7802 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7804         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
7805         (*cmp<mode>_cc_i387): Ditto.
7806         (*cmpu<mode>_cc_i387): Ditto.
7807         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
7808         * config/i386/i386.c (ix86_expand_fp_compare): Remove
7809         "scratch" argument.
7810         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
7811         Emit x86_sahf_1 pattern.
7812         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
7813         (ix86_expand_carry_flag_compare): Ditto.
7815 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7817         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
7818         to reg_or_0_operand.  Add "C" constraint.
7819         (*cmpxf_cc_i387): Ditto.
7820         (*cmp<mode>_i387): Change operand 2 predicate
7821         to nonimm_or_0_operand.  Add "C" constraint.
7822         (*cmp<mode>_cc_i387): Ditto.
7823         (*cmp<mode>_0_i387): Remove insn pattern.
7824         (*cmp<mode>_0_cc_i387): Ditto.
7826 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7828         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
7829         * config/i386/predicates.md (nonimm_or_0_operand): Rename
7830         from vector_move_operand.  Update all uses.
7832 2018-10-05  Martin Sebor  <msebor@redhat.com>
7834         PR tree-optimization/87490
7835         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
7836         consistently.
7838 2018-10-05  Richard Biener  <rguenther@suse.de>
7840         PR tree-optimization/63155
7841         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
7842         vertical space in dumpfiles.
7843         * tree-ssa-propagate.h
7844         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
7845         * tree-ssa-propagate.c (cfg_blocks_back): New global.
7846         (ssa_edge_worklist_back): Likewise.
7847         (curr_order): Likewise.
7848         (cfg_blocks_get): Remove abstraction.
7849         (cfg_blocks_add): Likewise.
7850         (cfg_blocks_empty_p): Likewise.
7851         (add_ssa_edge): Add to current or next worklist based on
7852         RPO index.
7853         (add_control_edge): Likewise.
7854         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
7855         into ...
7856         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
7857         iteration from CFG and SSA edge worklist so we process
7858         everything in RPO order, prioritizing forward progress
7859         over iteration.
7860         (ssa_prop_init): Allocate new worklists, do not dump
7861         immediate uses.
7862         (ssa_prop_fini): Free new worklists.
7864 2018-10-05  Richard Biener  <rguenther@suse.de>
7866         * tree-core.h (tree_block::abstract_flag): Remove.
7867         (tree_block::block_num): Make full 32bits.
7868         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
7869         * tree.h (BLOCK_ABSTRACT): Remove.
7870         * dwarf2out.c (gen_lexical_block_die): Remove dead code
7871         resulting from BLOCK_ABSTRACT being always false.
7872         (gen_inlined_subroutine_die): Likewise.
7873         (gen_block_die): Likewise.
7874         * tree.c (block_ultimate_origin): Likewise.
7875         * tree-pretty-print.c (dump_block_node): Remove code dealing
7876         with BLOCK_ABSTRACT.
7877         * tree-ssa-live.c (dump_scope_block): Likewise.
7878         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
7879         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
7881 2018-10-05   Richard Biener  <rguenther@suse.de>
7883         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
7884         is asked for initialize mode to the component mode of the
7885         vector type.
7887 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
7889         PR target/87522
7890         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
7891         assembler for -mavx.
7892         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
7894 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7896         PR target/87509
7897         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
7898         RS6000_BTM_DFP.
7899         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
7900         to be DImode.  When using mffscrn, force the operand to a register.
7902 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
7904         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
7905         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
7906         X87MODEF mode iterator.
7907         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
7908         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
7909         X87MODEF mode iterator.
7911 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
7913         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
7914         -Wno-prio-ctor-dtor.
7916 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7918         * Makefile.in (OBJS): Add opt-problem.o.
7919         * dump-context.h: Include "selftest.h.
7920         (selftest::temp_dump_context): New forward decl.
7921         (class dump_context): Make friend of class
7922         selftest::temp_dump_context.
7923         (dump_context::dump_loc_immediate): New decl.
7924         (class dump_pretty_printer): Move here from dumpfile.c.
7925         (class temp_dump_context): Move to namespace selftest.
7926         (temp_dump_context::temp_dump_context): Add param
7927         "forcibly_enable_dumping".
7928         (selftest::verify_dumped_text):
7929         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
7930         (selftest::verify_item):
7931         (ASSERT_IS_TEXT): Move here from dumpfile.c.
7932         (ASSERT_IS_TREE): Likewise.
7933         (ASSERT_IS_GIMPLE): Likewise.
7934         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
7935         to...
7936         (dump_context::dump_loc_immediate): ...this new function.
7937         (class dump_pretty_printer): Move to dump-context.h.
7938         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
7939         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
7940         (temp_dump_context::temp_dump_context): Move to "selftest"
7941         namespace.  Add param "forcibly_enable_dumping", and use it to
7942         conditionalize the use of m_pp;
7943         (selftest::verify_dumped_text): Make non-static.
7944         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
7945         (selftest::verify_item): Make non-static.
7946         (ASSERT_IS_TEXT): Move to dump-context.h.
7947         (ASSERT_IS_TREE): Likewise.
7948         (ASSERT_IS_GIMPLE): Likewise.
7949         (selftest::test_capture_of_dump_calls): Pass "true" for new
7950         param of temp_dump_context.
7951         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
7952         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
7953         TDF_COMPARE_DEBUG.
7954         * opt-problem.cc: New file.
7955         * opt-problem.h: New file.
7956         * optinfo-emit-json.cc
7957         (selftest::test_building_json_from_dump_calls): Pass "true" for
7958         new param of temp_dump_context.
7959         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
7960         (optinfo::emit_for_opt_problem): New function.
7961         (optinfo::emit): Clarity which emit_item is used.
7962         * optinfo.h (optinfo::get_dump_location): New accessor.
7963         (optinfo::emit_for_opt_problem): New decl.
7964         (optinfo::emit): Make const.
7965         * selftest-run-tests.c (selftest::run_tests): Call
7966         selftest::opt_problem_cc_tests.
7967         * selftest.h (selftest::opt_problem_cc_tests): New decl.
7968         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
7969         bool to opt_result, converting fprintf messages to
7970         opt_result::failure_at calls.  Add "stmt" param for use by the
7971         failure_at calls.
7972         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
7973         (runtime_alias_check_p): Convert return type from bool to
7974         opt_result, converting dump_printf calls to
7975         opt_result::failure_at, using the statement DDR_A for their
7976         location.
7977         (find_data_references_in_stmt): Convert return type from bool to
7978         opt_result, converting "return false" to opt_result::failure_at
7979         with a new message.
7980         * tree-data-ref.h: Include "opt-problem.h".
7981         (dr_analyze_innermost): Convert return type from bool to opt_result,
7982         and add a const gimple * param.
7983         (find_data_references_in_stmt): Convert return type from bool to
7984         opt_result.
7985         (runtime_alias_check_p): Likewise.
7986         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
7987         dr_analyze_innermost.
7988         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
7989         Convert return type from bool to opt_result, adding a message for
7990         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
7991         (vect_analyze_data_ref_dependence): Convert return type from bool
7992         to opt_result.  Change sense of return type from "false"
7993         effectively meaning "no problems" to "false" meaning a problem,
7994         so that "return false" becomes "return opt_result::success".
7995         Convert "return true" calls to opt_result::failure_at, using
7996         the location of statement A rather than vect_location.
7997         (vect_analyze_data_ref_dependences): Convert return type from bool
7998         to opt_result.
7999         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
8000         calls to opt_result::failure_at, using the stmt location rather
8001         than vect_location.
8002         (vect_verify_datarefs_alignment): Convert return type from bool
8003         to opt_result.
8004         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
8005         into multiple more-tightly-scoped copies.
8006         (vect_analyze_data_refs_alignment): Convert return type from bool
8007         to opt_result.
8008         (vect_analyze_data_ref_accesses): Likewise, converting a
8009         "return false" to a "return opt_result::failure_at", adding a
8010         new message.
8011         (vect_prune_runtime_alias_test_list): Convert return type from
8012         bool to opt_result, converting dump_printf_loc to
8013         opt_result::failure_at.  Add a %G to show the pertinent statement,
8014         and use the stmt's location rather than vect_location.
8015         (vect_find_stmt_data_reference): Convert return type from
8016         bool to opt_result, converting dump_printf_loc to
8017         opt_result::failure_at, using stmt's location.
8018         (vect_analyze_data_refs):  Convert return type from bool to
8019         opt_result.  Convert "return false" to "return
8020         opt_result::failure_at", adding messages as needed.
8021         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
8022         type from bool to opt_result.
8023         (vect_determine_vf_for_stmt): Likewise.
8024         (vect_determine_vectorization_factor): Likewise, converting
8025         dump_printf_loc to opt_result::failure_at, using location of phi
8026         rather than vect_location.
8027         (vect_analyze_loop_form_1): Convert return type from bool to
8028         opt_result, converting dump_printf_loc calls, retaining the use of
8029         vect_location.
8030         (vect_analyze_loop_form): Convert return type from loop_vec_info
8031         to opt_loop_vec_info.
8032         (vect_analyze_loop_operations): Convert return type from bool to
8033         opt_result, converting dump_printf_loc calls, using the location
8034         of phi/stmt rather than vect_location where available.  Convert
8035         various "return false" to "return opt_result::failure_at" with
8036         "unsupported phi" messages.
8037         (vect_get_datarefs_in_loop): Convert return type from bool to
8038         opt_result.  Add a message for the
8039         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
8040         (vect_analyze_loop_2): Convert return type from bool to
8041         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
8042         each "goto again;", adding new messages where needed.
8043         Add "unsupported grouped {store|load}" messages.
8044         (vect_analyze_loop): Convert return type from loop_vec_info to
8045         opt_loop_vec_info.
8046         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
8047         bool to opt_result.
8048         * tree-vect-stmts.c (process_use): Likewise, converting
8049         dump_printf_loc call and using stmt location, rather than
8050         vect_location.
8051         (vect_mark_stmts_to_be_vectorized): Likeise.
8052         (vect_analyze_stmt): Likewise, adding a %G.
8053         (vect_get_vector_types_for_stmt): Convert return type from bool to
8054         opt_result, converting dump_printf_loc calls and using stmt
8055         location, rather than vect_location.
8056         (vect_get_mask_type_for_stmt): Convert return type from tree to
8057         opt_tree, converting dump_printf_loc calls and using stmt location.
8058         * tree-vectorizer.c: Include "opt-problem.h.
8059         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
8060         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
8061         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
8062         enabled, use it to report at the top level "couldn't vectorize
8063         loop" followed by the problem.
8064         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
8065         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
8066         to opt_result.
8067         (vect_analyze_stmt): Likewise.
8068         (vect_get_vector_types_for_stmt): Likewise.
8069         (tree vect_get_mask_type_for_stmt): Likewise.
8070         (vect_analyze_data_ref_dependences): Likewise.
8071         (vect_enhance_data_refs_alignment): Likewise.
8072         (vect_analyze_data_refs_alignment): Likewise.
8073         (vect_verify_datarefs_alignment): Likewise.
8074         (vect_analyze_data_ref_accesses): Likewise.
8075         (vect_prune_runtime_alias_test_list): Likewise.
8076         (vect_find_stmt_data_reference): Likewise.
8077         (vect_analyze_data_refs): Likewise.
8078         (vect_analyze_loop): Convert return type from loop_vec_info to
8079         opt_loop_vec_info.
8080         (vect_analyze_loop_form): Likewise.
8081         (vect_analyze_slp): Convert return type from bool to opt_result.
8083 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
8085         * doc/invoke.texi (-fopt-info): Document new "internals"
8086         sub-option.
8087         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
8088         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
8089         MSG_ALL_KINDS.
8090         (optinfo_verbosity_options): Add "internals".
8091         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
8092         (dump_context::apply_dump_filter_p): New member function.
8093         (dump_context::dump_loc): Use apply_dump_filter_p rather than
8094         explicitly masking the dump_kind.
8095         (dump_context::begin_scope): Increment the scope depth first.  Use
8096         apply_dump_filter_p rather than explicitly masking the dump_kind.
8097         (dump_context::emit_item): Use apply_dump_filter_p rather than
8098         explicitly masking the dump_kind.
8099         (dump_dec): Likewise.
8100         (dump_hex): Likewise.
8101         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
8102         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
8103         (opt_info_switch_p): Update handling of default
8104         MSG_OPTIMIZED_LOCATIONS to cope with default of
8105         MSG_PRIORITY_USER_FACING.
8106         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
8107         masking the dump_kind.
8108         (selftest::test_capture_of_dump_calls): Update test_dump_context
8109         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
8110         than MSG_ALL.  Generalize scope test to be run at all four
8111         combinations of with/without MSG_PRIORITY_USER_FACING and
8112         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
8113         for each of the two values.
8114         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
8115         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
8116         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
8117         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
8118         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
8119         with MSG_PRIORITY_*.
8120         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
8121         dump messages as MSG_PRIORITY_USER_FACING.
8122         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
8123         about the interaction with MSG_PRIORITY_*.
8125 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8127         * varasm.c (output_constant): Add new parameter merge_strings.
8128         Make strings properly zero terminated in merge string sections.
8129         (mergeable_string_section): Don't fail if the last char is non-zero.
8130         (assemble_variable_contents): Handle merge string sections.
8131         (assemble_variable): Likewise.
8132         (assemble_constant_contents): Likewise.
8133         (output_constant_def_contents): Likewise.
8134         (output_constructor_array_range,
8135         output_constructor_regular_field): Adjust call to output_constant.
8136         (output_object_block): Adjust call to assemble_constant_contents
8137         and assemble_variable_contents.
8139 2018-10-04  Martin Liska  <mliska@suse.cz>
8141         PR c/87483
8142         * cgraphunit.c (process_function_and_variable_attributes):
8143         Warn about a function with alias attribute and a body.
8145 2018-10-04  Martin Liska  <mliska@suse.cz>
8147         PR ipa/82625
8148         * multiple_target.c (redirect_to_specific_clone): New function.
8149         (ipa_target_clone): Use it.
8150         * tree-inline.c: Fix comment.
8152 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
8154         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
8155         fields.
8156         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
8157         (gcc::dump_manager::register_pass): New member function, adapted
8158         from loop body in gcc::pass_manager::register_pass, adding a
8159         call to update_dfi_for_opt_info.
8160         (gcc::dump_manager::opt_info_enable_passes): Store the
8161         -fopt-info options into the new fields.  Move the loop
8162         bodies into...
8163         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
8164         function.
8165         * dumpfile.h (struct opt_pass): New forward decl.
8166         (gcc::dump_manager::register_pass): New decl.
8167         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
8168         (class gcc::dump_manager): Add fields "m_optgroup_flags",
8169         "m_optinfo_flags", and "m_optinfo_filename".
8170         * passes.c (gcc::pass_manager::register_pass): Move all of the
8171         dump-handling code to gcc::dump_manager::register_pass.
8173 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
8175         PR rtl-optimization/87466
8176         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
8177         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
8178         * doc/tm.texi: Regenerate.
8179         * ira-lives.c (process_bb_node_lives): Use the new target hook.
8180         * lra-lives.c (process_bb_lives): Likewise.
8181         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
8182         Define.
8184 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
8186         * params.c (add_params): Fix initialization.
8188 2018-10-04  Martin Liska  <mliska@suse.cz>
8190         PR gcov-profile/84107
8191         * tree-profile.c (init_ic_make_global_vars):
8192         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
8193         Come up with new ic_tuple* variables.  Emit
8194         __gcov_indirect_call{,_topn} variables.
8195         (gimple_gen_ic_profiler): Access the variable
8196         and emit gimple.
8197         (gimple_gen_ic_func_profiler): Access
8198         __gcov_indirect_call.callee field.
8199         (gimple_init_gcov_profiler): Use ptr_type_node.
8200         * value-prof.c (gimple_ic): Use ptr_type_node.
8202 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8204         PR tree-optimization/85787
8205         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
8206         malloc_candidate_p into this function and add support for
8207         detecting multiple phis.
8208         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
8210 2018-10-04  Martin Liska  <mliska@suse.cz>
8212         PR ipa/87491
8213         * ipa-inline.c (inline_to_all_callers_1):
8214         Call ultimate_alias_target for node being inlined.
8216 2018-10-03  Jeff Law  <law@redhat.com>
8218         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
8219         target's wchar_t.
8220         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
8221         * tree.h (get_typenode_from_name): Prototype.
8223 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
8225         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
8226         Change operand 2 predicate to nonimmediate_operand.
8227         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8229 2018-10-03  Martin Sebor  <msebor@redhat.com>
8230             Jeff Law  <law@redhat.com>
8232         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
8233         initialize it.
8234         (get_string_length): Detect unterminated arrays.
8235         (format_string): Same.
8236         (format_directive): Warn about unterminated arrays.
8237         (handle_gimple_call): Mark statements with no_warning as needed.
8239 2018-10-03  Jim Wilson  <jimw@sifive.com>
8241         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
8242         also define __riscv_abi_rve.  Delete trailing white space.
8244 2018-10-03  Paul Koning  <ni1d@arrl.net>
8246         Enable LRA register allocator for PDP11.
8247         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
8248         (R): Likewise.
8249         (D): Likewise.
8250         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
8251         * config/pdp11/pdp11.opt (-mlra): New option.
8252         * doc/invoke.texi (PDP-11 Options): Document -mlra.
8254 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
8256         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
8257         (*<absneg:code>extend<mode>xf2): Ditto.
8259 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
8261         PR tree-optimization/87415
8262         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
8263         precision fields.
8265 2018-10-02  Jeff Law  <law@redhat.com>
8267         * gimple-fold.c (get_range_strlen): Only set *nonstr when
8268         an unterminated string is discovered.  Bubble up range
8269         even for unterminated strings.
8270         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
8271         indicates the string was not terminated via NONSTR.
8273 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
8275         * tree-vrp.c (extract_range_from_unary_expr): Special case all
8276         pointer conversions.
8277         Do not do anything special for anti-ranges.
8279 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
8281         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
8282         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
8284 2018-10-03  Martin Liska  <mliska@suse.cz>
8286         PR gcov-profile/86109
8287         * coverage.c (coverage_begin_function): Do not
8288         mark lambdas as artificial.
8289         * tree-core.h (struct GTY): Remove tm_clone_flag
8290         and introduce new lambda_function.
8291         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
8293 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
8295         PR target/87474
8296         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
8297         P8_VECTOR and VSX are enabled.
8299 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
8301         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
8302         0x3907 as CPU model number.
8304 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
8306         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
8307         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
8308         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
8309         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
8310         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
8311         * config/s390/s390.md: Likewise. Rename also the cpu attribute
8312         value from arch12 to z14.
8314 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
8316         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
8317         (isinfxf2): Ditto.
8318         (isinf<mode>2): Ditto.
8320 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
8322         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
8323         before emitting fxam.  Perform calculations in XFmode.
8325 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
8327         * match.pd (((X /[ex] A) +- B) * A): New transformation.
8329 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
8331         PR middle-end/87319
8332         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
8333         * tree.c (signed_or_unsigned_type_for): Handle complex.
8335 2018-10-02  Jeff Law  <law@redhat.com>
8337         * gimple-fold.c (get_range_strlen): Remove dead code.
8339 2018-10-02  Martin Sebor  <msebor@redhat.com>
8340             Jeff Law  <law@redhat.com>
8342         * builtins.c (unterminated_array): Add new arguments.
8343         If argument is not terminated, bubble up size and exact
8344         state to callers.
8345         (expand_builtin_strnlen): Detect, avoid expanding
8346         and diagnose unterminated arrays.
8347         (c_strlen): Fill in offset of start of unterminated strings.
8348         * builtins.h (unterminated_array): Update prototype.
8350 2018-10-02  Richard Biener  <rguenther@suse.de>
8352         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
8353         of haddv4df, first reduce to SSE width and exploit the fact
8354         that we only need element zero with the reduction result.
8355         (reduc_plus_scal_v2df): Likewise.
8357 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
8359         * dojump.h (do_jump): Delete.
8360         (do_jump_1): Likewise.
8361         (split_comparison): Move around.
8362         * dojump.c (do_jump): Make static.
8363         (do_jump_1): Likewise.
8364         (jumpifnot): Move around.
8365         (jumpifnot_1): Likewise.
8366         (jumpif): Likewise.
8367         (jumpif_1): Likewise.
8368         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
8370 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
8372         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
8373         insns in the delay slot and add_insn_after for the jump insn.
8375 2018-10-02  Richard Biener  <rguenther@suse.de>
8377         * tree-inline.c (expand_call_inline): Use the location of
8378         the callee declaration for the inline-entry marker.
8379         * final.c (notice_source_line): Remove special-casing of
8380         NOTE_INSN_INLINE_ENTRY.
8382 2018-10-01  Carl Love  <cel@us.ibm.com>
8384         PR 69431
8385         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
8386         (__builtin_mtfsb0): New.
8387         (__builtin_mtfsb1): New.
8388         ( __builtin_set_fpscr_rn): New.
8389         (__builtin_set_fpscr_drn): New.
8390         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
8391         (rs6000_expand_set_fpscr_rn_builtin): Add.
8392         (rs6000_expand_set_fpscr_drn_builtin): Add.
8393         (rs6000_expand_builtin): Add case statement entries for
8394         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
8395         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
8396         RS6000_BUILTIN_MFFSL.
8397         (rs6000_init_builtins): Add ftype initialization and def_builtin
8398         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
8399         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
8400         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
8401         rs6000_mffscdrn): Add define_insn.
8402         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
8403         * doc/extend.texi: Add documentation for the builtins.
8405 2018-10-01  Richard Biener  <rguenther@suse.de>
8407         PR tree-optimization/87465
8408         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
8409         causing branch miscounts.
8411 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8413         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
8414         aarch64_option_default_param):  New.
8415         (params.h): Include.
8416         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
8417         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
8418         stack-clash protection validation code.
8420 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8422         * params.c (validate_param): New.
8423         (add_params): Use it.
8424         (set_param_value): Refactor param validation into validate_param.
8425         (diagnostic.h): Include.
8426         * diagnostic.h (diagnostic_ready_p): New.
8428 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8430         * params.c (set_param_value):
8431         Add index of parameter being validated.
8432         * common/common-target.def (option_validate_param): New.
8433         * common/common-targhooks.h (default_option_validate_param): New.
8434         * common/common-targhooks.c (default_option_validate_param): New.
8435         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
8436         * doc/tm.texi: Regenerate.
8438 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8440         PR target/86486
8441         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8442         Add validation for stack-clash parameters and set defaults.
8444 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8446         PR target/86486
8447         * configure.ac: Add stack-clash-protection-guard-size.
8448         * doc/install.texi: Document it.
8449         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
8450         * params.def: Update comment for guard-size.
8451         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
8452         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
8453         * configure: Regenerate.
8455 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8457         PR target/86486
8458         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
8459         STACK_DYNAMIC_OFFSET): New.
8460         * config/aarch64/aarch64.c (aarch64_layout_frame):
8461         Update outgoing args size.
8462         (aarch64_stack_clash_protection_alloca_probe_range,
8463         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
8465 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8467         PR target/86486
8468         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
8469         probe ranges.
8470         * target.def (stack_clash_protection_alloca_probe_range): New.
8471         (stack_clash_protection_final_dynamic_probe): Remove.
8472         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
8473         (default_stack_clash_protection_final_dynamic_probe): Remove.
8474         * targhooks.c: Likewise.
8475         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
8476         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
8477         * doc/tm.texi: Regenerate.
8479 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8481         PR target/86486
8482         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
8483         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
8484         aarch64_clamp_to_uimm12_shift): New.
8485         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
8486         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
8488 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
8490         PR target/86486
8491         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
8493 2018-10-01  Jeff Law  <law@redhat.com>
8494             Richard Sandiford <richard.sandiford@linaro.org>
8495             Tamar Christina  <tamar.christina@arm.com>
8497         PR target/86486
8498         * config/aarch64/aarch64.md
8499         (probe_stack_range): Add k (SP) constraint.
8500         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
8501         STACK_CLASH_MAX_UNROLL_PAGES): New.
8502         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
8503         stack probes for stack clash.
8504         (aarch64_allocate_and_probe_stack_space): New.
8505         (aarch64_expand_prologue): Use it.
8506         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
8507         (aarch64_sub_sp): Add emit_move_imm optional param.
8509 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
8511         PR tree-optimization/87261
8512         * match.pd: Remove trailing whitespace.
8513         Add (x & y) | ~(x | y) -> ~(x ^ y),
8514         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
8516 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
8518         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
8519         constraints.
8520         (ashlsi3_insn): Update instruction constraints.
8521         (ashrsi3_insn): Likewise.
8522         (rotrsi3): Likewise.
8523         (add_shift): Likewise.
8524         * config/arc/constraints.md (Csz): New 32 bit constraint. It
8525         avoids placing in the limm field small constants which, otherwise,
8526         could end into a small instruction.
8528 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
8530         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
8531         destination register is not odd-even.
8532         (umaddsidi4_split): Likewise.
8534 2018-10-01  Richard Biener  <rguenther@suse.de>
8536         * tree-inline.c (expand_call_inline): Store origin of fn
8537         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
8538         * tree.c (block_ultimate_origin): Simplify and do some
8539         checking.
8541 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
8543         * config/i386/mmx.md (EMMS): New int iterator.
8544         (emms): New int attribute.
8545         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
8546         EMMS int iterator.  Explicitly declare clobbers.
8547         (mmx_emms): Remove expander.
8548         (mmx_femms): Ditto.
8549         * config/i386/predicates.md (emms_operation): Remove predicate.
8550         (vzeroall_pattern): New predicate.
8551         (vzeroupper_pattern): Rename from vzeroupper_operation.
8552         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
8553         vzeroupper_pattern and vzeroall_pattern predicates.
8555 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
8557         PR rtl-optimization/86939
8558         * ira-lives.c (make_hard_regno_born): Rename from this...
8559         (make_hard_regno_live): ... to this.  Remove update to conflict
8560         information.  Update function comment.
8561         (make_hard_regno_dead): Add conflict information update.  Update
8562         function comment.
8563         (make_object_born): Rename from this...
8564         (make_object_live): ... to this.  Remove update to conflict information.
8565         Update function comment.
8566         (make_object_dead):  Add conflict information update.  Update function
8567         comment.
8568         (mark_pseudo_regno_live): Call make_object_live.
8569         (mark_pseudo_regno_subword_live): Likewise.
8570         (mark_hard_reg_dead): Update function comment.
8571         (mark_hard_reg_live): Call make_hard_regno_live.
8572         (process_bb_node_lives): Likewise.
8573         * lra-lives.c (make_hard_regno_born): Rename from this...
8574         (make_hard_regno_live): ... to this.  Remove update to conflict
8575         information.  Remove now uneeded check_pic_pseudo_p argument.
8576         Update function comment.
8577         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
8578         to conflict information.  Update function comment.
8579         (mark_pseudo_live): Remove update to conflict information.  Update
8580         function comment.
8581         (mark_pseudo_dead): Add conflict information update.
8582         (mark_regno_live): Call make_hard_regno_live.
8583         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
8584         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
8586 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
8588         PR target/87370
8589         * config/i386/i386.c (construct_container): Use TImode for
8590         BLKmode values in 2 integer registers.
8592 2018-09-29  Jeff Law  <law@redhat.com>
8594         * builtins.c (unterminated_array): Pass in c_strlen_data * to
8595         c_strlen rather than just a tree *.
8596         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
8597         Update recursive calls appropriately.  If caller did not provide a
8598         suitable data pointer, create a local one.  When a non-terminated
8599         string is discovered, bubble up information about the string via the
8600         c_strlen_data object.
8601         * builtins.h (c_strlen): Update prototype.
8602         (c_strlen_data): New structure.
8603         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
8604         For a type 2 call, if c_strlen indicates a non-terminated string
8605         use the length of the non-terminated string.
8606         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
8608 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
8610         PR target/87467
8611         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
8612         __m512d type for __A argument rather than __m512.
8614 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
8616         * match.pd (simple_comparison): Don't optimize if either operand is
8617         a function pointer when target needs function pointer canonicalization.
8619 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
8621         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
8622         power5 .. power9 to remove indirection.
8623         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
8624         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
8625         ASM_CPU_476_SPEC): Delete.
8626         (ASM_CPU_SPEC): Adjust.
8627         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
8628         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
8630 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
8632         * config.in: Delete HAVE_AS_DCI.
8633         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
8634         * config/rs6000/rs6000.h: Ditto.
8635         * configure.ac: Delete HAVE_AS_DCI.
8636         * configure: Regenerate.
8638 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
8640         * config.in (HAVE_AS_LWSYNC): Delete.
8641         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
8642         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
8643         do it as a .long .
8644         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
8645         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
8646         as a .long .
8647         * configure.ac: Delete HAVE_AS_LWSYNC.
8648         * configure: Regenerate.
8650 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
8651             Pierre-Marie de Rodat  <derodat@adacore.com>
8653         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
8654         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
8655         (cgraph_node::create_thunk): Add indirect_offset parameter.
8656         (thunk_adjust): Likewise.
8657         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
8658         and initialize the corresponding field with it.
8659         (cgraph_node::dump): Dump indirect_offset field.
8660         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
8661         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
8662         (thunk_adjust): Add indirect_offset parameter and deal with it.
8663         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
8664         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
8665         or if the thunk is external or local.  Fix formatting.  Do not chain
8666         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
8667         if any, in the GIMPLE representation.
8668         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
8669         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
8670         (input_node): Read indirect_offset field.
8671         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
8672         call to thunk_adjust.
8673         * tree-nested.c (struct nesting_info): Add thunk_p field.
8674         (create_nesting_tree): Set it.
8675         (convert_all_function_calls): Copy static chain from targets to thunks.
8676         (finalize_nesting_tree_1): Return early for thunks.
8677         (unnest_nesting_tree_1): Do not finalize thunks.
8678         (gimplify_all_functions): Do not gimplify thunks.
8680 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8682         * opt-suggestions.c (option_proposer::build_option_suggestions):
8683         Release "option_values".
8685 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8687         * coverage.c (get_coverage_counts): Convert problem-reporting dump
8688         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
8689         * dumpfile.c (kind_as_string): New function.
8690         (dump_loc): Rather than a hardcoded prefix of "note: ", use
8691         kind_as_string to vary the prefix based on dump_kind.
8692         (selftest::test_capture_of_dump_calls): Update for above.
8694 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8696         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
8697         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
8699 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8701         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
8702         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
8703         INVALID_REGNUM instead of FPSR_REG.
8704         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
8705         * config/i386/i386.md: Update comment of FP compares.
8706         (fldenv): Do not clobber FPSR_REG.
8708 2018-09-28  Richard Biener  <rguenther@suse.de>
8710         * tree.h (BLOCK_ORIGIN): New.
8711         * omp-expand.c (grid_expand_target_grid_body): Assign
8712         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
8713         * tree-inline.c (remap_block): Likewise.
8714         * auto-profile.c (get_function_decl_from_block): Simplify
8715         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
8716         * langhooks.c (lhd_print_error_function): Likewise.
8717         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
8718         Likewise.
8719         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8720         * tree.c (block_nonartificial_location): Likewise.
8721         (block_ultimate_origin): Likewise.
8722         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
8723         no longer needed LTO case.
8725 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
8726             Jan Hubicka  <jh@suse.cz>
8727             Martin Jambor  <mjambor@suse.cz>
8729         * simplify-rtx.c (simplify_merge_mask): New function.
8730         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
8731         same masks are used in op1 or op2.
8732         (test_vec_merge): New function.
8733         (test_vector_ops): Call test_vec_merge.
8735 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
8737         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
8738         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
8739         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
8740         (cypress_costs): Set it.
8741         (supersparc_costs): Likewise.
8742         (hypersparc_costs): Likewise.
8743         (leon_cost): Likewise.
8744         (leon3_costs): Likewise.
8745         (sparclet_costs): Likewise.
8746         (ultrasparc_costs): Likewise.
8747         (ultrasparc_costs): Likewise.
8748         (niagara_costs): Likewise.
8749         (niagara2_costs): Likewise.
8750         (niagara3_costs): Likewise.
8751         (niagara4_costs): Likewise.
8752         (niagara7_costs): Likewise.
8753         (m8_costs): Likewise.
8754         (TARGET_CAN_FOLLOW_JUMP): Define.
8755         (pass_work_around_errata::gate): Minor tweak.
8756         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
8757         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
8758         Automaitcally clear MASK_FSMULD mask for V7 processors.
8759         (sparc_can_follow_jump): New static function.
8760         (output_ubranch): Deal with CROSSING_JUMP_P.
8761         (sparc_use_sched_lookahead): Rewrite using switch statement.
8762         (sparc_issue_rate): Reorder.
8763         (sparc_branch_cost): New function.
8765 2018-09-27  Martin Sebor  <msebor@redhat.com>
8767         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
8768         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
8769         (int_fits_type_p): Same.
8771 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
8773         * config/i386/i386.md (FPCR_REG): Remove.
8774         (UNSPEC_FLDCW): Remove.
8775         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
8776         (x86_fldcw_1): Remove insn pattern.
8777         (fnstenv): Do not clobber FPCR_REG.
8778         (fldenv): Ditto.
8779         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
8780         (CALL_USED_REGISTERS): Ditto.
8781         (REG_ALLOC_ORDER): Ditto.
8782         (REG_CLASS_CONTENTS): Ditto.
8783         (HI_REGISTER_NAMES): Ditto.
8784         (ADDITIONAL_REGISTER_NAMES): Use defines instead
8785         of numerical constants.
8786         * config/i386/i386.c (regclass_map): Remove fpsr register.
8787         (dbx_register_map): Ditto.
8788         (dbx64_register_map): Ditto.
8789         (svr4_dbx_register_map): Ditto.
8790         (print_reg): Do not handle FPCR_REG.
8792 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
8794         PR target/87149
8795         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
8796         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
8797         Delete, always treat as true.
8798         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8799         Ditto.  Simplify remaining code.
8800         * config/powerpcspe/powerpcspe.h: Ditto.
8801         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
8802         Simplify remaining code.
8803         (rs6000_expand_builtin): Ditto.
8804         * config/rs6000/rs6000.h: Ditto.
8805         * configure.ac: Ditto.
8806         * configure: Regenerate.
8808 2018-09-27  Martin Liska  <mliska@suse.cz>
8810         * coverage.c (get_coverage_counts): Revert the formatting
8811         of missing profile opt info.
8813 2018-09-27  Richard Biener  <rguenther@suse.de>
8815         PR debug/37801
8816         PR debug/87440
8817         * dwarf2out.c (set_block_origin_self): Do not mark outermost
8818         block as we do not output that.
8819         (gen_inlined_subroutine_die): Elide the originally outermost
8820         block, matching what we do for concrete instances.
8821         (decls_for_scope): Add parameter specifying whether to recurse
8822         to subblocks.
8824 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
8825             Tom de Vries  <tom@codesourcery.com>
8827         PR 82089
8829         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
8830         STORE_FLAG_VALUE == 1.
8832 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
8834         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
8835         constant definitions.
8836         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
8837         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
8838         ("speculation_barrier"): New expander definition.
8840 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
8842         PR gcov-profile/86957
8843         * common.opt: New warning option -Wmissing-profile.
8844         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
8845         * doc/invoke.texi: Document -Wmissing-profile.
8847 2018-09-26  Jim Wilson  <jimw@sifive.com>
8849         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
8850         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
8852 2018-09-26  Martin Sebor  <msebor@redhat.com>
8854         * tree.c (zerop): Change return type to bool.
8855         (integer_zerop, integer_onep, integer_each_onep): Same.
8856         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8857         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8858         (real_onep, real_minus_onep, chain_index): Same.
8859         (print_type_hash_statistics, type_list_equal): Same.
8860         * tree.h (zerop): Same.
8861         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
8862         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8863         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8864         (real_onep, real_minus_onep, chain_index): Same.
8865         (print_type_hash_statistics, type_list_equal): Same.
8867 2018-09-26  Jim Wilson  <jimw@sifive.com>
8869         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
8871 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
8873         PR target/87414
8874         * config/i386/i386.c: Include debug.h and dwarf2out.h.
8875         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
8876         call.
8878 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
8880         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
8882 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8884         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
8885         and FP_SECOND_SSE_REGS.
8886         (REG_CLASS_NAMES): Ditto.
8887         (REG_CLASS_CONTENTS): Ditto.
8888         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
8889         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
8890         (ix86_preferred_output_reload_class): Ditto.
8891         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
8892         clobber constraint to "=&f".
8893         (fix_truncdi_i387): Ditto.
8894         (lrintxfdi2): Ditto.
8895         (fistdi2_<rounding>): Ditto.
8896         (fpremxf4_i387): Change "=u" constraint to "=f".
8897         (fprem1xf4_i387): Ditto.
8898         (sincosxf3): Ditto.
8899         (fptanxf4_i387): Ditto.
8900         (fxtractxf3_i387): Ditto.
8901         (fscalexf4_i387): Ditto.
8902         (atan2xf3): Change "u" constraint to "f".
8903         (fyl2xxf3_i387): Ditto.
8904         (fyl2xp1xf3_i387): Ditto.
8906 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8908         PR target/87439
8909         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
8910         for removed I387_MASK_PM entity.
8913 2018-09-26  Jeff Law  <law@redhat.com>
8914         Revert
8915         2018-09-26  Alexey Neyman  <stilor@att.net>
8917         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8918         headers are no longer pulled in by <isl/val.h>.
8920 2018-09-26  Richard Biener  <rguenther@suse.de>
8922         PR debug/87443
8923         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
8924         or concrete instance DIE to the tree.  Create abstract origin
8925         attributes also for concrete instances.
8927 2018-09-26  Alexey Neyman  <stilor@att.net>
8929         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8930         headers are no longer pulled in by <isl/val.h>.
8932 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
8934         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
8935         Use new helper functions.
8936         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
8937         Use new helper functions.
8938         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
8939         aarch_mm_needs_release): New declarations.
8940         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
8941         aarch_mm_needs_release): New.
8943 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
8945         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
8946         (arm32_output_mi_thunk): Deal with long calls.
8948 2018-09-26  Richard Biener  <rguenther@suse.de>
8950         PR debug/87428
8951         PR debug/87362
8952         * tree-inline.c (expand_call_inline): When the location
8953         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
8954         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
8955         the inserted BLOCK to make inlined_function_outer_scope_p
8956         recognize it.
8957         * dwarf2out.c (add_call_src_coords_attributes): Do not add
8958         coords for reserved locations.
8960 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8962         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
8963         (*call_indirect_nonlocal_sysv<mode>): Ditto.
8964         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
8965         (*sibcall_nonlocal_sysv<mode>): Ditto.
8966         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
8967         (<bd>_<mode>): Ditto.
8968         (<bd>tf_<mode>): Ditto.
8970 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8972         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
8973         control string as a list of templates instead of as C code.
8974         (*altivec_movti): Ditto.
8975         * config/rs6000/darwin.md (movdf_low_di): Ditto.
8977 2018-09-25  Jim Wilson  <jimw@sifive.com>
8979         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
8980         when target symbol is weak.
8982 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8984         PR c/87387
8985         * builtins.c (unterminated_array): Simplify.
8986         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
8987         where pointer arithmetic is safe.
8989 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8991         PR target/86987
8992         * config/rs6000/altivec.md (altivec_vspltb): Use
8993         const_0_to_15_operand instead of u5bit_cint_operand.
8994         (*altivec_vspltb_internal): Ditto.
8995         (altivec_vspltb_direct): Ditto.
8996         (altivec_vsplth): Use const_0_to_7_operand instead of
8997         u5bit_cint_operand.
8998         (*altivec_vsplth_internal): Ditto.
8999         (altivec_vsplth_direct): Ditto.
9000         (altivec_vspltw): Use const_0_to_3_operand instead of
9001         u5bit_cint_operand.
9002         (*altivec_vspltw_internal): Ditto.
9003         (altivec_vspltw_direct): Ditto.
9004         (altivec_vspltsf): Ditto.
9005         (*altivec_vspltsf_internal): Ditto.
9006         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
9007         various splats with the proper size immediate.  Reorder the various
9008         cases by ascending size of immediate, and put all such together.
9010 2018-09-25  Richard Biener  <rguenther@suse.de>
9012         PR debug/83941
9013         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
9014         GC-ification.
9015         (maybe_create_die_with_external_ref): Do not create
9016         DW_TAG_imported_unit here.
9017         (add_abstract_origin_attribute): Handle external BLOCK refs.
9018         (dwarf2out_abstract_function): Simplify LTO case.
9019         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
9020         rather than using maybe_create_die_with_external_ref.
9022 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
9024         PR target/71278
9025         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
9026         (frndintxf2_mask_pm_i387): Ditto.
9027         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
9028         Enable for !flag_trapping_math.
9029         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
9030         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
9031         Change operand 1 predicate to nonimmediate_operand.
9032         (attr "i387_cw"): Remove mask_pm.
9033         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
9034         (enum ix86_entity): Remove I387_MASK_PM.
9035         * config/i386/i386.c (ix86_i387_mode_needed): Do not
9036         handle I387_MASK_PM.
9037         (ix86_mode_needed): Ditto.
9038         (ix86_mode_after): Ditto.
9039         (ix86_mode_entry): Ditto.
9040         (ix86_mode_exit): Ditto.
9041         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
9043 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
9045         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
9046         to_update_switch_stmts to vNULL instead of calling create on them
9047         immediately.
9049 2018-09-25  Richard Biener  <rguenther@suse.de>
9051         PR tree-optimization/87402
9052         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
9053         (visit_phi): Re-instantiate handling of supposed to be VARYING
9054         but non-VARYING backedge value.
9056 2018-09-25  Richard Biener  <rguenther@suse.de>
9058         PR debug/83941
9059         * dwarf2out.c (struct sym_off_pair): New.
9060         (external_die_map): New global.
9061         (lookup_decl_die): When in LTO create DIEs lazily from the
9062         external_die_map.
9063         (lookup_block_die): New function, create DIEs lazily in LTO.
9064         (equate_block_to_die): New function.
9065         (dwarf2out_die_ref_for_decl): During WPA get the association
9066         from the external DIE map.
9067         (dwarf2out_register_external_die): Record mapping into the
9068         external DIE map.
9069         (maybe_create_die_with_external_ref): New function split out from
9070         DIE generation part of old dwarf2out_register_external_die.
9071         (add_abstract_origin_attribute): Do not return the DIE.  When
9072         in LTO reference externals directly.
9073         (dwarf2out_abstract_function): When in LTO ignore calls for
9074         decls with external DIEs (already present abstract instances).
9075         (gen_call_site_die): Adjust.
9076         (add_high_low_attributes): Likewise.
9077         (gen_lexical_block_die): Likewise.
9078         (gen_inlined_subroutine_die): Likewie.
9079         (gen_block_die): Likewise.
9080         (dwarf2out_inline_entry): Likewise.
9081         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
9082         DIEs.
9084 2018-09-25  Martin Liska  <mliska@suse.cz>
9086         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
9087         integers and not by a float value.
9089 2018-09-25  Martin Liska  <mliska@suse.cz>
9091         PR fortran/87394
9092         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
9093         instead of NULL.
9094         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
9095         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
9096         * godump.c (go_format_type): Remove extra parenthesis.
9098 2018-09-25  Martin Liska  <mliska@suse.cz>
9100         * alias.c (set_dest_equal_p): Remove unused function.
9101         * config/i386/i386.c (def_builtin_pure2): Likewise.
9102         * diagnostic-show-locus.c (class layout): Remove
9103         unused field.
9104         (layout::layout): Likewise here.
9105         * dump-context.h (class temp_dump_context): Likewise.
9106         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
9107         (add_AT_loclistsptr): Likewise.
9108         (add_AT_offset): Likewise.
9109         (get_AT_hi_pc): Likewise.
9110         (is_comdat_die): Likewise.
9111         (type_is_enum): Likewise.
9112         (ceiling): Likewise.
9113         (add_AT_vms_delta): Likewise.
9114         (is_class_die): Likewise.
9115         * edit-context.c (class line_event): Remove unused field.
9116         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
9117         unused function.
9118         * ipa-cp.c (ipa_get_vr_lat): Likewise.
9119         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
9120         (ok_for_base_p_nonstrict): Likewise.
9121         * tree-chrec.c (is_not_constant_evolution): Likewise.
9122         (chrec_fold_poly_cst): Likewise.
9123         * tree-if-conv.c (has_pred_critical_p): Likewise.
9124         * tree-ssa-coalesce.c (print_exprs): Likewise.
9125         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
9126         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
9127         * tree-vrp.c (value_ranges_intersect_p): Likewise.
9128         (value_range_nonnegative_p): Likewise.
9130 2018-09-25  Martin Liska  <mliska@suse.cz>
9132         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
9133         Do not handle "GNU Pascal".
9134         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
9135         Likewise.
9136         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
9137         from documentation. Likewise.
9138         * dbxout.c (dbxout_range_type): Likewise.
9139         * doc/cpp.texi: Likewise.
9140         * doc/extend.texi: Likewise.
9141         * doc/frontends.texi: Likewise.
9142         * doc/invoke.texi: Remove Pascal entry.
9143         * tree.def (CLEANUP_POINT_EXPR): Likewise.
9144         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
9146 2018-09-25  Martin Liska  <mliska@suse.cz>
9148         PR middle-end/86078
9149         * doc/invoke.texi: Document all parameters and remove default
9150         of the parameters.
9152 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
9154         PR bootstrap/87417
9155         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
9156         contains HOST_WIDE_INTs when computing its size.
9158 2018-09-24  Jim Wilson  <jimw@sifive.com>
9160         PR target/87391
9161         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
9162         not TARGET_RVE.
9163         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
9165 2018-09-24  Andrew Pinski  <apinski@marvell.com>
9167         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
9168         access prev before checking it for NULLness in the
9169         AARCH64_FUSE_CMP_BRANCH case.
9171 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
9173         PR target/82699
9174         * config/i386/i386.c (rest_of_insert_endbranch): Set
9175         endbr_queued_at_entrance to true and don't insert ENDBR if
9176         x86_function_profiler will be called.
9177         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
9178         is true.
9179         * config/i386/i386.h (machine_function): Add
9180         endbr_queued_at_entrance.
9182 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
9184         * genattrtab.c (mk_attr_alt): Use alternative_mask.
9185         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
9186         types.
9187         (check_attr_test): Use alternative_mask.
9188         (get_attr_value): Likewise.
9189         (compute_alternative_mask): Use alternative_mask and XWINT.
9190         (make_alternative_compare): Use alternative_mask.
9191         (attr_alt_subset_p): Use XWINT.
9192         (attr_alt_subset_of_compl_p): Likewise.
9193         (attr_alt_intersection): Use alternative_mask and XWINT.
9194         (attr_alt_union): Likewise.
9195         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
9196         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
9197         (simplify_test_exp): Use alternative_mask and XWINT.
9198         (write_test_expr): Use alternative_mask and XWINT, adjust bit
9199         number calculation to support 64 bits.  Generate code that
9200         checks 64-bit masks.
9201         (main): Use alternative_mask.
9202         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
9204 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
9206         PR target/80080
9207         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
9208         RETURN+USE when returning via %r14.
9210 2018-09-24  Martin Liska  <mliska@suse.cz>
9212         * gcov.c (output_lines): Print colorization legend
9213         for both flag_use_colors and flag_use_hotness_colors.
9214         Reword the help.
9216 2018-09-24  Martin Liska  <mliska@suse.cz>
9218         * coverage.c (get_coverage_counts): Use warning_at
9219         with current_function_decl location. Use %qD in warning
9220         message.
9222 2018-09-24  Martin Liska  <mliska@suse.cz>
9224         * memory-block.h (memory_block_pool::release): Annotate with
9225         valgrind that the memory is not accessible.
9227 2018-09-24  Martin Liska  <mliska@suse.cz>
9229         PR sanitizer/85774
9230         * asan.c: Make asan_handled_variables extern.
9231         * asan.h: Likewise.
9232         * cfgexpand.c (expand_stack_vars): Make sure
9233         a representative is unpoison if another
9234         variable in the partition is handled by
9235         use-after-scope sanitization.
9237 2018-09-24  Richard Biener  <rguenther@suse.de>
9239         PR tree-optimization/63155
9240         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
9241         the worklist when the edge of the respective argument isn't
9242         executable.
9244 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
9246         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
9247         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
9248         (MASK_CLASS_P): Update for rename.
9249         (MAYBE_MASK_CLASS_P): Ditto.
9250         (REG_CLASS_NAMES): Update.
9251         (REG_CLASS_CONTENT): Update.
9252         * config/i386/i386.c (regclass_map): Update for MASK_REG
9253         and ALL_MASK_REGS rename.
9254         * config/i386/constraints.md (Yk): Update for rename.
9255         (k): Ditto.
9257 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
9259         * config/i386/i386.h (enum reg_class): Remove
9260         EVEX_SSE_REGS and MOD4_SSE_REGS.
9261         (REG_CLASS_NAMES): Update.
9262         (REG_CLASS_CONTENT): Update.
9263         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
9264         registers as ALL_SSE_REGS.
9265         (ix86_additional_allocno_class_p): Remove.
9266         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
9267         (ix86_register_priority): Lower priority of EVEX SSE registers.
9268         Use IN_RANGE macro where appropriate.
9269         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
9270         AVX-5124VNNIW checks.
9271         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
9272         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
9273         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
9274         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
9275         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
9276         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
9277         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
9278         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
9279         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
9280         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
9281         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
9282         * config/i386/constraints.md (Yh): Remove.
9284 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
9286         * config/i386/i386.c (regclass_map): Declare integer REX registers
9287         as GENERAL_REGS.
9289 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
9291         * doc/service.texi (Service): Switch the fsf.org link to https.
9293 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
9295         PR target/86798
9296         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9297         Define to speculation_safe_value_not_needed.
9299 2018-09-21  Florian Weimer  <fweimer@redhat.com>
9301         PR middle-end/81035
9302         * doc/extend.texi (Common Function Attributes): Mention that
9303         noreturn suppresses tail call optimization.
9305 2018-09-21  Jeff Law  <law@redhat.com>
9307         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
9308         vr_values::cleanup_edges_and_switches.
9309         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
9310         vr_values class.
9311         (identify_jump_threads): Remove EDGE_IGNORE handling.
9312         (execute_vrp): Move handling of to_remove_edges and
9313         to_update_switch_stmts into vr_values class member functions.
9314         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
9315         (to_update_switch_stmts): Likewise.
9316         * vr-values.c: Include cfghooks.h.
9317         (vr_values::vr_values): Initialize to_remove_edges and
9318         to_update_switch_stmts.
9319         (vr_values::~vr_values): Verify to_remove_edges and
9320         to_update_switch_stmts are empty.
9321         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
9322         (vr_values::cleanup_edges_and_switches): New member function.
9323         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
9324         function.  Add new data members.
9326 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
9328         PR tree-optimization/87309
9329         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
9330         calls with pflags and alt_flags.
9331         (selftest::test_capture_of_dump_calls): Add test of interaction of
9332         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
9334 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9336         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
9338 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9340         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
9342 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9344         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
9345         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
9346         (VXWORKS_LIBS_RTP): Minor reordering.
9348 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9350         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
9351         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
9353 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9355         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
9356         (PTRDIFF_TYPE): Likewise.
9358 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9360         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
9361         triplet, similar to support for VxWorks7.
9362         * config/vxworks-dummy.h: Provide a default definition
9363         of TARGET_VXWORKS64 to 0.
9365 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9367         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
9368         * config/vxworks-dummy.h: here.
9370 2018-09-21  Olivier Hainque  <hainque@adacore.com>
9372         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
9374 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
9375             Bo Zhou  <zbo.zhou@hisilicon.com>
9377         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
9378         * config/aarch64/aarch64-tune.md: Regenerated.
9379         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
9380         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
9381         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
9383 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
9384             Julian Brown  <julian@codesourcery.com>
9386         * builtins.c (get_builtin_sync_mem): Handle address spaces.
9388 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
9390         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
9391         if the call takes a static chain.
9393 2018-09-21  Martin Liska  <mliska@suse.cz>
9395         * auto-profile.c (autofdo_source_profile::read): Do not
9396         set sum_all.
9397         (read_profile): Do not add working sets.
9398         (read_autofdo_file): Remove sum_all.
9399         (afdo_callsite_hot_enough_for_early_inline): Remove const
9400         qualifier.
9401         * coverage.c (struct counts_entry): Remove gcov_summary.
9402         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
9403         do not support GCOV_TAG_PROGRAM_SUMMARY.
9404         (get_coverage_counts): Remove summary and expected
9405         arguments.
9406         * coverage.h (get_coverage_counts): Likewise.
9407         * doc/gcov-dump.texi: Remove -w option.
9408         * gcov-dump.c (dump_working_sets): Remove.
9409         (main): Do not support '-w' option.
9410         (print_usage): Likewise.
9411         (tag_summary): Likewise.
9412         * gcov-io.c (gcov_write_summary): Do not dump
9413         histogram.
9414         (gcov_read_summary): Likewise.
9415         (gcov_histo_index): Remove.
9416         (gcov_histogram_merge): Likewise.
9417         (compute_working_sets): Likewise.
9418         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
9419         it not obsolete.
9420         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
9421         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
9422         (GCOV_HISTOGRAM_SIZE): Remove.
9423         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
9424         (struct gcov_summary): Simplify rapidly just
9425         to runs and sum_max fields.
9426         (gcov_histo_index): Remove.
9427         (NUM_GCOV_WORKING_SETS): Likewise.
9428         (compute_working_sets): Likewise.
9429         * gcov-tool.c (print_overlap_usage_message): Remove
9430         trailing empty line.
9431         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
9432         (output_lines): Remove program related line.
9433         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
9434         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
9435         histogram.
9436         (input_profile_summary): Do not read it.
9437         (merge_profile_summaries): And do not merge it.
9438         (input_symtab): Do not call removed function.
9439         * modulo-sched.c (sms_schedule): Do not print sum_max.
9440         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
9441         removed when histogram method was invented.
9442         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
9443         mode.
9444         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
9445         GCOV coding style.
9446         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
9447         and dump selected value.
9448         * profile.c (add_working_set): Remove.
9449         (get_working_sets): Likewise.
9450         (find_working_set): Likewise.
9451         (get_exec_counts): Do not work with working sets.
9452         (read_profile_edge_counts): Do not inform as sum_max is removed.
9453         (compute_branch_probabilities): Likewise.
9454         (compute_value_histograms): Remove argument for call of
9455         get_coverage_counts.
9456         * profile.h: Do not make gcov_summary const.
9458 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
9460         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
9462 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
9464         PR tree-optimization/86990
9465         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
9466         Check that the entire merged store group is made of constants only for
9467         overlapping stores.
9469 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
9471         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
9472         (VTABLE_VERIFICATION_SPEC): Likewise.
9473         (SANITIZER_EARLY_SPEC): Likewise.
9474         (SANITIZER_SPEC): Likewise.
9475         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
9476         * doc/invoke.texi (Link Options): Document -r.
9478 2018-09-20  Richard Biener <rguenther@suse.de>
9480         PR middle-end/87054
9481         * gimplify.c (gimplify_expr): Retain alignment of
9482         addressable lvalue in dereference.
9484 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
9486         PR bootstrap/87013
9487         * configure.ac: Check for .loc is_stmt support.
9488         * configure, config.in: Rebuilt.
9489         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
9490         if not supported.
9492 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
9494         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
9495         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
9496         -misel=no.
9498 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
9500         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9501         VECTOR_OTHER.
9502         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
9503         case VECTOR_OTHER.
9505 2018-09-20  Marek Polacek  <polacek@redhat.com>
9507         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
9509 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
9511         PR tree-optimization/87288
9512         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
9513         into account when determining PEELING_FOR_NITERS.
9515 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
9517         PR tree-optimization/86877
9518         * tree-vect-loop.c (vect_analyze_loop_2): Call
9519         vect_verify_datarefs_alignment.
9521 2018-09-19  Marek Polacek  <polacek@redhat.com>
9523         * doc/invoke.texi: Document -Wclass-conversion.
9525 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
9527         * config/pa/pa.c (pa_adjust_priority): Delete.
9528         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
9530         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
9531         (atomic_storehi): Likewise.
9532         (atomic_storesi): Likewise.
9533         (atomic_loaddi): Restore compare and swap exchange loop code.
9535 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
9537         PR rtl-optimization/86902
9538         * combine.c (try_combine): When changing the CC mode used, don't change
9539         an unrelated mode in other_insn to that new CC mode.
9541 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
9543         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
9544         with %T in place of calls to dump_generic_expr.
9545         (prune_runtime_alias_test_list): Likewise.
9546         (create_runtime_alias_checks): Likewise.
9547         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
9548         (vect_analyze_data_ref_dependence): Likewise.
9549         (vect_slp_analyze_data_ref_dependence): Likewise.
9550         (vect_record_base_alignment): Likewise.  Use %G in place of call
9551         to dump_gimple_stmt.
9552         (vect_compute_data_ref_alignment): Likewise.
9553         (verify_data_ref_alignment): Likewise.
9554         (vect_find_same_alignment_drs): Likewise.
9555         (vect_analyze_group_access_1): Likewise.
9556         (vect_analyze_data_ref_accesses): Likewise.
9557         (dependence_distance_ge_vf): Likewise.
9558         (dump_lower_bound): Likewise.
9559         (vect_prune_runtime_alias_test_list): Likewise.
9560         (vect_find_stmt_data_reference): Likewise.
9561         (vect_analyze_data_refs): Likewise.
9562         (vect_create_addr_base_for_vector_ref): Likewise.
9563         (vect_create_data_ref_ptr): Likewise.
9564         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
9565         (vect_can_advance_ivs_p): Likewise.
9566         (vect_update_ivs_after_vectorizer): Likewise.
9567         (vect_gen_prolog_loop_niters): Likewise.
9568         (vect_prepare_for_masked_peels): Likewise.
9569         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9570         (vect_determine_vectorization_factor): Likewise.
9571         (vect_is_simple_iv_evolution): Likewise.
9572         (vect_analyze_scalar_cycles_1): Likewise.
9573         (vect_analyze_loop_operations): Likewise.
9574         (report_vect_op): Likewise.
9575         (vect_is_slp_reduction): Likewise.
9576         (check_reduction_path): Likewise.
9577         (vect_is_simple_reduction): Likewise.
9578         (vect_create_epilog_for_reduction): Likewise.
9579         (vect_finalize_reduction:): Likewise.
9580         (vectorizable_induction): Likewise.
9581         (vect_transform_loop_stmt): Likewise.
9582         (vect_transform_loop): Likewise.
9583         (optimize_mask_stores): Likewise.
9584         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
9585         (vect_split_statement): Likewise.
9586         (vect_recog_over_widening_pattern): Likewise.
9587         (vect_recog_average_pattern): Likewise.
9588         (vect_determine_min_output_precision_1): Likewise.
9589         (vect_determine_precisions_from_range): Likewise.
9590         (vect_determine_precisions_from_users): Likewise.
9591         (vect_mark_pattern_stmts): Likewise.
9592         (vect_pattern_recog_1): Likewise.
9593         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9594         (vect_record_max_nunits): Likewise.
9595         (vect_build_slp_tree_1): Likewise.
9596         (vect_build_slp_tree_2): Likewise.
9597         (vect_print_slp_tree): Likewise.
9598         (vect_analyze_slp_instance): Likewise.
9599         (vect_detect_hybrid_slp_stmts): Likewise.
9600         (vect_detect_hybrid_slp_1): Likewise.
9601         (vect_slp_analyze_operations): Likewise.
9602         (vect_slp_analyze_bb_1): Likewise.
9603         (vect_transform_slp_perm_load): Likewise.
9604         (vect_schedule_slp_instance): Likewise.
9605         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9606         (vect_mark_stmts_to_be_vectorized): Likewise.
9607         (vect_init_vector_1): Likewise.
9608         (vect_get_vec_def_for_operand): Likewise.
9609         (vect_finish_stmt_generation_1): Likewise.
9610         (vect_check_load_store_mask): Likewise.
9611         (vectorizable_call): Likewise.
9612         (vectorizable_conversion): Likewise.
9613         (vectorizable_operation): Likewise.
9614         (vectorizable_load): Likewise.
9615         (vect_analyze_stmt): Likewise.
9616         (vect_is_simple_use): Likewise.
9617         (vect_get_vector_types_for_stmt): Likewise.
9618         (vect_get_mask_type_for_stmt): Likewise.
9619         * tree-vectorizer.c (increase_alignment): Likewise.
9621 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
9623         * doc/rtl.texi: Adjust vec_select description.
9624         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
9625         non-constant selectors.
9627 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
9629         * config/aarch64/aarch64-protos.h
9630         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
9631         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
9632         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
9633         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
9634         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
9635         (AARCH64_FL_PROFILE): Move index so flags are ordered.
9636         (AARCH64_ISA_RCPC8_4): New flag.
9637         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
9638         to aarch64_offset_9bit_signed_unscaled_p.
9639         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
9640         and use stlur.
9641         * config/aarch64/constraints.md (Ust): New constraint.
9642         * config/aarch64/predicates.md.
9643         (aarch64_9bit_offset_memory_operand): New predicate.
9644         (aarch64_rcpc_memory_operand): New predicate.
9646 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
9648         PR rtl-optimization/87361
9649         * rtlanal.c (nonzero_bits1): Revert accidental change.
9651 2018-09-19  Richard Biener  <rguenther@suse.de>
9653         PR tree-optimization/87349
9654         PR tree-optimization/87342
9655         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
9657 2018-09-18  Marek Polacek  <polacek@redhat.com>
9659         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
9660         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
9662 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
9664         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
9666 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
9668         PR rtl-optimization/86882
9669         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
9671 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
9673         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
9674         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
9676 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
9678         PR other/87353
9679         * doc/invoke.texi (Link Options): Fix formatting and grammar.
9681 2018-09-18  Richard Biener  <rguenther@suse.de>
9683         PR middle-end/63155
9684         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
9685         (compute_samebase_partition_bases): Likewise.
9686         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
9687         (gimple_can_coalesce_p): Simplify.
9689 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
9691         Handle a library implementation of ffs calling __builtin_ffs.
9692         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
9693         (mmix_init_libfuncs): New function: make __builtin_ffs expand
9694         to __ffsdi2.
9696 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9698         * diagnostic-show-locus.c (class layout_range): Add field
9699         "m_original_idx".
9700         (layout_range::layout_range): Add "original_idx" param and use it
9701         to initialize new field.
9702         (make_range): Use 0 for original_idx.
9703         (layout::layout): Pass in index to calls to
9704         maybe_add_location_range.
9705         (layout::maybe_add_location_range): Add param "original_idx" and
9706         pass it on to layout_range.
9707         (layout::print_any_labels): Pass on range->m_original_idx to
9708         get_text call.
9709         (gcc_rich_location::add_location_if_nearby): Use 0 for
9710         original_idx.
9711         * gcc-rich-location.h (text_range_label::get_text): Update for new
9712         param.
9713         (range_label_for_type_mismatch::get_text): Likewise.
9715 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9717         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
9719 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9721         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
9722         format_string_diagnostic_t.
9723         (fmtwarn_n): Likewise.
9724         * substring-locations.c
9725         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
9726         (format_warning_n_va): Convert to...
9727         (format_string_diagnostic_t::emit_warning_n_va): ...this.
9728         (format_warning_va): Convert to...
9729         (format_string_diagnostic_t::emit_warning_va): ...this.
9730         (format_warning_at_substring): Convert to...
9731         (format_string_diagnostic_t::emit_warning): ...this.
9732         (format_warning_at_substring_n): Convert to...
9733         (format_string_diagnostic_t::emit_warning_n): ...this.
9734         * substring-locations.h (class format_string_diagnostic_t): New
9735         class.
9736         (format_warning_va): Convert to
9737         format_string_diagnostic_t::emit_warning_va.
9738         (format_warning_n_va): Convert to
9739         format_string_diagnostic_t::emit_warning_n_va.
9740         (format_warning_at_substring): Convert to
9741         format_string_diagnostic_t::emit_warning.
9742         (format_warning_at_substring_n): Convert to
9743         format_string_diagnostic_t::emit_warning_n.
9745 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
9746             Bernd Schmidt <bernds_cb1@t-online.de>
9748         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
9749         SImode args.
9751 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9753         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
9754         operand 0 predicate to nonimmediate operand.
9755         (rint<mode>2_frndint): Remove insn pattern.
9756         (rint<mode>2): Change operand 1 predicate to general_operand.
9757         Extend operand 1 to XFmode and generate rintxf2 insn.
9758         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
9759         Do not use X87MODEF mode macro.
9760         (frndintxf2_<rounding>_i387): Rename from
9761         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
9762         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
9763         to XFmode and generate significandxf3 insn.
9765 2018-09-17  Richard Biener  <rguenther@suse.de>
9767         PR tree-optimization/87328
9768         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
9769         visiting unexecutable backedges when not iterating.
9770         (do_rpo_vn): Mark all edges not executable even when not
9771         iterating.
9773 2018-09-17  Martin Jambor  <mjambor@suse.cz>
9775         PR c/63886
9776         * doc/invoke.texi (Warning Options): Likewise.
9778 2018-09-17  Richard Biener  <rguenther@suse.de>
9780         PR tree-optimization/87301
9781         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
9782         clean EH info from leftover copy assignments.
9784 2018-09-17  Martin Liska  <mliska@suse.cz>
9786         PR gcov-profile/85871
9787         * gcov.c (output_intermediate_file): Fix out of bounds
9788         access.
9790 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
9792         * config/arc/arc.c: Object attributes for core4 not reflected
9793         correctly.
9794         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
9795         core3).
9797 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
9799         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
9801 2018-09-17  Martin Liska  <mliska@suse.cz>
9803         * doc/gcov.texi: Document new option --use-hotness-colors.
9804         * gcov.c (struct source_info): Declare new field.
9805         (source_info::source_info): Set default for maximum_count.
9806         (print_usage): Add new -q option.
9807         (process_args): Process it.
9808         (accumulate_line_info): Save src->maximum_count.
9809         (output_line_beginning): Make color line number if
9810         flag_use_hotness_colors is set.
9811         (output_line_details): Pass default argument value.
9812         (output_lines): Pass src->maximum_count.
9814 2018-09-17  Martin Liska  <mliska@suse.cz>
9816         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
9817         Use processor_names table.
9818         * config/i386/i386.c (ix86_default_align): Use
9819         processor_cost_table for alignment values.
9820         (ix86_option_override_internal): Use processor_names.
9821         (ix86_function_specific_print): Likewise.
9822         * config/i386/i386.h (struct processor_costs):
9823         Add alignment values.
9824         (struct ptt): Remove and replace with const char *.
9825         * config/i386/x86-tune-costs.h (struct processor_costs):
9826         Declare default alignments for all costs.
9828 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
9830         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
9831         symbolics or VR_VARYING ranges for ABS_EXPR.
9832         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
9833         when range will wrap.
9835 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
9837         PR middle-end/86864
9838         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
9839         before and after a JUMP_TABLE_DATA.
9841 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
9843         PR middle-end/87188
9844         * dojump.c (do_compare_and_jump): Canonicalize function pointers
9845         when one operand is a function pointer.  Use POINTER_TYPE_P and
9846         FUNC_OR_METHOD_TYPE_P.
9847         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
9848         * fold-const.c (build_range_check): Likewise.
9849         * match.pd (simple_comparison): Likewise.
9851 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
9853         PR c/82967
9854         * spellcheck.c (get_edit_distance_cutoff): New function.
9855         (selftest::test_edit_distance_unit_test_oneway): Rename to...
9856         (selftest::test_get_edit_distance_one_way): ...this.
9857         (selftest::test_get_edit_distance_unit): Rename to...
9858         (selftest::test_get_edit_distance_both_ways): ...this.
9859         (selftest::test_edit_distances): Move tests to this new function,
9860         and test some more pairs of strings.  Update for above renaming.
9861         (selftest::get_old_cutoff): New function.
9862         (selftest::test_get_edit_distance_cutoff): New function.
9863         (selftest::assert_suggested_for): New function.
9864         (ASSERT_SUGGESTED_FOR): New macro.
9865         (selftest::assert_not_suggested_for): New function.
9866         (ASSERT_NOT_SUGGESTED_FOR): New macro.
9867         (selftest::test_suggestions): New function.
9868         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
9869         tests to selftest::test_edit_distances and call it.  Add calls to
9870         selftest::test_get_edit_distance_cutoff and
9871         selftest::test_suggestions.
9872         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
9873         (best_match::consider): Replace hard-coded cutoff calculation with
9874         a call to...
9875         (best_match::get_cutoff): New declaration.
9876         (best_match::get_best_meaningful_candidate): Likewise.
9878 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9880         * builtins.c (fold_builtin_strlen): Remove TODO comment.
9882 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9884         revert:
9885         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9887         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9888         terminated string literal.
9890 2018-09-14  Martin Sebor  <msebor@redhat.com>
9892         * builtins.c (unterminated_array): Handle ARRAY_REF.
9893         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
9894         * builtins.h (unterminated_array): Declare extern.
9895         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
9896         arrays.
9897         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
9898         calls.
9900 2018-09-14  Martin Sebor  <msebor@redhat.com>
9901             Jeff Law  <law@redhat.com>
9903         * builtins.c (unterminated_array): New.
9904         (expand_builtin_strcpy): Adjust.
9905         (expand_builtin_strcpy_args): Detect unterminated arrays.
9906         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
9907         unterminated arrays.
9908         * gimple-fold.h (get_maxval_strlen): Add argument.
9909         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
9911         * gimple-fold.c (get_range_strlen): Add argument.
9912         (get_maxval_strlen): Adjust.
9913         * gimple-fold.h (get_range_strlen): Add argument.
9915 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
9917         * config/i386/movdirintrin.h: Fix copyright year.
9919 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
9921         * reg-stack.c: Include regs.h.
9922         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
9923         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
9924         FIRST_STACK_REG, not DFmode.
9925         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
9926         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
9927         (change stack): Default register mode to the reg_raw_mode of
9928         FIRST_STACK_REG, not DFmode.
9929         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
9930         (*swapxf): Rename from swapxf.
9932 2018-09-14  Carl Love  <cel@us.ibm.com>
9934         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
9935         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
9937 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
9939         PR target/87224
9940         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
9941         alternatives.
9943 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
9945         PR target/85628
9946         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
9948 2018-09-14  Jason Merrill  <jason@redhat.com>
9950         Fix --enable-gather-detailed-mem-stats.
9951         * hash-table.c (hash_table_usage): Change from variable to function.
9952         * hash-table.h: Adjust.
9953         * Makefile.in: Add missing dependencies on hash-table.h.
9955 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9957         PR tree-optimization/87259
9958         PR lto/87283
9959         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
9960         execute_cse_reciprocals_1 has tried transforming.
9962 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
9964         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
9965         VR_VARYING for PLUS/MINUS_EXPR.
9967 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9969         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
9970         formatting.
9972 2018-09-14  Richard Biener  <rguenther@suse.de>
9974         PR middle-end/63155
9975         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
9976         bits for the merged partition.
9978 2018-09-13  Martin Sebor  <msebor@redhat.com>
9979             Bernd Edlinger  <bernd.edlinger@hotmail.de>
9981         * builtins.h (c_srlen): Add argument.
9982         * builtins.c (warn_string_no_nul): New function.
9983         (c_strlen): Add argument and use it.  Update recursive calls.
9984         Pass DECL argument to string_constant to get info on non
9985         terminated strings.  Update *NONSTR as needed.
9986         (fold_builtin_strlen): Add argument to calls to c_strlen.
9987         Warn for unterminated arrays.
9988         (warn_string_no_null): Add prototype.
9989         * expr.c (string_constant): Update arguments.  Update recursive
9990         calls appropriately.  Detect missing NUL terminator and outermost
9991         declaration its missing in.
9992         Improve checks for arrays with nonzero lower bound or elements
9993         that are not a single byte.  Simplify offset computation.
9994         Simplify checks for non-NUL terminated strings.
9995         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
9996         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
9998 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10000         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
10001         correctly.
10002         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
10003         argument.  Fix range checks.
10004         * fold-const.h (c_getstr): Adjust protoype.
10005         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
10006         string is constant but contains no NUL byte.
10008         * expr.c (string_constant): Adjust function comment.
10009         Remove bogus check for zero termination.
10011         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
10013         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
10014         (get_constant_size): Don't make STRING_CSTs larger than they are.
10015         (check_string_literal): New check function for STRING_CSTs.
10016         (output_constant): Use it.
10018 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
10020         PR target/86812
10021         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
10023 2018-09-13  Richard Biener  <rguenther@suse.de>
10025         PR tree-optimization/87263
10026         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
10027         (struct unwind_state): Add max_rpo field.
10028         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
10029         Compute max_rpo, the max RPO number a block can be backwards reached
10030         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
10031         separating it from the iterating mode.
10033 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
10035         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
10036         (rfs_decision): New scheduling decision.
10038 2018-09-13  Richard Biener  <rguenther@suse.de>
10040         PR bootstrap/87134
10041         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
10042         (vn_nary_op_insert_pieces_predicated): Do not write useless
10043         valid_dominated_by_p entry outside of the allocated storage.
10045 2018-09-13  Omar Sandoval  <osandov@osandov.com>
10046             Tom de Vries  <tdevries@suse.de>
10048         PR debug/86985
10049         * dwarf2out.c (is_c): New function.
10050         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
10052 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
10054         PR target/85628
10055         * config/aarch64/aarch64.md (*aarch64_bfxil):
10056         Define.
10057         * config/aarch64/constraints.md (Ulc): Define.
10058         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
10059         Define.
10060         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
10061         New function.
10063 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
10065         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
10066         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
10067         aarch64_layout_frame call.
10068         (aarch64_expand_epilogue): Likewise.
10069         (aarch64_initial_elimination_offset): Likewise.
10070         (aarch64_get_separate_components): Likewise.
10071         (aarch64_use_return_insn_p): Likewise.
10072         (aarch64_layout_frame): Remove unneeded check.
10074 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10076         * configure.ac: Only append
10077         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
10078         gcc_config_arguments if it was never reconfigured or last reconfigure
10079         was with different arguments.
10080         * configure: Regenerated.
10082 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10083             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10085         PR middle-end/87290
10086         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
10087         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
10089 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
10091         PR tree-optimization/87287
10092         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
10093         X % C == 0 to X % (unsigned) C == 0 optimization to ...
10094         * match.pd (X % C == 0): ... here.  New optimization.
10096 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10098         PR middle-end/82853
10099         * expr.h (maybe_optimize_mod_cmp): Declare.
10100         * expr.c (mod_inv): New function.
10101         (maybe_optimize_mod_cmp): New function.
10102         (do_store_flag): Use it.
10103         * cfgexpand.c (expand_gimple_cond): Likewise.
10105 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
10106             Julian Brown  <julian@codesourcery.com>
10108         PR middle-end/86336
10109         * gimplify.c (gimplify_scan_omp_clauses): Set
10110         target_firstprivatize_array_bases in OpenACC parallel and kernels
10111         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
10112         OpenACC data regions.
10114 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
10116         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
10117         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
10119 2018-09-12  Richard Biener  <rguenther@suse.de>
10121         PR tree-optimization/87280
10122         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
10123         edge but unreachable target.
10124         (do_rpo_vn): For conservatively handling a PHI only mark
10125         the backedge executable but not the block reachable.
10127 2018-09-12  Richard Biener  <rguenther@suse.de>
10129         PR tree-optimization/87266
10130         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
10131         visited blocks.
10133 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
10135         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
10136         constants.
10137         ("trunc<BFP:mode><DFP_ALL:mode>2")
10138         ("trunc<DFP_ALL:mode><BFP:mode>2")
10139         ("extend<BFP:mode><DFP_ALL:mode>2")
10140         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
10141         according to the target operand type.
10143 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10144             Andreas Krebbel  <krebbel@linux.ibm.com>
10146         PR tree-optimization/86844
10147         * gimple-ssa-store-merging.c
10148         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
10149         there are any overlapping stores in between them, make sure they are
10150         also coalesced or we give up completely.
10152 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
10154         PR middle-end/87248
10155         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
10156         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
10158 2018-09-12  Tom de Vries  <tdevries@suse.de>
10160         * common.opt (gdescribe-dies): Add option.
10161         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
10162         attribute for artifical and nameless decls.
10163         (dwarf2out_register_external_die): Add description attribute to
10164         external reference die.
10165         (add_desc_attribute): New functions.
10166         (gen_subprogram_die): Add description attribute to
10167         DW_TAG_call_site_parameter.
10168         * tree-pretty-print.c (print_generic_expr_to_str): New function.
10169         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
10170         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
10171         -gno-describe-dies.
10172         (@item -gdescribe-dies): Add.
10174 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
10176         * tree-vrp.c (vrp_shift_undefined_p): Remove.
10177         (extract_range_from_binary_expr_1: Call
10178         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
10179         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
10180         depend on sign.
10182 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
10184         * gimple-ssa-warn-alloca.c
10185         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
10186         field for ALLOCA_BOUND_*_LARGE.
10188 2018-09-11  Nathan Sidwell  <nathan@acm.org>
10190         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
10192 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
10194         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
10195         for clobbers.  Remove obsolete comment.
10197 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
10199         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
10200         mpxchk, mpxld and mpxst types.
10201         (define_attr length_immediate): Remove all processing of mpx types.
10202         (define_attr prefix_0f): Ditto.
10203         (define_attr memory): Ditto.
10205 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
10207         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
10208         (log<mode>2): Change operand 1 predicate to general_operand.
10209         Extend operand 1 to XFmode and generate logxf3 insn.
10210         (log10<mode>2): Change operand 1 predicate to general_operand.
10211         Extend operand 1 to XFmode and generate log10xf3 insn.
10212         (log2<mode>2): Change operand 1 predicate to general_operand.
10213         Extend operand 1 to XFmode and generate log2xf3 insn.
10214         (fyl2xp1_extend<mode>xf3_i387): Remove.
10215         (log1p<mode>2): Change operand 1 predicate to general_operand.
10216         Extend operand 1 to XFmode and generate log1pxf3 insn.
10217         (fxtract_extend<mode>xf3_i387): Remove.
10218         (logb<mode>2): Change operand 1 predicate to general_operand.
10219         Extend operand 1 to XFmode and generate logbxf3 insn.
10220         (ilogb<mode>2): Change operand 1 predicate to general_operand.
10221         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
10222         (significand<mode>2): Change operand 1 predicate to general_operand.
10223         Extend operand 1 to XFmode and generate significandxf3 insn.
10225 2018-09-11  Nathan Sidwell  <nathan@acm.org>
10227         * gcc.c (perror_with_name, pfatal_with_name): Delete.
10228         (load_specs): Use fatal_error.
10229         (DELETE_IF_ORDINARY, process_command): Use error.
10230         (execute, run_attempt): Use fatal_error.
10232 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
10234         * diagnostic-core.h (sorry_at): New prototype.
10235         * diagnostic.c (sorry_at): New function.
10237 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
10239         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
10240         by zero as VR_UNDEFINED.
10242 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
10244         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
10245         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
10246         (<sincos>mode2): New expander.
10247         (sincos_extend<mode>xf3_i387): Remove insn pattern.
10248         (sincos -> sin, cos splitters): Remove splitter patterns.
10249         (sincos<mode>3): Change operand 2 predicate to general_operand.
10250         Extend operand 2 to XFmode and generate sincosxf3 insn.
10251         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
10252         Change operand 3 predicate to const1_operand.
10253         (fptan_extend<mode>xf4_i387): Remove insn pattern.
10254         (tanxf2): Update operands in the call to fptanxf4_i387.
10255         (tan<mode>2): Change operand 1 predicate to general_operand.
10256         Extend operand 1 to XFmode and generate tanxf3 insn.
10257         (atan2xf3): Rename from *fpatanxf3_i387.
10258         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
10259         (atan2xf3): Remove expander.
10260         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
10261         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
10262         (atan<mode>2): Change operand 1 predicate to general_operand.
10263         Extend operand 1 to XFmode and generate atanxf3 insn.
10265 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
10267         * config/i386/i386.md (x87/SSE constant load splitter): Use
10268         memory_operand instead of nonimmediate_operand for input operand
10269         predicate.
10271 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
10273         * config/i386/i386.md (float partial SSE register stall splitter): Move
10274         splitter near its instruction pattern.
10275         (float_extend partial SSE register stall splitter): Ditto.
10276         (float_truncate partial SSE register stall splitter): Ditto.
10278 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
10280         PR target/86794
10281         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10282         to speculation_safe_value_not_needed.
10284         PR target/85666
10285         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
10286         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
10287         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
10288         leaf_function_p, instead use has_hard_reg_initial_val.
10290 2018-09-09  Nathan Sidwell  <nathan@acm.org>
10292         * gcc.h (pfatal_with_name): Don't declare here.
10293         * gcc.c (pfatal_with_name): Make static.
10295 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
10297         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
10298         earlyclobber.
10300 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
10302         PR rtl-optimization/85458
10303         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
10304         priority hook to reduce the priority of EXPR.
10306 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
10308         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
10309         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
10310         with X87_ENABLE_FLOAT.  Remove preparation code.
10311         (*float<SWI48:mode><MODEF:mode>2): Rename from
10312         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
10313         math using "enabled" attribute.
10314         (*floatdi<MODEF:mode>2_i387): Rename from
10315         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
10316         enable for 32bit targets only.
10317         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
10318         splitter.
10319         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
10320         as operand 1 predicate.  Rewrite as define_insn_and_split.
10321         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
10323 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
10325         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
10326         to fallthru to FLOAT case.
10328 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
10330         PR target/86731
10331         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
10332         around folding of vec_sl to handle out of range shift values.
10334 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
10336         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
10337         Update callers to gen_fix_trunc<mode>_i387_fisttp
10338         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
10339         nonimmediate_operand.
10340         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
10341         and corresponding splitters.
10342         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
10343         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
10344         (fix_truncdi_i387_with_temp): Remove insn pattern
10345         and corresponding splitters.
10346         (fix_trunc<mode>_i387): Change operand 0 predicate to
10347         nonimmediate_operand.
10348         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
10349         and corresponding splitters.
10350         (*fistdi2_1): Remove.
10351         (fistdi2): Ditto.
10352         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
10353         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
10354         (*fist<mode>2_1): Remove.
10355         (fist<mode>2): Ditto.
10356         (fist<mode>2_with_temp): Remove insn pattern and corresponding
10357         splitters.
10358         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
10359         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
10360         (fistdi2_<rounding>): Change operand 0 predicate to
10361         nonimmediate_operand.
10362         (fistdi2_<rounding>_with_temp): Remove insn pattern
10363         and corresponding splitters.
10364         (fist<mode>2_<rounding>): Change operand 0 predicate to
10365         nonimmediate_operand.
10366         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
10367         and corresponding splitters.
10369         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
10371 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10373         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
10374         the init value.
10376 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
10378         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10379         early gimple folding of vec_splat().
10380         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
10381         * gimple-fold.h: Add an extern define for tree_vec_extract().
10383 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
10385         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
10386         wrappers around TREE_TYPE comparisons.
10388 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
10390         PR target/80080
10391         * config/s390/predicates.md: Add nonsym_memory_operand.
10392         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
10393         contains a SYMBOL_REF, load it into an intermediate pseudo.
10394         (s390_emit_compare_and_swap): Legitimize operand.
10395         * config/s390/s390.md: Use the new nonsym_memory_operand
10396         with UNSPECV_CAS patterns.
10398 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
10400         PR target/80080
10401         * config/s390/s390-passes.def: New file.
10402         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
10403         declaration.
10404         (make_pass_s390_early_mach): Add declaration.
10405         * config/s390/s390.c (make_pass_s390_early_mach):
10406         (s390_option_override): Remove dynamic registration.
10407         * config/s390/t-s390: Add s390-passes.def.
10409 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
10411         * config/s390/s390.c (s390_decompose_constant_pool_ref):
10412         Remove UNSPEC_LTREL_BASE check.
10413         (annotate_constant_pool_refs): Likewise.
10414         (find_constant_pool_ref): Likewise.
10415         (find_ltrel_base): Removed.
10416         (replace_ltrel_base): Removed.
10417         (s390_mainpool_finish): Remove replace_ltrel_base call.
10418         (s390_chunkify_start): Remove pending LTREL_BASE logic.
10419         (s390_chunkify_finish): Remove replace_ltrel_base call.
10420         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
10422 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
10424         PR target/86779
10425         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10426         to speculation_safe_value_not_needed.
10428 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
10429             Bernd Schmidt  <bernds_cb1@t-online.de>
10431         * config/nvptx/nvptx-opts.h: New file.
10432         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
10433         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
10434         (ASM_SPEC): Define.
10435         (TARGET_SM35): New macro.
10436         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
10437         correct predicate.
10438         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
10439         values.
10440         (misa=): New option.
10441         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
10443 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
10445         * config/i386/i386.md (truncdfsf2): Remove expander.
10446         (truncdfsf2_with_temp): Ditto.
10447         (truncxf<mode>2): Ditto.
10448         (*truncdfsf_fast_mixed): Remove insn pattern.
10449         (*truncdfsf_fast_i387): Ditto.
10450         (*truncdfsf_mixed): Ditto.
10451         (*truncdfsf_i387): Ditto.
10452         (*truncdfsf2_i387_1): Ditto.
10453         (*truncxfsf2_mixed): Ditto.
10454         (*truncxfdf2_mixed): Ditto.
10455         (*truncxf<mode>2_i387_noop): Ditto. Update callers
10456         to call gen_truncxf<mode>2 instead.
10457         (*truncxf<mode>2_i387): Remove.
10458         (reg->reg splitters): Remove splitter pattern.
10459         (reg->mem splitters): Ditto.
10461         (truncdfsf2): New insn pattern.
10462         (truncxf<mode>2): Ditto.
10464 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10466         * tree-ssa-math-opts.c (is_mult_by): New function.
10467         (is_square_of): Use the above.
10468         (optimize_recip_sqrt): New function.
10469         (pass_cse_reciprocals::execute): Use the above.
10471 2018-09-05  Richard Biener  <rguenther@suse.de>
10473         PR bootstrap/87134
10474         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
10475         to zero-init the emplaced vec.
10477 2018-09-05  Martin Liska  <mliska@suse.cz>
10479         PR tree-optimization/87205
10480         * tree-switch-conversion.c (pass_lower_switch::execute):
10481         Group cases for switch statements.
10483 2018-09-05  Richard Biener  <rguenther@suse.de>
10485         PR tree-optimization/87217
10486         * tree-ssa-sccvn.c (vuse_valueize): New.
10487         (vn_reference_lookup_pieces): Use it.
10488         (vn_reference_lookup): Likewise.
10490 2018-09-05  Nathan Sidwell  <nathan@acm.org>
10492         PR c++/87137
10493         * stor-layout.c (place_field): Scan forwards to check last
10494         bitfield when ms_bitfield_placement is in effect.
10496 2018-09-05  Richard Biener  <rguenther@suse.de>
10498         PR bootstrap/87225
10499         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
10500         return.
10502 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10503             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
10505         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
10506         * config.gcc (extra_objs): Build it.
10507         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
10508         Likewise.
10509         * config/aarch64/aarch64-passes.def
10510         (pass_tag_collision_avoidance): New pass.
10511         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
10512         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
10513         (aarch64_classify_address): Remove static qualifier.
10514         (aarch64_address_info, aarch64_address_type): Move to...
10515         * config/aarch64/aarch64-protos.h: ... here.
10516         (make_pass_tag_collision_avoidance): New function.
10517         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
10518         New tuning flag.
10520 2018-09-05  Martin Liska  <mliska@suse.cz>
10522         * doc/gcov.texi: Update documentation of humar
10523         readable mode.
10524         * gcov.c (format_count): Print one decimal place, it provides
10525         more fine number of situations like '1G' vs. '1.4G'.
10527 2018-09-05  Martin Liska  <mliska@suse.cz>
10529         PR target/87164
10530         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
10531         * optc-gen.awk: Allow 'Var' for Deprecated options in order
10532         to generate a MASK value.
10534 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
10536         PR debug/86593
10537         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
10538         if frame pointer isn't used.
10539         (compute_frame_pointer_to_fb_displacement): Likewise.
10541 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
10543         PR target/87198
10544         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
10545         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
10546         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
10547         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
10548         and OPTION_MASK_ISA_XSAVEC_UNSET.
10550 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
10552         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
10553         XOR operations in NAND case.
10555 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
10557         * wide-int-range.cc (wide_int_range_convert): New.
10558         * wide-int-range.h (wide_int_range_convert): New.
10559         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
10560         code into wide_int_range_convert.
10561         (extract_range_into_wide_ints): Do not munge anti range constants
10562         into the entire domain.  Just return the range back.
10564 2018-09-04  Martin Liska  <mliska@suse.cz>
10566         * genmatch.c (output_line_directive): Add new argument
10567         fnargs.
10568         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
10570 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
10572         * doc/invoke.texi (Option Summary): Add whitespace.
10574         * doc/invoke.texi (Option Summary): Add -Waligned-new.
10576 2018-09-04  Richard Biener  <rguenther@suse.de>
10578         PR tree-optimization/87211
10579         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
10580         backedge value we're supposed to treat as VARYING also number
10581         the PHI to VARYING in case it got a different value-number already.
10583 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
10585         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
10586         (extract_range_from_binary_expr_1): Do not call
10587         vrp_can_optimize_bit_op.
10588         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
10589         static.
10590         (wide_int_range_get_mask_and_bounds): New.
10591         (wide_int_range_optimize_bit_op): New.
10592         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
10593         (wide_int_range_bit_and): Same.
10594         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
10595         (wide_int_range_optimize_bit_op): New.
10596         (wide_int_range_get_mask_and_bounds): New.
10598 2018-09-04  Richard Biener  <rguenther@suse.de>
10600         PR tree-optimization/87176
10601         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
10602         variable.  When value-numbering a virtual PHI node make sure
10603         to not value-number to the backedge value.
10605 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
10607         * doc/extend.texi (Long Long, Hex Floats): Document support for
10608         long long and hex floats in more recent versions of ISO C++.
10610 2018-09-03  Richard Biener  <rguenther@suse.de>
10612         PR tree-optimization/87177
10613         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
10614         cleanup.
10616 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
10618         * bb-reorder.c (edge_order): Convert to C-qsort-style
10619         tri-state comparator.
10620         (reorder_basic_blocks_simple): Change std::stable_sort to
10621         gcc_stablesort.
10623 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
10625         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
10626         tri-state comparator.
10627         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
10629 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
10631         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
10632         (mergesort): ... here as maximum count for using netsort.
10633         (gcc_qsort): Set nlim to 3 if stable sort is requested.
10634         (gcc_stablesort): New.
10635         * system.h (gcc_stablesort): Declare.
10637 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
10639         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
10640         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
10641         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
10643 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
10645         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
10646         lxsdx and stxsdx alternatives.
10647         (*mov<mode>_hardfloat64): Ditto.
10648         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
10650 2018-09-03  Richard Biener  <rguenther@suse.de>
10652         PR tree-optimization/87200
10653         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
10654         simplify result.
10656 2018-09-03  Martin Liska  <mliska@suse.cz>
10658         PR tree-optimization/87201
10659         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
10660         Fix parenthesis in an expression.
10662 2018-09-03  Richard Biener  <rguenther@suse.de>
10664         PR tree-optimization/87197
10665         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
10666         visited.  CSE the VN_INFO hashtable lookup.
10668         PR tree-optimization/87169
10669         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
10670         iterating make sure there's no extra backedges from irreducible
10671         regions feeding the header.  Mark the destination block
10672         executable.
10674 2018-09-03  Martin Liska  <mliska@suse.cz>
10676         PR driver/83193
10677         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
10678         * common/common-targhooks.c (default_get_valid_option_values):
10679         New function.
10680         * common/common-targhooks.h (default_get_valid_option_values):
10681         Likewise.
10682         * common/config/i386/i386-common.c: Move processor_target_table
10683         from i386.c.
10684         (ix86_get_valid_option_values): New function.
10685         (TARGET_GET_VALID_OPTION_VALUES): New macro.
10686         * config/i386/i386.c (struct ptt): Move to i386-common.c.
10687         (PTA_*): Move all defined masks into i386-common.c.
10688         (ix86_function_specific_restore): Use new processor_cost_table.
10689         * config/i386/i386.h (struct ptt): Moved from i386.c.
10690         (struct pta): Likewise.
10691         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
10692         * doc/tm.texi.in: Likewise.
10693         * opt-suggestions.c (option_proposer::suggest_option):
10694         Pass prefix to build_option_suggestions.
10695         (option_proposer::get_completions): Likewise.
10696         (option_proposer::build_option_suggestions): Use the new target
10697         hook.
10698         * opts.c (struct option_help_tuple): New struct.
10699         (print_filtered_help): Use the new target hook.
10701 2018-09-03  Martin Liska  <mliska@suse.cz>
10703         PR middle-end/59521
10704         * predict.c (set_even_probabilities): Add likely_edges
10705         argument and handle cases where we have precisely one
10706         likely edge.
10707         (combine_predictions_for_bb): Catch also likely_edges.
10708         (tree_predict_by_opcode): Handle gswitch statements.
10709         * tree-cfg.h (find_case_label_for_value): New declaration.
10710         (find_taken_edge_switch_expr): Likewise.
10711         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
10712         Find pivot in decision tree based on probabily, not by number of
10713         nodes.
10715 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
10717         * doc/standards.texi (Standards): Update Objective-C reference.
10719 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10721         * doc/install.texi (Prerequisites): Update link for MPC.
10723 2018-09-01  Michael Matz  <matz@suse.de>
10725         PR tree-optimization/87074
10726         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
10727         PHIs for outer-loop uses.
10729 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10731         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
10732         * doc/invoke.texi (C Dialect Options): Ditto.
10734 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10736         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
10738 2018-08-31  Richard Biener  <rguenther@suse.de>
10740         PR tree-optimization/87168
10741         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
10742         (rpo_elim::eliminate_avail): When OP was not visited it must
10743         be available.
10745 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
10747         * tree-vrp.c (copy_value_range): Convert param "from" from
10748         "value_range *" to "const value_range *".
10749         (range_is_null): Likewise for param "vr".
10750         (range_int_cst_p): Likewise.
10751         (range_int_cst_singleton_p): Likewise.
10752         (symbolic_range_p): Likewise.
10753         (value_ranges_intersect_p): Likewise for both params.
10754         (value_range_nonnegative_p): Likewise for param "vr".
10755         (value_range_constant_singleton): Likewise.
10756         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
10757         (extract_range_into_wide_ints): Likewise for param "vr".
10758         (extract_range_from_multiplicative_op): Likewise for params "vr0"
10759         and "vr1".
10760         (vrp_can_optimize_bit_op): Likewise.
10761         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
10762         "vr1_".
10763         (extract_range_from_unary_expr): Likewise.
10764         (debug_value_range): Likewise for param "vr".
10765         (value_range::dump): Add "const" qualifier.
10766         (vrp_prop::check_array_ref): Convert local "vr" from
10767         "value_range *" to "const value_range *".
10768         (vrp_prop::check_mem_ref): Likewise.
10769         (vrp_prop::visit_stmt): Likewise for local "old_vr".
10770         (vrp_intersect_ranges_1): Likewise for param "vr_1".
10771         (vrp_intersect_ranges): Likewise.
10772         (simplify_stmt_for_jump_threading): Likewise for local "vr".
10773         (vrp_prop::vrp_finalize): Likewise.
10774         * tree-vrp.h (value_range::dump): Add "const" qualifier.
10775         (vrp_intersect_ranges): Add "const" qualifier to params as above.
10776         (extract_range_from_unary_expr): Likewise.
10777         (value_range_constant_singleton): Likewise.
10778         (symbolic_range_p): Likewise.
10779         (copy_value_range): Likewise.
10780         (extract_range_from_binary_expr_1): Likewise.
10781         (range_int_cst_p): Likewise.
10782         (vrp_set_zero_nonzero_bits): Likewise.
10783         (range_int_cst_singleton_p): Likewise.
10785 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
10787         * config/aarch64/arm_neon.h (vabsd_s64): New.
10788         (vnegd_s64): Likewise.
10790 2018-08-31  Martin Jambor  <mjambor@suse.cz>
10792         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
10794 2018-08-31  Martin Liska  <mliska@suse.cz>
10796         * ipa-icf.c (sem_item::add_type): Use
10797         sem_item::m_type_hash_cache.
10798         * ipa-icf.h: Move the cache from sem_item_optimizer
10799         to sem_item.
10801 2018-08-31  Nathan Sidwell  <nathan@acm.org>
10803         * doc/extend.texi (Backwards Compatibility): Remove implicit
10804         extern C leeway of () being (...).
10806 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10808         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
10810 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
10812         PR target/86684
10813         PR target/87149
10814         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
10816 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
10818         PR middle-end/87138
10819         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
10820         gen_int_mode.  Formatting fixes.
10822 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
10824         * target.def (custom_function_descriptors): Improve documentation.
10825         * doc/tm.texi.in (Trampolines): Expand discussion of function
10826         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
10827         beginning of the section.
10828         * doc/tm.texi: Regenerated.
10830 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
10832         * cfg.h (class auto_edge_flag): Spell out the template-id of the
10833         base class in the initializer list.  This is a workaround for
10834         building with older GCC.
10835         (class auto_bb_flag): Likewise.
10837 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
10839         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
10840         (altivec_vcmpequ<VI_char>_p): Remove star.
10841         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
10842         vector load modes.
10843         (expand_strncmp_vec_sequence): New function.
10844         (emit_final_str_compare_vec): New function.
10845         (expand_strn_compare): Add support for vector strncmp.
10846         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
10847         length specification to bytes.
10848         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
10849         (vcmpnezb_p): New pattern.
10850         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
10851         for option -mstring-compare-inline-limit.
10853 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
10855         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
10856         (PTA_SKYLAKE): Add PTA_AES.
10857         (PTA_GOLDMONT): Likewise.
10859 2018-08-29  Jan Hubicka  <jh@suse.cz>
10861         PR lto/86517
10862         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
10863         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
10865 2018-08-29  Jan Hubicka  <jh@suse.cz>
10867         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
10868         TYPE_STUB_DECL.
10869         (hash_tree): Do not visit TYPE_STUB_DECL.
10870         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
10871         stream TYPE_STUB_DECL.
10872         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
10873         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
10874         after free_lang_data.
10875         (type_in_anonymous_namespace_p): Likewise.
10877 2018-08-29  Jan Hubicka  <jh@suse.cz>
10879         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10880         comment that it has to be even number.
10881         (class sreal): Change m_sig type to int32_t.
10882         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10883         int64_t for temporary calculations.
10884         (sreal_verify_basics): Drop one bit from minimum and maximum.
10886 2018-08-30  Richard Biener  <rguenther@suse.de>
10888         PR tree-optimization/87147
10889         * tree-ssa-sccvn.c (SSA_VISITED): New function.
10890         (visit_phi): When the degenerate result is from the backedge and
10891         we didn't visit its definition yet drop to VARYING.
10892         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
10894 2018-08-29  Jan Hubicka  <jh@suse.cz>
10896         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
10897         DECL_VINDEX.
10898         (hash_tree): Likewise.
10900 2018-08-29  Jan Hubicka  <jh@suse.cz>
10902         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
10903         and TYPE_NEXT_REF_TO.
10905 2018-08-29  Jan Hubicka  <jh@suse.cz>
10907         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10908         comment that it has to be even number.
10909         (class sreal): Change m_sig type to int32_t.
10910         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10911         int64_t for temporary calculations.
10912         (sreal_verify_basics): Drop one bit from minimum and maximum.
10914 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
10916         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
10918 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
10920         PR middle-end/86995
10921         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
10922         if to_add is negative.
10924 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10926         PR middle-end/87053
10927         * builtins.c (c_strlen): Improve range checks.
10929 2018-08-29  Martin Sebor  <msebor@redhat.com>
10930             Jeff Law  <law@redhat.com>
10932         PR tree-optimization/86714
10933         PR tree-optimization/86711
10934         * builtins.c (c_strlen): Add arguments to call to string_constant.
10935         * expr.c (string_constant): Add argument.  Detect missing nul
10936         terminator and outermost declaration it's missing in.
10937         * expr.h (string_constant): Add argument.
10938         * fold-const.c (read_from_constant_string): Add arguments to call to
10939         string_constant.
10940         (c_getstr): Likewise.
10941         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
10942         to string_constant.
10943         * tree-ssa-strlen.c (get_stridx): Likewise.
10945 2018-08-29  Jan Hubicka  <jh@suse.cz>
10947         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
10948         Do not stream DECL_VINDEX.
10949         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
10950         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
10951         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
10953 2018-08-29  Richard Biener  <rguenther@suse.de>
10955         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
10956         virtual operands that are not default defs to honor region
10957         boundaries.
10958         (rpo_vn_valueize): Remove ineffective code here.
10960 2018-08-29  Richard Biener  <rguenther@suse.de>
10962         PR tree-optimization/87132
10963         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
10964         when skipping defs reachable over backedges.
10966 2018-08-29  Richard Biener  <rguenther@suse.de>
10968         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
10969         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
10970         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
10971         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
10972         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
10973         (vn_reference_lookup_3): Remove use of const_parms.
10974         (free_rpo_vn): Do not free const_parms.
10975         (do_rpo_vn): Do not call init_const_parms.
10976         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
10977         SSA_NAME_POINTS_TO_READONLY_MEMORY.
10978         (call_may_clobber_ref_p_1): Likewise.
10980 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
10982         PR other/86726
10983         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
10984         (-O): Ditto.
10985         (-ftree-scev-cprop): Document.
10987 2018-08-29  Jan Hubicka  <jh@suse.cz>
10989         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
10990         parameters.
10991         (sreal constructor): Update.
10992         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
10993         sreal:operator/): Update.
10995 2018-08-29  Martin Liska  <mliska@suse.cz>
10997         * tree-switch-conversion.c (switch_conversion::expand):
10998         Strenghten assumption about gswitch statements.
11000 2018-08-29  Richard Biener  <rguenther@suse.de>
11002         PR tree-optimization/87117
11003         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
11004         re-value-number released SSA VDEFs.
11006 2018-08-29  Richard Biener  <rguenther@suse.de>
11008         PR tree-optimization/87126
11009         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
11011 2018-08-28  Jim Wilson  <jimw@sifive.com>
11013         * config/riscv/pic.md: Rewrite.
11014         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
11015         invalid address.
11016         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
11017         (SOFTF, default_load, softload, softstore): New.
11019 2018-08-28  Jeff Law  <law@redhat.com>
11021         * fold-const.c (fold_binary_loc): Remove recently added assert.
11023 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
11025         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
11026         to OP parmeter of generated function.
11028 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
11030         PR tree-optimization/87009
11031         * match.pd: Add boolean optimizations.
11033 2018-08-28  Martin Sebor  <msebor@redhat.com>
11035         PR middle-end/86631
11036         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
11037         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
11038         (pass_walloca::gate): Use it.
11039         (alloca_call_type): Same.
11040         (pass_walloca::execute): Same.
11041         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
11043 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
11045         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
11046         GCC_VERSION for usage of "__gcc_dump_printf__" format from
11047         >= 3005 to >= 9000.
11049 2018-08-28  Richard Biener  <rguenther@suse.de>
11051         PR tree-optimization/87124
11052         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
11053         constants before looking up avail.
11055 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
11057         PR middle-end/87099
11058         * calls.c (maybe_warn_nonstring_arg): Punt early if
11059         warn_stringop_overflow is zero.  Don't call get_range_strlen
11060         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
11061         Swap comparison operands to have constants on rhs.  Only use
11062         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
11063         increment lenrng[0].
11065 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
11067         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
11068         use of tree_to_shwi.  Remove duplicated test for the size being
11069         a whole number of bytes.
11071 2018-08-28  Richard Biener  <rguenther@suse.de>
11073         PR tree-optimization/87117
11074         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
11075         Handle removed stmt without LHS (GIMPLE_NOP).
11077 2018-08-28  Richard Biener  <rguenther@suse.de>
11079         PR tree-optimization/87117
11080         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
11081         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
11083 2018-08-28  Richard Biener  <rguenther@suse.de>
11085         PR tree-optimization/87117
11086         * tree-ssa-pre.c (compute_avail): Do not make expressions
11087         with predicated values available.
11088         (get_expr_value_id): Assert we do not run into predicated value
11089         expressions.
11091 2018-08-28  Richard Biener  <rguenther@suse.de>
11093         PR tree-optimization/87117
11094         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
11095         get virtual operands.
11096         (get_expr_operands): Handle STRING_CST like other decls.
11098 2018-08-28  Martin Liska  <mliska@suse.cz>
11100         * tree.h: Update documentation of fndecl_built_in_p
11101         functions.
11104 2018-08-27  Jeff Law  <law@redhat.com>
11105         PR tree-optimization/87110
11106         * tree-ssa-dse.c (compute_trims): Handle non-constant
11107         TYPE_SIZE_UNIT.
11109 2018-08-27  Martin Sebor  <msebor@redhat.com>
11111         PR tree-optimization/86914
11112         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
11114 2018-08-27  Martin Sebor  <msebor@redhat.com>
11116         PR tree-optimization/87112
11117         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
11118         the type of the bound argument.
11120 2018-08-27  Jeff Law  <law@redhat.com>
11122         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
11123         type does not have a TYPE_SIZE_UNIT.
11125 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
11127         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
11128         with include of backend.h.
11130 2018-08-27  Richard Biener  <rguenther@suse.de>
11132         PR tree-optimization/86927
11133         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
11134         use const cond reduction code.
11136 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
11138         PR tree-optimization/85758
11139         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
11141 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
11143         PR c++/87091
11144         * diagnostic-show-locus.c (class layout_range): Update for
11145         conversion of show_caret_p to a tri-state.
11146         (layout_range::layout_range): Likewise.
11147         (make_range): Likewise.
11148         (layout::maybe_add_location_range): Likewise.
11149         (layout::should_print_annotation_line_p): Don't show annotation
11150         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
11151         (layout::get_state_at_point): Update for conversion of
11152         show_caret_p to a tri-state.  Bail out early for
11153         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
11154         underlining or source colorization.
11155         (gcc_rich_location::add_location_if_nearby): Update for conversion
11156         of show_caret_p to a tri-state.
11157         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
11158         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
11159         Likewise.
11160         (selftest::test_one_liner_labels): Likewise.
11161         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
11162         conversion of show_caret_p to a tri-state.
11163         * pretty-print.c (text_info::set_location): Likewise.
11164         * pretty-print.h (text_info::set_location): Likewise.
11165         * substring-locations.c (format_warning_n_va): Likewise.
11166         * tree-diagnostic.c (default_tree_printer): Likewise.
11167         * tree-pretty-print.c (newline_and_indent): Likewise.
11169 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
11171         PR c++/87091
11172         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
11173         line above for line-insertion fix-it hints.
11174         (selftest::test_fixit_insert_containing_newline): Update the
11175         expected results, and add a test with line-numbering enabled.
11177 2018-08-27  Martin Liska  <mliska@suse.cz>
11179         PR sanitizer/86962
11180         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
11181         params with DECL_HAS_VALUE_EXPR_P.
11183 2018-08-27  Martin Liska  <mliska@suse.cz>
11185         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
11186         selected expansion strategy.
11188 2018-08-27  Martin Liska  <mliska@suse.cz>
11190         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
11191         * builtins.c (is_builtin_fn): Likewise.
11192         * attribs.c (diag_attr_exclusions): Use new function
11193         fndecl_built_in_p and remove check for FUNCTION_DECL if
11194         possible.
11195         (builtin_mathfn_code): Likewise.
11196         (fold_builtin_expect): Likewise.
11197         (fold_call_expr): Likewise.
11198         (fold_builtin_call_array): Likewise.
11199         (fold_call_stmt): Likewise.
11200         (set_builtin_user_assembler_name): Likewise.
11201         (is_simple_builtin): Likewise.
11202         * calls.c (gimple_alloca_call_p): Likewise.
11203         (maybe_warn_nonstring_arg): Likewise.
11204         * cfgexpand.c (expand_call_stmt): Likewise.
11205         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
11206         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
11207         (cgraph_node::verify_node): Likewise.
11208         * cgraphclones.c (build_function_decl_skip_args): Likewise.
11209         (cgraph_node::create_clone): Likewise.
11210         * config/arm/arm.c (arm_insert_attributes): Likewise.
11211         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
11212         * dse.c (scan_insn): Likewise.
11213         * expr.c (expand_expr_real_1): Likewise.
11214         * fold-const.c (operand_equal_p): Likewise.
11215         (fold_binary_loc): Likewise.
11216         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
11217         * gimple-low.c (lower_stmt): Likewise.
11218         * gimple-pretty-print.c (dump_gimple_call): Likewise.
11219         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11220         Likewise.
11221         * gimple.c (gimple_build_call_from_tree): Likewise.
11222         (gimple_call_builtin_p): Likewise.
11223         (gimple_call_combined_fn): Likewise.
11224         * gimplify.c (gimplify_call_expr): Likewise.
11225         (gimple_boolify): Likewise.
11226         (gimplify_modify_expr): Likewise.
11227         (gimplify_addr_expr): Likewise.
11228         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
11229         * ipa-cp.c (determine_versionability): Likewise.
11230         * ipa-fnsummary.c (compute_fn_summary): Likewise.
11231         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
11232         * ipa-split.c (visit_bb): Likewise.
11233         (split_function): Likewise.
11234         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11235         * lto-cgraph.c (input_node): Likewise.
11236         * lto-streamer-out.c (write_symbol): Likewise.
11237         * omp-low.c (setjmp_or_longjmp_p): Likewise.
11238         (lower_omp_1): Likewise.
11239         * predict.c (strip_predict_hints): Likewise.
11240         * print-tree.c (print_node): Likewise.
11241         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
11242         * trans-mem.c (is_tm_irrevocable): Likewise.
11243         (is_tm_load): Likewise.
11244         (is_tm_simple_load): Likewise.
11245         (is_tm_store): Likewise.
11246         (is_tm_simple_store): Likewise.
11247         (is_tm_abort): Likewise.
11248         (tm_region_init_1): Likewise.
11249         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
11250         * tree-cfg.c (verify_gimple_call): Likewise.
11251         (move_stmt_r): Likewise.
11252         (stmt_can_terminate_bb_p): Likewise.
11253         * tree-eh.c (lower_eh_constructs_2): Likewise.
11254         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
11255         * tree-inline.c (remap_gimple_stmt): Likewise.
11256         (copy_bb): Likewise.
11257         (estimate_num_insns): Likewise.
11258         (fold_marked_statements): Likewise.
11259         * tree-sra.c (scan_function): Likewise.
11260         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
11261         (optimize_stack_restore): Likewise.
11262         (pass_fold_builtins::execute): Likewise.
11263         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11264         (mark_all_reaching_defs_necessary_1): Likewise.
11265         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
11266         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11267         (pass_forwprop::execute): Likewise.
11268         * tree-ssa-loop-im.c (stmt_cost): Likewise.
11269         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
11270         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
11271         * tree-ssa-strlen.c (get_string_length): Likewise.
11272         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
11273         (find_func_aliases_for_call): Likewise.
11274         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
11275         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
11276         * tree-tailcall.c (find_tail_calls): Likewise.
11277         * tree.c (need_assembler_name_p): Likewise.
11278         (free_lang_data_in_decl): Likewise.
11279         (get_call_combined_fn): Likewise.
11280         * ubsan.c (is_ubsan_builtin_p): Likewise.
11281         * varasm.c (incorporeal_function_p): Likewise.
11282         * tree.h (DECL_BUILT_IN): Remove and replace with
11283         fndecl_built_in_p.
11284         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
11285         (fndecl_built_in_p): New.
11287 2018-08-27  Martin Liska  <mliska@suse.cz>
11289         PR tree-optimization/86847
11290         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
11291         Dump also subtree probability.
11292         (switch_decision_tree::do_jump_if_equal): New function.
11293         (switch_decision_tree::emit_case_nodes): Handle special
11294         situations in balanced tree that can be emitted much simpler.
11295         Fix calculation of probabilities that happen in tree expansion.
11296         * tree-switch-conversion.h (struct cluster): Add
11297         is_single_value_p.
11298         (struct simple_cluster): Likewise.
11299         (struct case_tree_node): Add new function has_child.
11300         (do_jump_if_equal): New.
11302 2018-08-27  Martin Liska  <mliska@suse.cz>
11304         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
11305         Add new argument to bit_test_cluster constructor.
11306         (bit_test_cluster::emit): Set bits really number of values
11307         handlel by a test.
11308         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
11309         probability argument.
11310         * tree-switch-conversion.h (struct bit_test_cluster):
11311         Add m_handles_entire_switch.
11313 2018-08-27  Martin Liska  <mliska@suse.cz>
11315         PR tree-optimization/86702
11316         * tree-switch-conversion.c (jump_table_cluster::emit):
11317         Make probabilities even for values in jump table
11318         according to number of cases handled.
11319         (switch_decision_tree::compute_cases_per_edge): Pass
11320         argument to reset_out_edges_aux function.
11321         (switch_decision_tree::analyze_switch_statement): Likewise.
11322         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
11323         Make it static.
11325 2018-08-27  Martin Liska  <mliska@suse.cz>
11327         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
11328         cfun argument explicitly.
11329         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
11330         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
11331         function gimple_switch_default_bb.
11332         (convert_switch_statements):
11333         (expand_builtins):
11334         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
11335         * stmt.c (label_to_block_fn): Use label_to_block and pass
11336         cfun argument explicitly and use gimple_switch_label_bb.
11337         (expand_case): Likewise.
11338         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
11339         cfun argument explicitly. Likewise.
11340         (make_edges_bb): Likewise.
11341         (make_cond_expr_edges): Likewise.
11342         (get_cases_for_edge): Likewise.
11343         (make_gimple_switch_edges): Likewise.
11344         (label_to_block_fn): Likewise.
11345         (label_to_block): Likewise.
11346         (make_goto_expr_edges): Likewise.
11347         (make_gimple_asm_edges): Likewise.
11348         (main_block_label): Likewise.
11349         (group_case_labels_stmt): Likewise.
11350         (find_taken_edge_computed_goto): Likewise.
11351         (find_taken_edge_switch_expr): Likewise.
11352         (gimple_verify_flow_info): Likewise.
11353         (gimple_redirect_edge_and_branch): Likewise.
11354         (gimple_switch_label_bb): New function.
11355         (gimple_switch_default_bb): Likewise.
11356         (gimple_switch_edge): Likewise.
11357         (gimple_switch_default_edge): Likewise.
11358         * tree-cfg.h (label_to_block_fn): Remove and replace ...
11359         (label_to_block): ... with this.
11360         (gimple_switch_label_bb): New.
11361         (gimple_switch_default_bb): Likewise.
11362         (gimple_switch_edge): Likewise.
11363         (gimple_switch_default_edge): Likewise.
11364         * tree-cfgcleanup.c (convert_single_case_switch): Use
11365         new gimple functions and pass new argument to label_to_block.
11366         (cleanup_control_flow_bb):
11367         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
11368         cfun argument explicitly.
11369         (make_eh_edges): Likewise.
11370         (redirect_eh_dispatch_edge): Likewise.
11371         (lower_resx): Likewise.
11372         (lower_eh_dispatch): Likewise.
11373         (maybe_remove_unreachable_handlers): Likewise.
11374         (unsplit_eh): Likewise.
11375         (cleanup_empty_eh): Likewise.
11376         (verify_eh_edges): Likewise.
11377         (verify_eh_dispatch_edge): Likewise.
11378         * tree-ssa-dom.c (record_edge_info): Likewise.
11379         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
11380         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
11381         (thread_through_normal_block): Likewise.
11382         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
11383         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
11384         * tree-switch-conversion.c (switch_conversion::collect): Use new
11385         gimple functions.
11386         (switch_conversion::check_final_bb): Likewise.
11387         (switch_conversion::gather_default_values): Pass new argument
11388         to label_to_block.
11389         (switch_conversion::build_constructors): Likewise.
11390         (switch_decision_tree::compute_cases_per_edge): Use new
11391         gimple_switch_edge function.
11392         (switch_decision_tree::analyze_switch_statement): Pass new argument
11393         to label_to_block.
11394         (switch_decision_tree::try_switch_expansion): Use
11395         gimple_switch_default_edge.
11396         * tree-vrp.c (find_switch_asserts): Pass new argument
11397         to label_to_block.
11398         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
11399         (vr_values::simplify_switch_using_ranges): Likewise.
11401 2018-08-27  Richard Biener  <rguenther@suse.de>
11403         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
11404         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
11406         * tree-ssa-sccvn.h (struct vn_pval): New structure.
11407         (struct vn_nary_op_s): Add unwind_to member.  Add
11408         predicated_values flag and put result into a union together
11409         with a linked list of vn_pval.
11410         (struct vn_ssa_aux): Add name member to make maintaining
11411         a map of SSA name to vn_ssa_aux possible.  Remove no longer
11412         needed info, dfsnum, low, visited, on_sccstack, use_processed
11413         and range_info_anti_range_p members.
11414         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
11415         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
11416         New functions.
11417         (vn_valueize): New global.
11418         (vn_context_bb): Likewise.
11419         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
11420         VN_INFO_PTR_INFO): Remove.
11421         * tree-ssa-sccvn.c: ... (rewrite)
11422         (pass_fre::execute): For -O2+ initialize loops and run
11423         RPO VN in optimistic mode (iterating).  For -O1 and -Og
11424         run RPO VN in non-optimistic mode.
11425         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
11426         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
11427         * doc/invoke.texi (sccvn-max-scc-size): Remove.
11428         (rpo-vn-max-loop-depth): Document.
11429         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
11430         when valuezing the VUSE signals we walked out of the region.
11431         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
11432         (phi_translate): Set VN context block to use for availability
11433         lookup.
11434         (compute_avail): Likewise.
11435         (pre_valueize): New function.
11436         (pass_pre::execute): Adjust to the RPO VN API.
11438         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
11439         (propagate_constants_for_unrolling): Remove.
11440         (tree_unroll_loops_completely): Perform value-numbering
11441         on the unrolled bodies loop parent.
11443 2018-08-27  Richard Biener  <rguenther@suse.de>
11445         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
11446         for partial antic compute.
11448 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
11450         PR rtl-optimization/87065
11451         * combine.c (simplify_if_then_else): Formatting fix.
11452         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
11453         check.
11454         (known_cond): Don't return const_true_rtx for vector modes.  Use
11455         CONST0_RTX instead of const0_rtx.  Formatting fixes.
11457 2018-08-27  Martin Liska  <mliska@suse.cz>
11459         PR gcov-profile/87069
11460         * gcov.c (process_file): Record files already processed
11461         and warn about a file being processed multiple times.
11463 2018-08-27  Martin Liska  <mliska@suse.cz>
11465         PR driver/83193
11466         * config/aarch64/aarch64.c (aarch64_override_options_internal):
11467         Set default values for x_aarch64_*_string strings.
11468         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
11469         prefix.  For -mabi do not print '=ABI' in help and use
11470         <option_value> format for -msve-vector-bits and -moverride
11471         options.
11473 2018-08-26  Jeff Law  <law@redhat.com>
11475         * config/mips/frame-header-opt.c: Include "backend.h" rather than
11476         "cfg.h"
11478 2018-08-26  Marek Polacek  <polacek@redhat.com>
11480         PR c++/87029, Implement -Wredundant-move.
11481         * doc/invoke.texi: Document -Wredundant-move.
11483 2018-08-25  Martin Sebor  <msebor@redhat.com>
11485         PR tree-optimization/87059
11486         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
11487         to the same type as the other.
11488         * fold-const.c (fold_binary_loc): Assert expectation.
11490 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
11492         * config/darwin.c (machopic_legitimize_pic_address): Clean up
11493         extraneous parentheses, dead code section and formatting.
11495 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
11497         PR c++/87091
11498         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
11499         wide enough for jumps in the line-numbering to be visible.
11500         (layout::print_gap_in_line_numbering): New member function.
11501         (layout::calculate_line_spans): When using line numbering, merge
11502         line spans that are only 1 line apart.
11503         (diagnostic_show_locus): When printing line numbers, show gaps in
11504         line numbering directly, rather than printing headers.
11505         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
11506         line-numbering with multiple line spans.
11507         (selftest::test_fixit_insert_containing_newline_2): Add test of
11508         line-numbering, in which the spans are close enough to be merged.
11510 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
11512         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
11513         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
11514         * tree-vrp.c (range_is_nonnull): Remove.
11515         (range_includes_zero_p): Accept value_range instead of min/max.
11516         (extract_range_from_binary_expr_1): Do not early bail on
11517         POINTER_PLUS_EXPR.
11518         Use range_includes_zero_p instead of range_is_nonnull.
11519         (extract_range_from_unary_expr): Use range_includes_zero_p instead
11520         of range_is_nonnull.
11521         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
11522         special case VR_ANTI_RANGE.
11523         (vrp_finalize): Same.
11524         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
11525         instead of min/max.
11526         (range_is_nonnull): Remove.
11527         * vr-values.c (vrp_stmt_computes_nonzero): Use
11528         range_includes_zero_p instead of range_is_nonnull.
11529         (extract_range_basic): Pass value_range to range_includes_zero_p
11530         instead of range_is_nonnull.
11532 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
11534         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
11535         * emit-rtl.h (rtl_data): Remove return_bnd.
11536         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
11537         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
11538         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
11539         (POINTER_BOUNDS_MODE): Remove definition.
11540         (make_pointer_bounds_mode): Remove.
11541         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
11542         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
11543         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
11544         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
11545         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
11546         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
11547         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
11549         * config/i386/i386-modes.def (BND32, BND64): Remove.
11550         * config/i386/i386.c (dbx_register_map): Remove bound registers.
11551         (dbx64_register_map): Ditto.
11552         (svr4_dbx_register_map): Ditto.
11553         (indirect_thunk_bnd_needed): Remove.
11554         (indirect_thunks_bnd_used): Ditto.
11555         (indirect_return_bnd_needed): Ditto.
11556         (indirect_return_via_cx_bnd): Ditto.
11557         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
11558         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
11559         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
11560         (output_indirect_thunk_function): Remove handling of
11561         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
11562         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
11563         (ix86_save_reg): Remove handling of crtl->return_bnd.
11564         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
11565         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
11566         and UNSPEC_BNDLX_ADDR.
11567         (ix86_output_indirect_branch_via_reg): Remove handling of
11568         indirect_thunk_prefix_bnd.
11569         (ix86_output_indirect_branch_via_push): Ditto.
11570         (ix86_output_function_return): Ditto.
11571         (ix86_output_indirect_function_return): Ditto.
11572         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
11573         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
11574         (CALL_USED_REGISTERS): Ditto.
11575         (REG_ALLOC_ORDER): Update for removal of bound registers.
11576         (HI_REGISTER_NAMES): Ditto.
11577         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
11578         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
11579         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
11580         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
11581         (FIRST_PSEUDO_REG): Update.
11582         (BND): Remove mode iterator.
11583         * config/i386/predicates.md (bnd_mem_operator): Remove.
11585 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
11587         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
11588         vectors.
11590 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
11592         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
11593         the case in which the permute needs only a single element and
11594         repeats for every vector of the result.  Extend that case to
11595         handle variable-length vectors.
11596         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11598 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
11600         PR debug/79342
11601         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
11602         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
11604 2018-08-24  Richard Biener  <rguenther@suse.de>
11606         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
11607         bb_flags_allocated members.
11608         (auto_flag): New RAII class for allocating flags.
11609         (auto_edge_flag): New RAII class for allocating edge flags.
11610         (auto_bb_flag): New RAII class for allocating bb flags.
11611         * cfgloop.c (verify_loop_structure): Allocate temporary edge
11612         flag dynamically.
11613         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
11614         in favor of temporarily allocated BB flag.
11615         * hsa-brig.c: Re-order includes.
11616         * hsa-dump.c: Likewise.
11617         * hsa-regalloc.c: Likewise.
11618         * print-rtl.c: Likewise.
11619         * profile-count.c: Likewise.
11621 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
11623         PR target/86989
11624         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
11625         the TOC register.
11627 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
11629         PR 87073/bootstrap
11630         * wide-int-range.cc (wide_int_range_div): Do not ignore result
11631         from wide_int_range_multiplicative_op.
11633 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11635         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
11636         "permutaion".
11638 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
11640         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
11641         to 'expanded'.
11643 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
11645         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
11646         full GENERIC expression used for replacement.
11648 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
11650         * tree-vrp.c (abs_extent_range): Remove.
11651         (extract_range_into_wide_ints): Pass wide ints by reference.
11652         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
11653         Pass wide ints by reference in all calls to
11654         extract_range_into_wide_ints.
11655         * wide-int-range.cc (wide_int_range_div): New.
11656         * wide-int-range.h (wide_int_range_div): New.
11657         (wide_int_range_includes_zero_p): New.
11658         (wide_int_range_zero_p): New.
11660 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
11662         * config/aarch64/aarch64.md (arches): New enum.
11663         (arch): New enum attr.
11664         (arch_enabled): New attr.
11665         (enabled): Now uses arch_enabled only.
11666         (simd, sve, fp16): Removed attribute.
11667         (fp): Attr now defined in terms of 'arch'.
11668         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
11669         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
11670         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
11671         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
11672         attributes into 'arch'.
11673         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
11674         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
11675         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
11676         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
11677         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
11678         'simd' attribute into 'arch'.
11679         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
11680         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
11681         Convert use of 'fp' attribute to 'arch'.
11682         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
11683         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
11684         into 'arch'.
11685         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
11686         (different modes) Merge 'fp' and 'simd' into 'arch'.
11687         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
11688         'simd' into 'arch'.
11690 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
11692         PR rtl-optimization/87026
11693         * expmed.c (canonicalize_comparison): If we can no longer create
11694         pseudoregisters, don't.
11696 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
11698         PR target/86951
11699         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
11700         prototype.
11701         * config/arm/arm.c (speculation_barrier_libfunc): New static
11702         variable.
11703         (arm_init_libfuncs): Initialize it.
11704         (arm_emit_speculation_barrier): New function.
11705         * config/arm/arm.md (speculation_barrier): Call
11706         arm_emit_speculation_barrier for architectures that do not have
11707         DSB or ISB.
11708         (speculation_barrier_insn): Only match on Armv7 or later.
11710 2018-08-23  Richard Biener  <rguenther@suse.de>
11712         PR middle-end/87024
11713         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
11714         calls.
11716 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11718         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
11719         of single-vector TBLs.
11720         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
11721         one input is given.
11723 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11725         PR target/85910
11726         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
11727         aarch64_evpc_tbl guard.
11729 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11731         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
11732         behaviour.
11734 2018-08-22  Martin Sebor  <msebor@redhat.com>
11736         PR middle-end/87052
11737         * tree-pretty-print.c (pretty_print_string): Add argument.
11738         (dump_generic_node): Call to pretty_print_string with string size.
11740 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
11742         PR rtl-optimization/86771
11743         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
11744         of two SETs into those two SETs, one to be placed at i2, if that SETs
11745         destination is modified between i2 and i3.
11747 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11749         PR tree-optimization/86725
11750         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
11751         function.
11752         (vect_analyze_scalar_cycles_1): Check it.
11754 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11756         PR tree-optimization/86725
11757         * tree-vect-loop.c (vect_is_simple_reduction): When treating
11758         an outer loop phi as a double reduction, make sure that the
11759         single user of the phi result is an inner loop phi.
11761 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11763         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
11764         grouped stores with gaps to a strided group.
11766 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11768         * tree-vect-stmts.c (get_group_load_store_type)
11769         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
11770         first statement in a group.
11772 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11774         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
11775         the sequence used in gcc/gcc.c.
11777 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11779         PR other/704
11780         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
11781         building it.
11783 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11785         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
11786         Darwin10-specific unwinder-shim.
11787         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
11788         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
11789         New to cater for Darwin10 Rosetta.
11791 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11793         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
11794         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
11796 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11798         PR bootstrap/81033
11799         PR target/81733
11800         PR target/52795
11801         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
11802         (dwarf2out_switch_text_section): Generate a local label for the
11803         second function sub-section and apply it as the second FDE start
11804         label.
11805         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
11806         second sub-section start.
11808 2018-08-22  Richard Biener  <rguenther@suse.de>
11810         PR tree-optimization/86988
11811         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
11813 2018-08-22  Richard Biener  <rguenther@suse.de>
11815         PR tree-optimization/86945
11816         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
11818 2018-08-22  Alexandre Oliva <oliva@adacore.com>
11820         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
11821         a comment about how uses of r2 for .sdata2 come about.
11823 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
11825         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
11827 2018-08-21  Marek Polacek  <polacek@redhat.com>
11829         PR c++/86981, Implement -Wpessimizing-move.
11830         * doc/invoke.texi: Document -Wpessimizing-move.
11832 2018-08-21  Jan Hubicka  <jh@suse.cz>
11834         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
11835         * tree.h (is_redundant_typedef): Remove.
11836         * dwarf2out.c (is_redundant_typedef): Turn into static function.
11838 2018-08-21  Jan Hubicka  <jh@suse.cz>
11840         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
11841         when possible.
11843 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
11845         * expmed.c (extract_low_bits): Reject invalid subregs early.
11847 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11849         PR middle-end/86121
11850         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
11851         behaviour.
11853 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11855         * config/vxworks.h: Guard vxworks_asm_out_constructor and
11856         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
11857         * config/vxworks.c: Likewise.
11859 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11861         * config/vxworks.c: Set targetm.have_ctors_dtors
11862         if HAVE_INITFINI_ARRAY_SUPPORT.
11863         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
11864         if HAVE_INITFINI_ARRAY_SUPPORT.
11866 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11868         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
11869         default search path for VxWorks < 7.
11871 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11873         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
11874         (get_format_string): Refer to c_getstr.
11876 2018-08-21  Tom de Vries  <tdevries@suse.de>
11878         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11879         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11880         (debuginfo_early_stop): Declare.
11881         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11882         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11883         (debuginfo_early_stop): New function.
11884         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
11885         and debuginfo_early_stop.
11886         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
11887         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
11888         (general_init): Call debuginfo_early_init.
11889         (finalize): Call debuginfo_fini.
11890         (do_compile): Call debuginfo_init.
11891         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
11892         -fdump-early-debug.
11893         (@item -fdump-debug, @item -fdump-earlydebug): Add.
11895 2018-08-21  Tom de Vries  <tdevries@suse.de>
11897         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
11898         flag_dump_noaddr and flag_dump_unnumbered.
11900 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
11902         * wide-int-range.cc (wide_int_range_abs): New.
11903         (wide_int_range_order_set): Rename from wide_int_range_min_max.
11904         * wide-int-range.h (wide_int_range_abs): New.
11905         (wide_int_range_min_max): New.
11906         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
11907         case to call wide_int_range_abs.
11908         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
11909         (extract_range_from_abs_expr): Delete.
11911 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
11913         PR target/87033
11914         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
11915         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
11916         for indexed loads.
11918 2018-08-20  Nathan Sidwell  <nathan@acm.org>
11919             Jeff Law <law@redhat.com>
11921         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
11922         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
11924 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
11926         PR other/84889
11927         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
11928         (decl_attributes): Likewise.
11929         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
11930         instance.
11931         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11932         * diagnostic-core.h (class auto_diagnostic_group): New class.
11933         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
11934         (diagnostic_report_diagnostic): Handle the first diagnostics within
11935         a group.
11936         (emit_diagnostic): Add auto_diagnostic_group instance.
11937         (inform): Likewise.
11938         (inform_n): Likewise.
11939         (warning): Likewise.
11940         (warning_at): Likewise.
11941         (warning_n): Likewise.
11942         (pedwarn): Likewise.
11943         (permerror): Likewise.
11944         (error): Likewise.
11945         (error_n): Likewise.
11946         (error_at): Likewise.
11947         (sorry): Likewise.
11948         (fatal_error): Likewise.
11949         (internal_error): Likewise.
11950         (internal_error_no_backtrace): Likewise.
11951         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
11952         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
11953         * diagnostic.h (struct diagnostic_context): Add fields
11954         "diagnostic_group_nesting_depth",
11955         "diagnostic_group_emission_count", "begin_group_cb",
11956         "end_group_cb".
11957         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
11958         Add auto_diagnostic_group instance(s).
11959         (find_explicit_erroneous_behavior): Likewise.
11960         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
11961         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
11962         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
11963         (gimplify_va_arg_expr): Likewise.
11964         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
11965         (HSA_SORRY_AT): Likewise.
11966         * ipa-devirt.c (compare_virtual_tables): Likewise.
11967         (warn_odr): Likewise.
11968         * multiple_target.c (expand_target_clones): Likewise.
11969         * opts-common.c (cmdline_handle_error): Likewise.
11970         * reginfo.c (globalize_reg): Likewise.
11971         * substring-locations.c (format_warning_n_va): Likewise.
11972         * tree-inline.c (expand_call_inline): Likewise.
11973         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
11974         * tree-ssa-loop-niter.c
11975         (do_warn_aggressive_loop_optimizations): Likewise.
11976         * tree-ssa-uninit.c (warn_uninit): Likewise.
11977         * tree.c (warn_deprecated_use): Likewise.
11979 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
11981         PR target/87014
11982         * config/i386/i386.md (eh_return): Always update EH return
11983         address in word_mode.
11985 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
11987         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
11988         TARGET_SPLIT_COMPLEX_ARG is defined.
11990 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11992         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
11994 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11996         PR target/86984
11997         * expr.c (expand_assignment): Assert that bitpos is positive.
11998         (store_field): Likewise
11999         (expand_expr_real_1): Make sure that bitpos is positive.
12000         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
12001         integer overflow.
12003 2018-08-20  Nathan Sidwell  <nathan@acm.org>
12005         * Makefile.in (CPP_ID_DATA_H): Delete.
12006         (CPP_INTERNAL_H): Don't add it.
12007         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
12008         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
12010 2018-08-20  Richard Biener  <rguenther@suse.de>
12012         PR tree-optimization/78655
12013         * tree-vrp.c (extract_range_from_binary_expr_1): Make
12014         pointer + offset nonnull if either operand is nonnull work.
12016 2018-08-20  Tom de Vries  <tdevries@suse.de>
12018         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
12019         unless the referenced die describes the added property using
12020         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
12021         Otherwise, add a DW_AT_location to the referenced die.
12023 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
12025         PR target/86994
12026         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
12027         register_operand when calling ix86_set_reg_reg_cost.
12028         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
12029         Set *total to 0 for operands that satisfy x86_64_immediate_operand
12030         predicate and to 1 otherwise.
12032 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
12034         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
12035         emit a diagnostic that it is not supported and reset the option.
12036         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
12037         supported and consume the option.  (ASM_FINAL_SPEC): New.
12039 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
12041         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
12042         a sentence.
12044 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
12046         C-SKY port: Documentation
12048         * doc/extend.texi (C-SKY Function Attributes): New section.
12049         * doc/invoke.texi (Option Summary): Add C-SKY options.
12050         (C-SKY Options): New section.
12051         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
12053 2018-08-17  Jojo  <jijie_rong@c-sky.com>
12054             Huibin Wang  <huibin_wang@c-sky.com>
12055             Sandra Loosemore  <sandra@codesourcery.com>
12056             Chung-Lin Tang  <cltang@codesourcery.com>
12058         C-SKY port: Backend implementation
12060         * config/csky/*: New.
12061         * common/config/csky/*: New.
12063 2018-08-17  Jojo  <jijie_rong@c-sky.com>
12064             Huibin Wang  <huibin_wang@c-sky.com>
12065             Sandra Loosemore  <sandra@codesourcery.com>
12066             Chung-Lin Tang  <cltang@codesourcery.com>
12067             Andrew Jenner  <andrew@codesourcery.com>
12069         C-SKY port: Configury
12071         * config.gcc (csky-*-*): New.
12072         * configure.ac: Add csky to targets for dwarf2 debug_line support.
12073         * configure: Regenerated.
12075 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
12077         * dump-context.h: Include "dumpfile.h".
12078         (dump_context::dump_printf_va): Convert final param from va_list
12079         to va_list *.  Convert from ATTRIBUTE_PRINTF to
12080         ATTRIBUTE_GCC_DUMP_PRINTF.
12081         (dump_context::dump_printf_loc_va): Likewise.
12082         * dumpfile.c: Include "stringpool.h".
12083         (make_item_for_dump_printf_va): Delete.
12084         (make_item_for_dump_printf): Delete.
12085         (class dump_pretty_printer): New class.
12086         (dump_pretty_printer::dump_pretty_printer): New ctor.
12087         (dump_pretty_printer::emit_items): New member function.
12088         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
12089         function.
12090         (dump_pretty_printer::emit_item): New member function.
12091         (dump_pretty_printer::stash_item): New member function.
12092         (dump_pretty_printer::format_decoder_cb): New member function.
12093         (dump_pretty_printer::decode_format): New member function.
12094         (dump_context::dump_printf_va): Reimplement in terms of
12095         dump_pretty_printer.
12096         (dump_context::dump_printf_loc_va): Convert final param from va_list
12097         to va_list *.
12098         (dump_context::begin_scope): Reimplement call to
12099         make_item_for_dump_printf.
12100         (dump_printf): Update for change to dump_printf_va.
12101         (dump_printf_loc): Likewise.
12102         (selftest::test_capture_of_dump_calls): Convert "stmt" from
12103         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
12104         with %T, %E, and %G.
12105         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
12106         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
12107         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
12108         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
12109         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
12110         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
12111         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
12112         within a dump_printf_loc call to "%wu".
12113         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
12114         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
12115         missing space after "=".
12116         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
12117         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
12118         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
12119         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
12120         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
12121         duplicate "vectorized" from message.
12123 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
12125         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
12126         polyNxK_t element's TYPE_STRING_FLAG.
12128 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
12130         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
12131         (they were unnamed before).  Fix comments.
12133 2018-08-17  Nathan Sidwell  <nathan@acm.org>
12135         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
12137 2018-08-17  Richard Biener  <rguenther@suse.de>
12139         PR tree-optimization/86841
12140         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
12142 2018-08-17  Martin Liska  <mliska@suse.cz>
12144         * common.opt: Remove Warn, Init and Report for options with
12145         Ignore/Deprecated flag. Warning is done automatically for
12146         Deprecated flags.
12147         * config/i386/i386.opt: Likewise.
12148         * config/ia64/ia64.opt: Likewise.
12149         * config/rs6000/rs6000.opt: Likewise.
12150         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
12151         Remove usage of flag_check_pointer_bounds.
12152         * lto-wrapper.c (merge_and_complain): Do not handle
12153         OPT_fcheck_pointer_bounds.
12154         (append_compiler_options): Likewise.
12155         * opt-functions.awk: Do not handle Deprecated.
12156         * optc-gen.awk: Check that Var, Report and Init are not
12157         used for an option with Ignore/Deprecated flag.
12158         * opts-common.c (decode_cmdline_option): Do not report
12159         CL_ERR_DEPRECATED.
12160         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
12161         options.
12162         * opts.h (struct cl_option): Remove cl_deprecated flag.
12163         (CL_ERR_DEPRECATED): Remove error enum value.
12165 2018-08-17  Richard Biener  <rguenther@suse.de>
12167         PR middle-end/86505
12168         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
12169         across a va-arg-pack using call adjust its return value accordingly.
12171 2018-08-16  Martin Sebor  <msebor@redhat.com>
12173         PR tree-optimization/86853
12174         * gimple-ssa-sprintf.c (struct format_result): Rename member.
12175         (struct fmtresult): Add member and initialize it in ctors.
12176         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
12177         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
12178         (format_directive): Set POSUNDER4K when MAYFAIL is set.
12179         (parse_directive): Handle %C same as %c.
12180         (sprintf_dom_walker::compute_format_length): Adjust.
12181         (is_call_safe): Adjust.
12183 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12185         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
12186         for determining how to count the elements.
12187         * builtins.h (c_strlen): Adjust prototype.
12188         * expr.c (string_constant): Add new parameter mem_size.
12189         Set *mem_size appropriately.
12190         * expr.h (string_constant): Adjust protoype.
12191         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
12192         * gimple-fold.h (get_range_strlen): Adjust prototype.
12193         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
12194         (format_string): Call get_string_length with eltsize.
12196 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
12198         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
12199         to emit the span, rather than setting it as the prefix.
12201 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
12203         * diagnostic-show-locus.c (layout::start_annotation_line): Add
12204         "margin_char" parameter, defaulting to space.  Use it in place
12205         of pp_space for the initial part of the margin.
12206         (layout::print_leading_fixits): Use '+' when filling the margin
12207         of line-insertion fix-it hints.
12209 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
12211         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
12212         Delete.
12214 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
12216         * config/rs6000/altivec.md: Don't set length attribute to the default
12217         value.
12218         * config/rs6000/darwin.md: Ditto.
12219         * config/rs6000/dfp.md: Ditto.
12220         * config/rs6000/htm.md: Ditto.
12221         * config/rs6000/rs6000.md: Ditto.
12222         * config/rs6000/sync.md: Ditto.
12223         * config/rs6000/vsx.md: Ditto.
12225 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
12227         * config/rs6000/altivec.md: Don't set length attribute to the default
12228         value, for branch instructions.
12229         * config/rs6000/darwin.md: Ditto.
12230         * config/rs6000/rs6000.md: Ditto.
12232 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
12234         * config/rs6000/rs6000.md (length): Always define as const_int 4.
12235         (unnamed conditional branch define_insn): Set length to 4 or 8
12236         depending on offset.
12237         (<bd>_<mode>): Similar, for alternative 0.
12238         (<bd>tf_<mode>): Ditto.
12240 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
12242         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
12244 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
12246         * doc/rtl.texi: Replace old RTX class names with new names.
12249 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
12251         * expmed.h (canonicalize_comparison): New declaration.
12252         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
12253         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
12254         * optabs.c (prepare_cmp_insn): Likewise.
12255         * rtl.h (unsigned_condition_p): New function which checks if a
12256         comparison operator is unsigned.
12258 2018-08-16  Nathan Sidwell  <nathan@acm.org>
12260         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
12261         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
12263 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
12265         PR target/84711
12266         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
12267         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
12268         (mov<mov>): ..this and enable unconditionally.
12270 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
12272         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
12274 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
12276         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
12277         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
12278         "Common" with "Target".
12280 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
12282         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
12283         * doc/invoke.texi (mmitigate-rop): Remove.
12284         * config/i386/i386.c: Do not include "regrename.h".
12285         (ix86_rop_should_change_byte_p, reg_encoded_number)
12286         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
12287         Remove.
12288         (ix86_reorg): Remove call to ix86_mitigate_rop.
12289         * config/i386/i386.md (attr "modrm_class"): Remove.
12290         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
12291         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
12292         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
12294 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
12296         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
12297         allow folding of mergeh() and mergel() for the float and double types.
12298         (fold_mergehl_helper): Rework to handle building a permute tree
12299         for float vectors.
12301 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
12303         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
12304         for TARGET_SSE.
12306 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
12308         * common.opt (fdiagnostics-show-labels): New option.
12309         * diagnostic-show-locus.c (class layout_range): Add field
12310         "m_label".
12311         (class layout): Add field "m_show_labels_p".
12312         (layout_range::layout_range): Add param "label" and use it to
12313         initialize m_label.
12314         (make_range): Pass in NULL for new "label" param of layout_range's
12315         ctor.
12316         (layout::layout): Initialize m_show_labels_p.
12317         (layout::maybe_add_location_range): Pass in loc_range->m_label
12318         when constructing layout_range instances.
12319         (struct line_label): New struct.
12320         (layout::print_any_labels): New member function.
12321         (layout::print_line): Call it if label-printing is enabled.
12322         (selftest::test_one_liner_labels): New test.
12323         (selftest::test_diagnostic_show_locus_one_liner): Call it.
12324         * diagnostic.c (diagnostic_initialize): Initialize
12325         context->show_labels_p.
12326         * diagnostic.h (struct diagnostic_context): Add field
12327         "show_labels_p".
12328         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
12329         -fno-diagnostics-show-labels.
12330         * dwarf2out.c (gen_producer_string): Add
12331         OPT_fdiagnostics_show_labels to the ignored options.
12332         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
12333         param.
12334         (gcc_rich_location::maybe_add_expr): Likewise.
12335         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
12336         label" param, defaulting to NULL.
12337         (gcc_rich_location::add_expr): Add "label" param.
12338         (gcc_rich_location::maybe_add_expr): Likewise.
12339         (class text_range_label): New class.
12340         (class range_label_for_type_mismatch): New class.
12341         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
12342         of format_warning_va.
12343         (fmtwarn_n): Likewise for new params of format_warning_n_va.
12344         * lto-wrapper.c (merge_and_complain): Add
12345         OPT_fdiagnostics_show_labels to the "pick one setting" options.
12346         (append_compiler_options): Likewise to the dropped options.
12347         (append_diag_options): Likewise to the passed-on options.
12348         * opts.c (common_handle_option): Handle the new option.
12349         * selftest-diagnostic.c
12350         (test_diagnostic_context::test_diagnostic_context): Enable
12351         show_labels_p.
12352         * substring-locations.c: Include "gcc-rich-location.h".
12353         (format_warning_n_va): Add "fmt_label" and "param_label" params
12354         and use them as appropriate.
12355         (format_warning_va): Add "fmt_label" and "param_label" params,
12356         passing them on to format_warning_n_va.
12357         (format_warning_at_substring): Likewise.
12358         (format_warning_at_substring_n): Likewise.
12359         * substring-locations.h (format_warning_va): Add "fmt_label" and
12360         "param_label" params.
12361         (format_warning_n_va): Likewise.
12362         (format_warning_at_substring): Likewise.
12363         (format_warning_at_substring_n): Likewise.
12364         * toplev.c (general_init): Initialize global_dc->show_labels_p.
12366 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
12368         PR testsuite/86519
12369         * builtins.c (expand_builtin_memcmp): Do not expand the call
12370         when overflow is detected.
12372 2018-08-15  Martin Sebor  <msebor@redhat.com>
12374         PR tree-optimization/71625
12375         * config/aarch64/aarch64-builtins.c
12376         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
12378 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
12380         * config/s390/s390.c (s390_reorg): Remove loop.
12382 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
12384         * config/darwin.c
12385          (darwin_function_switched_text_sections): Delete.
12386         * gcc/config/darwin.h
12387          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
12389 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
12391         PR target/81685
12392         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
12393         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
12394         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
12396 2018-08-15  Martin Liska  <mliska@suse.cz>
12398         PR tree-optimization/86925
12399         * predict.c (expr_expected_value_1): When taking
12400         later predictor, assign also probability.
12401         Use fold_build2_initializer_loc in order to fold
12402         the expression in -frounding-math.
12404 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
12406         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
12407         patterns.
12408         (expand_vec_perm_1): Try the new method.
12410 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
12412         PR target/86547
12413         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
12414         Check whether lra_live_max_point is 0 before dividing.
12416 2018-08-14  Martin Sebor  <msebor@redhat.com>
12418         PR tree-optimization/86650
12419         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
12420         (vrp_prop::check_mem_ref): Same.
12422 2018-08-13  Liu Hao <lh_mouse@126.com>
12424         * pretty-print.c (eat_esc_sequence): Swap the foreground and
12425         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
12426         and clear it thereafter, as it only works for DBCS.
12428 2018-08-13  Liu Hao <lh_mouse@126.com>
12430         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
12431         handle returned by _get_osf_handle().
12433 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
12435         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
12436         for folding vec_perm.
12438 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
12440         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
12441         Add support for gimple-folding of vec_pack() and vec_unpack()
12442         intrinsics.
12444 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
12446         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
12447         vec_xst variants to the list.
12448         (rs6000_gimple_fold_builtin): Add support for folding unaligned
12449         vector loads and stores.
12451 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
12453         * config.gcc (rs6000-ibm-aix4.x): Delete.
12454         (rs6000-ibm-aix5.1): Delete.
12455         (rs6000-ibm-aix5.2): Delete.
12456         (rs6000-ibm-aix5.3): Delete.
12457         * config/rs6000/aix43.h: Delete.
12458         * config/rs6000/aix51.h: Delete.
12459         * config/rs6000/aix52.h: Delete.
12460         * config/rs6000/t-aix43: Delete.
12462 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
12464         * config/s390/s390.c (s390_decompose_constant_pool_ref):
12465         New function.
12466         (s390_decompose_address): Factor out constant pool ref
12467         decomposition.
12469 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12471         * config/nds32/nds32-predicates.c
12472         (nds32_can_use_bclr_p): Change return type as bool.
12473         (nds32_can_use_bset_p): Ditto.
12474         (nds32_can_use_btgl_p): Ditto.
12475         (nds32_can_use_bitci_p): Ditto.
12476         * config/nds32/nds32-protos.h
12477         (nds32_can_use_bclr_p): Change declaration.
12478         (nds32_can_use_bset_p): Ditto.
12479         (nds32_can_use_btgl_p): Ditto.
12480         (nds32_can_use_bitci_p): Ditto.
12482 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12484         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
12485         Support -msched-prolog-epilog option.
12486         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
12488 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12490         * common/config/nds32/nds32-common.c
12491         (nds32_option_optimization_table): Enalbe -malways-align.
12493 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12495         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
12496         extra_headers.
12497         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
12498         OPT_misr_secure_ case.
12499         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
12500         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
12501         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
12502         secure attribute.
12503         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
12504         (nds32_isr_info): New field security_level.
12505         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
12506         * config/nds32/nds32.md (return_internal): Consider critical attribute.
12507         * config/nds32/nds32.opt (misr-secure): New option.
12508         * config/nds32/nds32_init.inc: New file.
12509         * config/nds32/nds32_isr.h: New file.
12511 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
12513         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
12514         Update comment for atomic instructions.
12515         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
12516         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
12517         Remove.
12518         (atomic_loaddi): Revise fence expansion to only emit fence prior to
12519         load for __ATOMIC_SEQ_CST model.
12520         (atomic_loaddi_1): Remove float register target.
12521         (atomic_storedi): Handle CONST_INT values.
12522         (atomic_storedi_1): Remove float register source.  Add special case
12523         for zero value.
12524         (memory_barrier): New expander and insn.
12526 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
12528         PR tree-optimization/86835
12529         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
12530         new_stmt after def_gsi, make sure to insert new_square_stmt after
12531         that stmt, not 2 stmts before it.
12533 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
12535         PR target/82418
12536         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
12537         instead of SWI48.
12539 2018-08-10  Martin Liska  <mliska@suse.cz>
12541         PR target/83610
12542         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
12543         function type.
12544         * builtins.c (expand_builtin_expect_with_probability):
12545         New function.
12546         (expand_builtin_expect_with_probability): New function.
12547         (build_builtin_expect_predicate): Add new argumnet probability
12548         for BUILT_IN_EXPECT_WITH_PROBABILITY.
12549         (fold_builtin_expect):
12550         (fold_builtin_2):
12551         (fold_builtin_3):
12552         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
12553         * builtins.h (fold_builtin_expect): Set new argument.
12554         * doc/extend.texi: Document __builtin_expect_with_probability.
12555         * doc/invoke.texi: Likewise.
12556         * gimple-fold.c (gimple_fold_call): Pass new argument.
12557         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
12558         also BUILT_IN_EXPECT_WITH_PROBABILITY.
12559         * predict.c (get_predictor_value): New function.
12560         (expr_expected_value): Add new argument probability. Assume
12561         that predictor and probability are always non-null.
12562         (expr_expected_value_1): Likewise.  For __builtin_expect and
12563         __builtin_expect_with_probability set probability.  Handle
12564         combination in binary expressions.
12565         (tree_predict_by_opcode): Simplify code by simply calling
12566         get_predictor_value.
12567         (pass_strip_predict_hints::execute): Add handling of
12568         BUILT_IN_EXPECT_WITH_PROBABILITY.
12569         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
12570         new predictor.
12571         * tree.h (DECL_BUILT_IN_P): New function.
12573 2018-08-10  Martin Liska  <mliska@suse.cz>
12575         PR tree-optimization/85799
12576         * passes.def: Add argument for pass_strip_predict_hints.
12577         * predict.c (class pass_strip_predict_hints): Add new argument
12578         early_p.
12579         (strip_predictor_early): New function.
12580         (pass_strip_predict_hints::execute): Call the function to
12581         strip predictors.
12582         (strip_predict_hints): New function.
12583         * predict.def: Fix comment.
12585 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
12587         * Makefile.in: Clarify which tm.texi to copy over to assert the
12588         right to grant a GFDL license for all.
12590 2018-08-09  Jeff Law  <law@redhat.com>
12592         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
12593         unused variable.
12595 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
12597         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
12598         prototype.
12600 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
12602         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
12603         reductions for variable-length vectors.
12605 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
12607         PR other/84889
12608         * common.opt (fdiagnostics-show-line-numbers): New option.
12609         * diagnostic-show-locus.c (class layout): Add fields
12610         "m_show_line_numbers_p" and "m_linenum_width";
12611         (num_digits): New function.
12612         (test_num_digits): New function.
12613         (layout::layout): Initialize new fields.  Update m_x_offset
12614         logic to handle any left margin.
12615         (layout::print_source_line): Print line number when requested.
12616         (layout::start_annotation_line): New member function.
12617         (layout::print_annotation_line): Call it.
12618         (layout::print_leading_fixits): Likewise.
12619         (layout::print_trailing_fixits): Likewise.  Update calls to
12620         move_to_column for new parameter.
12621         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
12622         it to potentially call start_annotation_line.
12623         (layout::show_ruler): Call start_annotation_line.
12624         (selftest::test_line_numbers_multiline_range): New selftest.
12625         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
12626         and selftest::test_line_numbers_multiline_range.
12627         * diagnostic.c (diagnostic_initialize): Initialize
12628         show_line_numbers_p.
12629         * diagnostic.h (struct diagnostic_context): Add field
12630         "show_line_numbers_p".
12631         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
12632         -fno-diagnostics-show-line-numbers.
12633         * dwarf2out.c (gen_producer_string): Add
12634         OPT_fdiagnostics_show_line_numbers to the ignored options.
12635         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
12636         one setting" options.
12637         (append_compiler_options): Likewise to the dropped options.
12638         (append_diag_options): Likewise to the passed-on options.
12639         * opts.c (common_handle_option): Handle the new option.
12640         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
12642 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12644         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
12645         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
12646         third argument of type "const signed char" to descriptions of
12647         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
12648         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
12649         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
12650         __builtin_bcdsub_ov functions.
12652 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
12654         PR tree-optimization/86858
12655         * tree-vect-loop.c (vect_is_simple_reduction): Restore
12656         flow_bb_inside_loop_p calls.
12658 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
12660         PR tree-optimization/86871
12661         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
12662         instead of gimple_assign_lhs.
12664 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
12666         PR target/86887
12667         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
12668         register constraint to operand 0.
12669         (add<mode>3_carryinC): Likewise.
12670         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
12672 2018-08-09  Martin Liska  <mliska@suse.cz>
12674         PR c/86895
12675         * common.opt: Remove extra line.
12677 2018-08-09  Martin Liska  <mliska@suse.cz>
12679         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
12680         at the end of a line, make first letter capital and end up
12681         a sentence with a dot.
12682         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
12683         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
12684         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
12685         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
12686         (PARAM_MAX_ISL_OPERATIONS): Likewise.
12687         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
12688         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
12689         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
12690         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
12691         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
12692         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
12693         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
12694         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
12695         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
12696         (PARAM_TREE_REASSOC_WIDTH): Likewise.
12697         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
12698         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
12699         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
12701 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
12703         PR target/84332
12704         * config/s390/s390.c (s390_option_override_internal): Reduce the
12705         stack-clash-protection-probe-interval param if it would be too big
12706         for z900.
12708 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
12710         PR target/46179
12711         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
12712         * config/m68k/m68k.c (handle_move_double): Don't call
12713         m68k_final_prescan_insn.
12714         (m68k_adjust_decorated_operand): Renamed from
12715         m68k_final_prescan_insn, remove first and third operand and
12716         simplify.
12717         (print_operand): Call it.
12718         (print_operand_address): Call it.
12720 2018-08-08  Nathan Sidwell  <nathan@acm.org>
12722         * diagnostic.c (diagnostic_report_current_module): Use
12723         linemap_included_from & linemap_included_from_linemap.
12725 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
12727         * config/aarch64/aarch64-cores.def: Add phecda core.
12728         * config/aarch64/aarch64-tune.md: Regenerate.
12729         * doc/invoke.texi: Add phecda core.
12731 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12733         PR target/85295
12734         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
12735         definitions.
12736         * config/s390/s390.md ("movti"): Add more alternatives for
12737         constant to GPR copies.
12739 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12741         * config/s390/s390.c: Fix whitespace damage throughout the file.
12742         * config/s390/s390.h: Likewise.
12743         * config/s390/tpf.h: Likewise.
12745 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12747         * config/s390/s390.c (s390_loadrelative_operand_p):
12748         Remove TARGET_CPU_ZARCH usages.
12749         (s390_rtx_costs): Likewise.
12750         (s390_legitimate_constant_p): Likewise.
12751         (s390_cannot_force_const_mem): Likewise.
12752         (legitimate_reload_constant_p): Likewise.
12753         (s390_preferred_reload_class): Likewise.
12754         (legitimize_pic_address): Likewise.
12755         (legitimize_tls_address): Likewise.
12756         (s390_split_branches): Removed.
12757         (s390_add_execute): Removed.
12758         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
12759         (s390_mainpool_start): Likewise.
12760         (s390_mainpool_finish): Likewise.
12761         (s390_mainpool_cancel): Removed.
12762         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
12763         (s390_chunkify_cancel): Likewise.
12764         (s390_return_addr_rtx): Likewise.
12765         (s390_register_info): Remove split_branches_pending_p uages.
12766         (s390_optimize_register_info): Likewise.
12767         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
12768         split_branches_pending_p usages.
12769         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
12770         (s390_load_got): Likewise.
12771         (s390_expand_split_stack_prologue): Likewise.
12772         (output_asm_nops): Likewise.
12773         (s390_function_profiler): Likewise.
12774         (s390_emit_call): Likewise.
12775         (s390_conditional_register_usage): Likewise.
12776         (s390_optimize_prologue): Likewise.
12777         (s390_reorg): Remove TARGET_CPU_ZARCH and
12778         split_branches_pending_p usages.
12779         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
12780         usages.
12781         (s390_output_indirect_thunk_function): Likewise.
12782         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
12783         (TARGET_CPU_ZARCH_P): Removed.
12784         (struct machine_function): Remove split_branches_pending_p.
12785         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
12787 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12789         * common/config/s390/s390-common.c (processor_flags_table):
12790         Remove flags.
12791         * config.gcc: Remove with_arch/with_tune support.
12792         * config/s390/2064.md: Remove cpu attribute comparisons.
12793         * config/s390/driver-native.c (s390_host_detect_local_cpu):
12794         Remove MTN.
12795         * config/s390/linux.h (ASM_SPEC):
12796         Remove -march support.
12797         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
12798         Use a table to get an arch level.
12799         * config/s390/s390-opts.h (enum processor_type):
12800         Remove enum values.
12801         * config/s390/s390.c
12802         (processor_table): Remove entries, add arch_level values.
12803         (s390_issue_rate): Remove cases.
12804         (s390_option_override): Adjust
12805         s390_option_override_internal() call.
12806         (s390_option_override_internal): Remove deprecation warning.
12807         (s390_valid_target_attribute_tree): Adjust
12808         s390_option_override_internal() call.
12809         * config/s390/s390.h (struct s390_processor):
12810         Share with s390-c.c, add arch_level field.
12811         * config/s390/s390.md:
12812         Remove occurrences in cpu attribute.
12813         * config/s390/s390.opt: Remove -march/-mtune support.
12814         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
12815         * doc/invoke.texi: Remove deprecation warning.
12817 2018-08-08  Luis Machado  <luis.machado@linaro.org>
12819         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
12820         global.
12821         (qdf24xx_tunings): Set vector cost structure to
12822         qdf24xx_vector_cost.
12824         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
12825         <register_sextend>: Set to 3.
12827 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
12829         PR target/86838
12830         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
12831         * config/aarch64/aarch64-simd.md
12832         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
12833         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
12834         (aarch64_frecpx<mode>): ...this new pattern.
12835         * config/aarch64/aarch64-simd-builtins.def: Remove comment
12836         about aarch64_frecp<FRECP:frecp_suffix><mode>.
12838 2018-08-07  Martin Liska  <mliska@suse.cz>
12840         PR middle-end/83023
12841         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
12842         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
12843         * predict.def (PRED_MALLOC_NONNULL): New predictor.
12844         * doc/extend.texi: Document that malloc attribute adds
12845         hit to compiler.
12847 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
12849         PR target/86785
12850         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12851         Define to speculation_safe_value_not_needed.
12853 2018-08-06  Jeff Law  <law@redhat.com>
12855         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
12856         the vr_values instance to cprop_into_stmt.
12857         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
12858         (cprop_operand): Also query EVRP to determine if OP is a constant.
12860 2018-08-06  Nathan Sidwell  <nathan@acm.org>
12862         * diagnostic.c (diagnostic_report_current_module): Reroll
12863         included-at loop.  Translate text.
12865 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
12867         * function-tests.c (selftest::test_expansion_to_rtl): Call
12868         free_after_compilation.
12870 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12872         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
12874 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
12876         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
12877         loops with memory block operations from getting unrolled.
12879 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
12881         PR target/86807
12882         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12883         Define to speculation_safe_value_not_needed.
12885 2018-08-06  Jeff Law  <law@redhat.com>
12887         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
12888         assert.
12890 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12892         PR target/86662
12893         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
12894         with all enabled __intN types.
12896         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
12898 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12900         * alias.c (record_set): Check for clobber high.
12901         * cfgexpand.c (expand_gimple_stmt): Likewise.
12902         * combine-stack-adj.c (single_set_for_csa): Likewise.
12903         * combine.c (find_single_use_1): Likewise.
12904         (set_nonzero_bits_and_sign_copies): Likewise.
12905         (get_combine_src_dest): Likewise.
12906         (is_parallel_of_n_reg_sets): Likewise.
12907         (try_combine): Likewise.
12908         (record_dead_and_set_regs_1): Likewise.
12909         (reg_dead_at_p_1): Likewise.
12910         (reg_dead_at_p): Likewise.
12911         * dce.c (deletable_insn_p): Likewise.
12912         (mark_nonreg_stores_1): Likewise.
12913         (mark_nonreg_stores_2): Likewise.
12914         * df-scan.c (df_find_hard_reg_defs): Likewise.
12915         (df_uses_record): Likewise.
12916         (df_get_call_refs): Likewise.
12917         * dwarf2out.c (mem_loc_descriptor): Likewise.
12918         * haifa-sched.c (haifa_classify_rtx): Likewise.
12919         * ira-build.c (create_insn_allocnos): Likewise.
12920         * ira-costs.c (scan_one_insn): Likewise.
12921         * ira.c (equiv_init_movable_p): Likewise.
12922         (rtx_moveable_p): Likewise.
12923         (interesting_dest_for_shprep): Likewise.
12924         * jump.c (mark_jump_label_1): Likewise.
12925         * postreload-gcse.c (record_opr_changes): Likewise.
12926         * postreload.c (reload_cse_simplify): Likewise.
12927         (struct reg_use): Add source expr.
12928         (reload_combine): Check for clobber high.
12929         (reload_combine_note_use): Likewise.
12930         (reload_cse_move2add): Likewise.
12931         (move2add_note_store): Likewise.
12932         * print-rtl.c (print_pattern): Likewise.
12933         * recog.c (decode_asm_operands): Likewise.
12934         (store_data_bypass_p): Likewise.
12935         (if_test_bypass_p): Likewise.
12936         * regcprop.c (kill_clobbered_value): Likewise.
12937         (kill_set_value): Likewise.
12938         * reginfo.c (reg_scan_mark_refs): Likewise.
12939         * reload1.c (maybe_fix_stack_asms): Likewise.
12940         (eliminate_regs_1): Likewise.
12941         (elimination_effects): Likewise.
12942         (mark_not_eliminable): Likewise.
12943         (scan_paradoxical_subregs): Likewise.
12944         (forget_old_reloads_1): Likewise.
12945         * reorg.c (find_end_label): Likewise.
12946         (try_merge_delay_insns): Likewise.
12947         (redundant_insn): Likewise.
12948         (own_thread_p): Likewise.
12949         (fill_simple_delay_slots): Likewise.
12950         (fill_slots_from_thread): Likewise.
12951         (dbr_schedule): Likewise.
12952         * resource.c (update_live_status): Likewise.
12953         (mark_referenced_resources): Likewise.
12954         (mark_set_resources): Likewise.
12955         * rtl.c (copy_rtx): Likewise.
12956         * rtlanal.c (reg_referenced_p): Likewise.
12957         (single_set_2): Likewise.
12958         (noop_move_p): Likewise.
12959         (note_stores): Likewise.
12960         * sched-deps.c (sched_analyze_reg): Likewise.
12961         (sched_analyze_insn): Likewise.
12963 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12965         * cse.c (invalidate_reg): New function extracted from...
12966         (invalidate): ...here.
12967         (canonicalize_insn): Check for clobber high.
12968         (invalidate_from_clobbers): invalidate clobber highs.
12969         (invalidate_from_sets_and_clobbers): Likewise.
12970         (count_reg_usage): Check for clobber high.
12971         (insn_live_p): Likewise.
12972         * cselib.c (cselib_expand_value_rtx_1):Likewise.
12973         (cselib_invalidate_regno): Check for clobber in setter.
12974         (cselib_invalidate_rtx): Pass through setter.
12975         (cselib_invalidate_rtx_note_stores):
12976         (cselib_process_insn): Check for clobber high.
12977         * cselib.h (cselib_invalidate_rtx): Add operand.
12979 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12981         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
12982         (mark_not_eliminable): Likewise.
12983         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
12984         * lra-lives.c (process_bb_lives): Check for clobber high.
12985         * lra.c (new_insn_reg): Remember clobber highs.
12986         (collect_non_operand_hard_regs): Check for clobber high.
12987         (lra_set_insn_recog_data): Likewise.
12988         (add_regs_to_insn_regno_info): Likewise.
12989         (lra_update_insn_regno_info): Likewise.
12991 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12993         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
12994         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
12996 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12998         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
12999         (copy_insn_1): Likewise.
13000         (gen_hard_reg_clobber_high): New gen function.
13001         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
13002         * genemit.c (gen_exp): Likewise.
13003         (gen_emit_seq): Pass through info.
13004         (gen_insn): Check for CLOBBER_HIGH.
13005         (gen_expand): Pass through info.
13006         (gen_split): Likewise.
13007         (output_add_clobbers): Likewise.
13008         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
13009         (remove_clobbers): Likewise.
13010         * rtl.h (gen_hard_reg_clobber_high): New declaration.
13012 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
13014         * doc/rtl.texi (clobber_high): Add.
13015         (parallel): Add in clobber high
13016         * rtl.c (rtl_check_failed_code3): Add function.
13017         * rtl.def (CLOBBER_HIGH): Add expression.
13018         * rtl.h (RTL_CHECKC3): Add macro.
13019         (rtl_check_failed_code3): Add declaration.
13020         (XC3EXP): Add macro.
13022 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
13024         PR target/86386
13025         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
13026         cfun->machine->max_used_stack_alignment if needed.
13028 2018-08-04  Martin Sebor  <msebor@redhat.com>
13030         PR tree-optimization/86571
13031         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
13032         NaN output to 4.
13034 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
13036         PR target/86799
13037         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13038         Define.
13040 2018-08-03  Jeff Law  <law@redhat.com>
13042         PR target/86795
13043         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13044         Define to speculation_safe_value_not_needed.
13046 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
13048         * doc/gcov.texi (-x): Remove duplicate "to".
13049         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
13050         (-Wif-not-aligned): Remove duplicate "is".
13051         (-flto): Remove duplicate "the".
13052         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
13053         duplicate "v5.00.b".
13054         (MSP430 Options): Remove duplicate "and" from the description
13055         of "-mgprel-sec=regexp".
13056         (x86 Options): Remove duplicate copies of "vmldLog102" and
13057         vmlsLog104 from description of "-mveclibabi=type".
13059 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
13061         * internal-fn.h (first_commutative_argument): Declare.
13062         * internal-fn.c (first_commutative_argument): New function.
13063         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
13064         restrictions for pattern statements.  Use first_commutative_argument
13065         to look for commutative operands in calls to internal functions.
13067 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
13069         * Makefile.in (wide-int-range.o): New.
13070         * tree-vrp.c: Move all the wide_int_* functions to...
13071         * wide-int-range.cc: ...here.
13072         * tree-vrp.h: Move all the wide_int_* prototypes to...
13073         * wide-int-range.h: ...here.
13075 2018-08-03  Tom de Vries  <tdevries@suse.de>
13077         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
13078         UI_NONE.
13079         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
13080         * except.c (output_function_exception_table): Do early exit if
13081         targetm_common.except_unwind_info (&global_options) == UI_NONE.
13083 2018-08-03  Martin Liska  <mliska@suse.cz>
13085         * predict.c (dump_prediction): Change to 2 digits
13086         in fraction part.
13088 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
13090         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
13091         neon_dup_q to...
13092         (falkor_am_1_gtov_gtov): ... a new insn reservation.
13094 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
13096         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
13097         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
13098         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
13099         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
13100         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
13102 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
13104         * diagnostic-show-locus.c (diagnostic_show_locus): Use
13105         pp_take_prefix when saving the existing prefix.
13106         * diagnostic.c (diagnostic_append_note): Likewise.
13107         * langhooks.c (lhd_print_error_function): Likewise.
13108         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
13109         param's type.  Free the existing prefix.
13110         (pp_take_prefix): New function.
13111         (pretty_printer::pretty_printer): Drop the prefix parameter.
13112         Rename the length parameter to match the comment.
13113         (pretty_printer::~pretty_printer): Free the prefix.
13114         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
13115         parameter.
13116         (struct pretty_printer): Drop the "const" from "prefix" field's
13117         type and clarify memory management.
13118         (pp_set_prefix): Drop the "const" from the 2nd param.
13119         (pp_take_prefix): New decl.
13121 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
13123         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
13124         for word_mode_ok here instead of passing as argument.
13125         (expand_block_compare): Change select_block_compare_mode() call.
13126         (expand_strncmp_gpr_sequence): New function.
13127         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
13129 2018-08-02  Jeff Law  <law@redhat.com>
13131         PR target/86790
13132         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13133         Define to speculation_safe_value_not_needed.
13135         PR target/86784
13136         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13137         Define to speculation_safe_value_not_needed.
13139 2018-08-02  Tom de Vries  <tdevries@suse.de>
13141         PR target/86660
13142         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
13143         function.  Return UI_TARGET unconditionally.
13144         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
13145         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
13147 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
13149         * genemit.c (print_overload_test): Fix typo.
13151 2018-08-02  Richard Biener  <rguenther@suse.de>
13153         PR tree-optimization/86816
13154         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
13155         which checks for value availability before querying it.
13156         (gvn_uses_equal): Use it.
13157         (same_succ_hash): Likewise.
13158         (gimple_equal_p): Likewise.
13160 2018-08-02  Nick Clifton  <nickc@redhat.com>
13162         PR target/86813
13163         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13164         Define to speculation_safe_value_not_needed.
13166         PR target/86810
13167         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13168         Define to speculation_safe_value_not_needed.
13170         PR target/86810
13171         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13172         Define to speculation_safe_value_not_needed.
13174         PR target/86803
13175         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13176         Define to speculation_safe_value_not_needed.
13178         PR target/86797
13179         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13180         Define to speculation_safe_value_not_needed.
13182         PR target/86791
13183         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13184         Define to speculation_safe_value_not_needed.
13186         PR target/86789
13187         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13188         Define to speculation_safe_value_not_needed.
13190         PR target/86787
13191         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13192         Define to speculation_safe_value_not_needed.
13194         PR target/86782
13195         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
13196         speculation_safe_value_not_needed.
13198         PR target/86781
13199         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
13200         to speculation_safe_value_not_needed.
13202 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
13204         * doc/md.texi: Expand the documentation of instruction names
13205         to mention port-local uses.  Document '@' in pattern names.
13206         * read-md.h (overloaded_instance, overloaded_name): New structs.
13207         (mapping): Declare.
13208         (md_reader::handle_overloaded_name): New member function.
13209         (md_reader::get_overloads): Likewise.
13210         (md_reader::m_first_overload): New member variable.
13211         (md_reader::m_next_overload_ptr): Likewise.
13212         (md_reader::m_overloads_htab): Likewise.
13213         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
13214         m_next_overload_ptr and m_overloads_htab.
13215         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
13216         (get_mode_token, get_code_token, get_int_token): New functions.
13217         (map_attr_string): Add an optional argument that passes back
13218         the associated iterator.
13219         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
13220         (md_reader::handle_overloaded_name, add_overload_instance): New
13221         functions.
13222         (apply_iterators): Handle '@' names.  Report an error if '@'
13223         is used without iterators.
13224         (initialize_iterators): Initialize the new iterator_group fields.
13225         * genopinit.c (handle_overloaded_code_for)
13226         (handle_overloaded_gen): New functions.
13227         (main): Use them to print declarations of maybe_code_for_* and
13228         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
13229         * genemit.c (print_overload_arguments, print_overload_test)
13230         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
13231         (main): Use it to print definitions of maybe_code_for_* and
13232         maybe_gen_* functions.
13233         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
13234         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
13235         instead of explicit mode checks.
13236         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
13237         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
13238         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
13239         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
13240         (aarch64_expand_compare_and_swap): Likewise
13241         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
13242         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
13243         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
13244         (aarch64_constant_pool_reload_icode): Delete.
13245         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
13246         instead of aarch64_constant_pool_reload_icode.  Use
13247         code_for_aarch64_reload_mov instead of explicit mode checks.
13248         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
13249         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
13250         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
13251         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
13252         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
13253         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
13254         (aarch64_atomic_load_op_code): Delete.
13255         (aarch64_emit_atomic_load_op): Likewise.
13256         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
13257         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
13258         instead of aarch64_emit_atomic_load_op.
13259         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
13260         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
13261         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
13262         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
13263         character before the pattern name.
13264         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
13265         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
13266         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
13267         (aarch64_frecps<mode>): Likewise.
13268         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
13269         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
13270         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
13271         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
13272         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
13274 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
13276         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
13277         Allow HFmode constants if TARGET_FP_F16INST.
13279 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
13281         PR target/86014
13282         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
13283         No longer check last store for clobber of address register.
13285 2018-08-02  Martin Liska  <mliska@suse.cz>
13287         PR gcov-profile/86817
13288         * gcov.c (process_all_functions): New function.
13289         (main): Call it.
13290         (process_file): Move functions processing to
13291         process_all_functions.
13293 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
13295         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
13296         "const" to the "gimple *" and "rtx_insn *" parameters.
13297         * dumpfile.h (dump_user_location_t::dump_user_location_t):
13298         Likewise.
13299         (dump_location_t::dump_location_t): Likewise.
13301 2018-08-01  Martin Sebor  <msebor@redhat.com>
13303         PR tree-optimization/86650
13304         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
13305         rather than a "gcall *".  Directly pass the data of interest
13306         to percent_K_format, rather than building a temporary CALL_EXPR
13307         to hold it.
13308         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
13309         (gimple_fold_builtin_strncat): Adjust.
13310         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
13311         gcall* argument with gimple*.
13312         * gimple-ssa-warn-restrict.c (check_call): Same.
13313         (wrestrict_dom_walker::before_dom_children): Same.
13314         (builtin_access::builtin_access): Same.
13315         (check_bounds_or_overlap): Same
13316         (maybe_diag_overlap): Same.
13317         (maybe_diag_offset_bounds): Same.
13318         * tree-diagnostic.c (default_tree_printer): Move usage of
13319         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
13320         to this callsite.
13321         * tree-pretty-print.c (percent_K_format): Add argument.
13322         * tree-pretty-print.h: Add argument.
13323         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
13324         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
13325         (maybe_diag_stxncpy_trunc): Same.
13326         (handle_builtin_stxncpy): Same.
13327         (handle_builtin_strcat): Same.
13329 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13331         * match.pd: Optimise pointer range checks.
13333 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13335         PR tree-optimization/86758
13336         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
13337         to remove pattern statements.
13339 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13341         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
13342         result of dfs_enumerate_from when constructing stmt_vec_infos,
13343         instead of additionally calling get_loop_body.
13345 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13347         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
13348         parameter.
13349         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
13350         When creating an iv, assert that the step is not known to be zero.
13351         (vect_setup_realignment): Update call accordingly.
13352         * tree-vect-stmts.c (vectorizable_store): Likewise.
13353         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
13355 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13357         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
13358         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
13359         (vectorizable_reduction): Likewise.
13360         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13361         (vect_detect_hybrid_slp_stmts): Likewise.
13362         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13364 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
13366         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
13367         (wide_int_set_zero_nonzero_bits): ...this.
13368         (zero_nonzero_bits_from_vr): Rename to...
13369         (vrp_set_zero_nonzero_bits): ...this.
13370         (extract_range_from_multiplicative_op_1): Abstract wide int
13371         code...
13372         (wide_int_range_multiplicative_op): ...here.
13373         (extract_range_from_binary_expr_1): Extract wide int binary
13374         operations into their own functions.
13375         (wide_int_range_lshift): New.
13376         (wide_int_range_can_optimize_bit_op): New.
13377         (wide_int_range_shift_undefined_p): New.
13378         (wide_int_range_bit_xor): New.
13379         (wide_int_range_bit_ior): New.
13380         (wide_int_range_bit_and): New.
13381         (wide_int_range_trunc_mod): New.
13382         (extract_range_into_wide_ints): New.
13383         (vrp_shift_undefined_p): New.
13384         (extract_range_from_multiplicative_op): New.
13385         (vrp_can_optimize_bit_op): New.
13386         * tree-vrp.h (value_range::dump): New.
13387         (wide_int_range_multiplicative_op): New.
13388         (wide_int_range_lshift):New.
13389         (wide_int_range_shift_undefined_p): New.
13390         (wide_int_range_bit_xor): New.
13391         (wide_int_range_bit_ior): New.
13392         (wide_int_range_bit_and): New.
13393         (wide_int_range_trunc_mod): New.
13394         (zero_nonzero_bits_from_bounds): Rename to...
13395         (wide_int_set_zero_nonzero_bits): ...this.
13396         (zero_nonzero_bits_from_vr): Rename to...
13397         (vrp_set_zero_nonzero_bits): ...this.
13398         (range_easy_mask_min_max): Rename to...
13399         (wide_int_range_can_optimize_bit_op): this.
13400         * vr-values.c (simplify_bit_ops_using_ranges): Rename
13401         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
13403 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13405         * tree-vectorizer.h (vect_orig_stmt): New function.
13406         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
13407         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
13408         (vect_create_epilog_for_reduction): Likewise.
13409         (vectorizable_live_operation): Likewise.
13410         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
13411         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
13412         * tree-vect-stmts.c (vectorizable_call): Likewise.
13413         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
13415 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13417         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
13418         argument.
13419         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
13420         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
13421         (vect_transform_loop): Likewise.
13422         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
13424 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13426         * tree-vectorizer.h (vect_schedule_slp): Return void.
13427         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
13428         (vect_schedule_slp): Likewise.
13430 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13432         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
13433         argument.
13434         (vect_transform_loop): Update calls accordingly.  Schedule SLP
13435         instances before the main loop, if any exist.
13437 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
13439         PR tree-optimization/86749
13440         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
13441         If the lhs is used in a COND_EXPR, check that it is being used
13442         as the "then" or "else" value.
13444 2018-08-01  Tom de Vries  <tdevries@suse.de>
13446         PR target/86800
13447         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
13448         speculation_safe_value_not_needed.
13450 2018-08-01  Richard Biener  <rguenther@suse.de>
13452         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
13453         as base and offset.
13455 2018-08-01  Martin Liska  <mliska@suse.cz>
13457         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
13458         format how successful transformation is dumped.
13459         (gimple_mod_pow2_value_transform): Likewise.
13460         (gimple_mod_subtract_transform): Likewise.
13461         (gimple_stringops_transform): Likewise.
13463 2018-08-01  Martin Liska  <mliska@suse.cz>
13465         PR value-prof/35543
13466         * value-prof.c (interesting_stringop_to_profile_p):
13467         Simplify the code and add BUILT_IN_MEMMOVE.
13468         (gimple_stringops_transform): Likewise.
13470 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
13472         * config/aarch64/aarch64-simd.md
13473         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
13474         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
13475         use GPI iterator instead of SI mode.
13477 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
13479         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
13480         rs6000_speculation_barrier.
13481         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
13482         new barrier pattern name.
13484 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
13486         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
13487         (speculation_barrier): New insn.
13489 2018-08-01  Richard Biener  <rguenther@suse.de>
13491         PR bootstrap/86724
13492         * graphite.h: Include isl/id.h and isl/space.h to allow build
13493         with ISL 0.20.
13495 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
13497         PR target/86651
13498         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13499         mode for COFF targets.
13500         * defaults.h (TARGET_COFF): Define.
13501         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
13502         TARGET_COFF): Define.
13503         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
13504         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
13505         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
13507 2018-07-31  Alexandre Oliva <oliva@adacore.com>
13509         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
13510         * gimple-streamer-out.c (output_bb): Save it.
13511         * lto-streamer-in.c (input_struct_function_base): Restore
13512         instance discriminator if available.  Create map on demand.
13513         * lto-streamer-out.c (output_struct_function_base): Save it if
13514         available.
13515         * final.c (decl_to_instance_map): Document LTO strategy.
13517 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
13518             Olivier Hainque  <hainque@adacore.com>
13520         * debug.h (decl_to_instance_map_t): New type.
13521         (decl_to_instance_map): Declare.
13522         (maybe_create_decl_to_instance_map): New inline function.
13523         * final.c (bb_discriminator, last_bb_discriminator): New statics,
13524         to track basic block discriminators.
13525         (final_start_function_1): Initialize them.
13526         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
13527         bb_discriminator.
13528         (decl_to_instance_map): New variable.
13529         (map_decl_to_instance, maybe_set_discriminator): New functions.
13530         (notice_source_line): Set discriminator.
13532 2018-07-31  Ian Lance Taylor  <iant@golang.org>
13534         * targhooks.c (default_have_speculation_safe_value): Add
13535         ATTRIBUTE_UNUSED.
13537 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
13539         * dump-context.h: Include "pretty-print.h".
13540         (dump_context::refresh_dumps_are_enabled): New decl.
13541         (dump_context::emit_item): New decl.
13542         (class dump_context): Add fields "m_test_pp" and
13543         "m_test_pp_flags".
13544         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
13545         (temp_dump_context::get_dumped_text): New decl.
13546         (class temp_dump_context): Add field "m_pp".
13547         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
13548         (dump_context::refresh_dumps_are_enabled): ...and add a test for
13549         m_test_pp.
13550         (set_dump_file): Update for above change.
13551         (set_alt_dump_file): Likewise.
13552         (dump_loc): New overload, taking a pretty_printer *.
13553         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
13554         to any test pretty-printer.
13555         (make_item_for_dump_gimple_stmt): New function, adapted from
13556         optinfo::add_gimple_stmt.
13557         (dump_context::dump_gimple_stmt): Call it, and use the result,
13558         eliminating the direct usage of dump_file and alt_dump_file in
13559         favor of indirectly using them via emit_item.
13560         (make_item_for_dump_gimple_expr): New function, adapted from
13561         optinfo::add_gimple_expr.
13562         (dump_context::dump_gimple_expr): Call it, and use the result,
13563         eliminating the direct usage of dump_file and alt_dump_file in
13564         favor of indirectly using them via emit_item.
13565         (make_item_for_dump_generic_expr): New function, adapted from
13566         optinfo::add_tree.
13567         (dump_context::dump_generic_expr): Call it, and use the result,
13568         eliminating the direct usage of dump_file and alt_dump_file in
13569         favor of indirectly using them via emit_item.
13570         (make_item_for_dump_printf_va): New function, adapted from
13571         optinfo::add_printf_va.
13572         (make_item_for_dump_printf): New function.
13573         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
13574         and use the result, eliminating the direct usage of dump_file and
13575         alt_dump_file in favor of indirectly using them via emit_item.
13576         (make_item_for_dump_dec): New function.
13577         (dump_context::dump_dec): Call it, and use the result,
13578         eliminating the direct usage of dump_file and alt_dump_file in
13579         favor of indirectly using them via emit_item.
13580         (make_item_for_dump_symtab_node): New function, adapted from
13581         optinfo::add_symtab_node.
13582         (dump_context::dump_symtab_node): Call it, and use the result,
13583         eliminating the direct usage of dump_file and alt_dump_file in
13584         favor of indirectly using them via emit_item.
13585         (dump_context::begin_scope): Reimplement, avoiding direct usage
13586         of dump_file and alt_dump_file in favor of indirectly using them
13587         via emit_item.
13588         (dump_context::emit_item): New member function.
13589         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
13590         Set up test pretty-printer on the underlying context.  Call
13591         refresh_dumps_are_enabled.
13592         (temp_dump_context::~temp_dump_context): Call
13593         refresh_dumps_are_enabled.
13594         (temp_dump_context::get_dumped_text): New member function.
13595         (selftest::verify_dumped_text): New function.
13596         (ASSERT_DUMPED_TEXT_EQ): New macro.
13597         (selftest::test_capture_of_dump_calls): Run all tests twice, with
13598         and then without optinfo enabled.  Add uses of
13599         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
13600         * dumpfile.h: Update comment for the dump_* API.
13601         * optinfo-emit-json.cc
13602         (selftest::test_building_json_from_dump_calls): Update for new
13603         param for temp_dump_context ctor.
13604         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
13605         and "m_owned" field.
13606         (optinfo_item::~optinfo_item): Likewise.
13607         (optinfo::add_item): New member function.
13608         (optinfo::emit): Update comment.
13609         (optinfo::add_string): Delete.
13610         (optinfo::add_printf): Delete.
13611         (optinfo::add_printf_va): Delete.
13612         (optinfo::add_gimple_stmt): Delete.
13613         (optinfo::add_gimple_expr): Delete.
13614         (optinfo::add_tree): Delete.
13615         (optinfo::add_symtab_node): Delete.
13616         (optinfo::add_dec): Delete.
13617         * optinfo.h (class dump_context): New forward decl.
13618         (optinfo::add_item): New decl.
13619         (optinfo::add_string): Delete.
13620         (optinfo::add_printf): Delete.
13621         (optinfo::add_printf_va): Delete.
13622         (optinfo::add_gimple_stmt): Delete.
13623         (optinfo::add_gimple_expr): Delete.
13624         (optinfo::add_tree): Delete.
13625         (optinfo::add_symtab_node): Delete.
13626         (optinfo::add_dec): Delete.
13627         (optinfo::add_poly_int): Delete.
13628         (optinfo_item::optinfo_item): Remove "owned" param.
13629         (class optinfo_item): Remove field "m_owned".
13631 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13633         PR middle-end/86705
13634         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
13635         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
13636         requested variable alignment.
13637         (expand_one_ssa_partition): Likewise.
13638         (expand_one_var): Likewise.
13640 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13642         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
13643         to speculation_safe_value_not_needed.
13645 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13647         * targhooks.h (speculation_safe_value_not_needed): New prototype.
13648         * targhooks.c (speculation_safe_value_not_needed): New function.
13649         * target.def (have_speculation_safe_value): Update documentation.
13650         * doc/tm.texi: Regenerated.
13652 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13654         * config/aarch64/iterators.md (ALLI_TI): New iterator.
13655         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
13656         expand.
13657         (despeculate_copy<ALLI:mode>_insn): New insn.
13658         (despeculate_copyti_insn): New insn.
13659         (despeculate_simple<ALLI:mode>): New insn
13660         (despeculate_simpleti): New insn.
13661         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
13662         function.
13663         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
13664         aarch64_speculation_safe_value.
13665         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
13667 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13669         * config/aarch64/aarch64-speculation.cc: New file.
13670         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
13671         before pass_reorder_blocks.
13672         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
13673         prototype.
13674         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
13675         X14 and X15 when tracking speculation.
13676         * config/aarch64/aarch64.md (register name constants): Add
13677         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
13678         (unspec): Add UNSPEC_SPECULATION_TRACKER.
13679         (speculation_barrier): New insn attribute.
13680         (cmp<mode>): Allow SP in comparisons.
13681         (speculation_tracker): New insn.
13682         (speculation_barrier): Add speculation_barrier attribute.
13683         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
13684         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
13685         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
13687 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13689         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
13690         aarch64_track_speculation is true.
13691         (tb<optab><mode>1): Likewise.
13692         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
13693         generate CB[N]Z when tracking speculation.
13694         (aarch64_split_compare_and_swap): Likewise.
13695         (aarch64_split_atomic_op): Likewise.
13697 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13699         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
13701 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13703         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
13704         (speculation_barrier): New insn.
13706 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13708         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
13709         * config/arm/arm.md (speculation_barrier): New expand.
13710         (speculation_barrier_insn): New pattern.
13712 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13714         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
13715         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
13716         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
13717         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
13718         list.
13719         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
13720         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
13721         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
13722         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
13723         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
13724         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
13725         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
13726         * builtins.c (expand_speculation_safe_value): New function.
13727         (expand_builtin): Call it.
13728         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
13729         * doc/extend.texi: Document __builtin_speculation_safe_value.
13730         * doc/md.texi: Document "speculation_barrier" pattern.
13731         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
13732         TARGET_HAVE_SPECULATION_SAFE_VALUE.
13733         * doc/tm.texi: Regenerated.
13734         * target.def (have_speculation_safe_value, speculation_safe_value): New
13735         hooks.
13736         * targhooks.c (default_have_speculation_safe_value): New function.
13737         (default_speculation_safe_value): New function.
13738         * targhooks.h (default_have_speculation_safe_value): Add prototype.
13739         (default_speculation_safe_value): Add prototype.
13741 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
13743         * dump-context.h (dump_context::dump_loc): New decl.
13744         * dumpfile.c (dump_context::dump_loc): New member function.
13745         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
13746         and dump_gimple_stmt.
13747         (dump_context::dump_gimple_expr_loc): Likewise, using
13748         dump_gimple_expr.
13749         (dump_context::dump_generic_expr_loc): Likewise, using
13750         dump_generic_expr.
13751         (dump_context::dump_printf_loc_va): Likewise, using
13752         dump_printf_va.
13753         (dump_context::begin_scope): Explicitly using the global function
13754         "dump_loc", rather than the member function.
13756 2018-07-31  Martin Sebor  <msebor@redhat.com>
13758         PR tree-optimization/86741
13759         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
13761 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
13763         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
13764         depend on whether prefetch instructions will be emitted or not.
13765         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
13766         will be emitted or not.
13767         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
13768         (TARGET_SETMEM_PFD): New macros.
13770 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13772         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
13773         (NULL_STMT_VEC_INFO): Delete.
13774         (stmt_vec_info::operator*): Likewise.
13775         (stmt_vec_info::operator gimple *): Likewise.
13776         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
13777         of NULL_STMT_VEC_INFO.
13778         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13779         (vect_reassociating_reduction_p): Likewise.
13780         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13781         (vectorizable_store): Likewise.
13782         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
13783         (vec_info::free_stmt_vec_infos): Likewise.
13785 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13787         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
13788         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
13790 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13792         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
13793         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13794         (vec_info::free_stmt_vec_info): New private member functions.
13795         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
13796         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
13797         * tree-parloops.c (gather_scalar_reductions): Remove calls to
13798         set_stmt_vec_info_vec and free_stmt_vec_infos.
13799         * tree-vect-loop.c (_loop_vec_info): Remove call to
13800         set_stmt_vec_info_vec.
13801         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
13802         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
13803         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
13804         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13805         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
13806         assignments in {vec_info::,}new_stmt_vec_info that are redundant
13807         with the clearing in the xcalloc.
13808         (stmt_vec_info_vec): Delete.
13809         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
13810         (vectorize_loops): Likewise.
13811         (vec_info::~vec_info): Remove argument from call to
13812         free_stmt_vec_infos.
13813         (vec_info::add_stmt): Remove vinfo argument from call to
13814         new_stmt_vec_info.
13816 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13818         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
13819         rather than a gimple stmt.
13820         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
13821         information for pattern statements when passed the original
13822         statement; instead wait to be passed the pattern statement itself.
13823         Don't call set_vinfo_for_stmt here.
13824         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
13825         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
13826         stmt_vec_infos here.
13827         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
13828         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
13829         stmt_vec_infos entry.
13831 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13833         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
13834         * tree-vectorizer.c (vec_info::replace_stmt): New function.
13835         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
13836         * tree-vect-stmts.c (vectorizable_call): Likewise.
13837         (vectorizable_simd_clone_call): Likewise.
13839 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13841         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
13842         * tree-vectorizer.c (vec_info::remove_stmt): New function.
13843         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
13844         * tree-vect-loop.c (vect_transform_loop): Likewise.
13845         * tree-vect-slp.c (vect_schedule_slp): Likewise.
13846         * tree-vect-stmts.c (vect_remove_stores): Likewise.
13848 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13850         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
13851         (vect_dr_stmt): Delete.
13852         * tree-vectorizer.c (vec_info::lookup_dr): New function.
13853         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
13854         of DR_VECT_AUX.
13855         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13856         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
13857         (vect_verify_datarefs_alignment, vect_peeling_supportable)
13858         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13859         (vect_analyze_data_refs): Likewise.
13860         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
13861         argument.
13862         (vect_find_same_alignment_drs): Likewise.
13863         (vect_slp_analyze_node_dependences): Update calls accordingly.
13864         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13865         instead of DR_VECT_AUX.
13866         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
13867         of a vector data references.  Use vec_info::lookup_dr instead of
13868         DR_VECT_AUX.
13869         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
13870         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13871         instead of DR_VECT_AUX.
13873 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13875         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
13876         dr_vec_info.
13877         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
13878         accordingly.
13879         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13880         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13881         (vect_gen_prolog_loop_niters): Likewise.
13883 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13885         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
13886         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
13887         (vect_known_alignment_in_bytes, vect_dr_behavior)
13888         (vect_get_scalar_dr_size): Take references as dr_vec_infos
13889         instead of data_references.  Update calls to other routines for
13890         which the same change has been made.
13891         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
13892         dr_vec_infos instead of stmt_vec_infos.
13893         (vect_analyze_data_ref_dependence): Update call accordingly.
13894         (vect_slp_analyze_data_ref_dependence)
13895         (vect_record_base_alignments): Use DR_VECT_AUX.
13896         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
13897         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
13898         (vector_alignment_reachable_p, vect_get_data_access_cost)
13899         (vect_peeling_supportable, vect_analyze_group_access_1)
13900         (vect_analyze_group_access, vect_analyze_data_ref_access)
13901         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
13902         (vect_compile_time_alias, vect_small_gap_p)
13903         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
13904         (vect_supportable_dr_alignment): Take references as dr_vec_infos
13905         instead of data_references.  Update calls to other routines for
13906         which the same change has been made.
13907         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
13908         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13909         (vect_slp_analyze_and_verify_node_alignment)
13910         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13911         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
13912         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
13913         above changes.
13914         (_vect_peel_info::dr): Replace with...
13915         (_vect_peel_info::dr_info): ...this new field.
13916         (vect_peeling_hash_get_most_frequent)
13917         (vect_peeling_hash_choose_best_peeling): Update accordingly.
13918         (vect_peeling_hash_get_lowest_cost):
13919         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
13920         routines for which the same change has been made.
13921         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
13922         data_reference.
13923         * tree-vect-loop-manip.c (get_misalign_in_elems)
13924         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
13925         above changes.
13926         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13927         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
13928         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
13929         (get_group_load_store_type, get_negative_load_store_type)
13930         (vect_get_data_ptr_increment, vectorizable_store)
13931         (vectorizable_load): Likewise.
13932         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
13933         Update calls to other routines for which the same change has been made.
13935 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13937         * tree-vectorizer.h (vec_info::move_dr): New member function.
13938         (dataref_aux): Rename to...
13939         (dr_vec_info): ...this and add "dr" and "stmt" fields.
13940         (_stmt_vec_info::dr_aux): Update accordingly.
13941         (_stmt_vec_info::data_ref_info): Delete.
13942         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
13943         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
13944         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
13945         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
13946         of data_ref.
13947         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
13948         (STMT_VINFO_DR_INFO): New macro.
13949         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
13950         (set_dr_misalignment): Update after rename of dataref_aux.
13951         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
13952         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
13953         initialization of STMT_VINFO_DATA_REF.
13954         * tree-vectorizer.c (vec_info::move_dr): New function.
13955         * tree-vect-patterns.c (vect_recog_bool_pattern)
13956         (vect_recog_mask_conversion_pattern)
13957         (vect_recog_gather_scatter_pattern): Use it.
13958         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
13959         the "dr" and "stmt" fields of dr_vec_info instead of
13960         STMT_VINFO_DATA_REF.
13962 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13964         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
13965         (is_pattern_stmt_p): Use it.
13966         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
13967         on pattern statements.
13969 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13971         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
13972         original stmt as a stmt_vec_info rather than a gimple stmt.
13973         (vect_pattern_recog_1): Take the statement directly as a
13974         stmt_vec_info, rather than via a gimple_stmt_iterator.
13975         Update call to vect_mark_pattern_stmts.
13976         (vect_pattern_recog): Update calls accordingly.
13978 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13980         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
13981         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
13982         a vect_def_type for the first argument.
13983         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
13984         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
13985         operand if it isn't defined by a vectorized statement.
13986         (vect_build_gather_load_calls): Remove the mask_dt argument and
13987         update calls to vect_get_vec_def_for_stmt_copy.
13988         (vectorizable_bswap): Likewise the dt argument.
13989         (vectorizable_call): Update calls to vectorizable_bswap and
13990         vect_get_vec_def_for_stmt_copy.
13991         (vectorizable_simd_clone_call, vectorizable_assignment)
13992         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
13993         (vectorizable_comparison): Update calls to
13994         vect_get_vec_def_for_stmt_copy.
13995         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
13996         vect_is_simple_use.
13997         (vect_get_loop_based_defs): Remove dt argument and update call
13998         to vect_get_vec_def_for_stmt_copy.
13999         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
14000         and vect_get_vec_def_for_stmt_copy.
14001         (vectorizable_load): Update calls to vect_build_gather_load_calls
14002         and vect_get_vec_def_for_stmt_copy.
14003         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14004         (vectorizable_reduction, vectorizable_live_operation): Update calls
14005         to vect_get_vec_def_for_stmt_copy.
14007 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14009         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
14010         and gimple stmt arguments with a stmt_vec_info.
14011         (vect_record_base_alignments): Update calls accordingly.
14012         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
14013         and gimple stmt arguments with a stmt_vec_info.
14014         (vect_build_slp_tree_1): Remove vinfo argument and update call
14015         to vect_record_max_nunits.
14016         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
14017         and vect_record_max_nunits.
14019 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14021         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
14022         file and take a stmt_vec_info instead of a gimple stmt.
14023         (supportable_widening_operation, vect_finish_replace_stmt)
14024         (vect_finish_stmt_generation, vect_get_store_rhs)
14025         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
14026         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
14027         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
14028         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
14029         (vect_create_data_ref_ptr, bump_vector_ptr)
14030         (vect_permute_store_chain, vect_setup_realignment)
14031         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
14032         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
14033         (vectorizable_reduction, vectorizable_induction)
14034         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
14035         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
14036         than gimple stmts as arguments.
14037         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
14038         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
14039         (can_group_stmts_p, vect_check_gather_scatter)
14040         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14041         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
14042         (vect_permute_load_chain, vect_shift_permute_load_chain)
14043         (vect_transform_grouped_load)
14044         (vect_record_grouped_load_vectors): Likewise.
14045         * tree-vect-loop.c (vect_fixup_reduc_chain)
14046         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
14047         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
14048         (vectorizable_reduction, vectorizable_induction)
14049         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
14050         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
14051         (vect_get_load_store_mask): Likewise.
14052         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
14053         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
14054         * tree-vect-stmts.c (vect_mark_relevant)
14055         (is_simple_and_all_uses_invariant)
14056         (exist_non_indexing_operands_for_use_p, process_use)
14057         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
14058         (vect_get_vec_def_for_operand, vect_get_vec_defs)
14059         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
14060         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
14061         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
14062         (get_negative_load_store_type, get_load_store_type)
14063         (vect_check_load_store_mask, vect_check_store_rhs)
14064         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
14065         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14066         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14067         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14068         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
14069         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
14070         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
14071         (supportable_widening_operation): Likewise.
14073 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14075         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
14076         a stmt_vec_info instead of a gcall.
14077         (vect_check_gather_scatter): Update call accordingly.
14078         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
14079         of a gphi.
14080         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
14081         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
14082         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
14083         instead of a gimple stmt.
14084         (vect_transform_loop): Update calls accordingly.
14085         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
14086         stmt_vec_infos instead of gimple stmts.
14087         (vect_analyze_slp_instance): Update use accordingly.
14088         * tree-vect-stmts.c (read_vector_array, write_vector_array)
14089         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
14090         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
14091         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
14092         (vect_gen_widened_results_half, vect_get_loop_based_defs)
14093         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
14094         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
14095         down to subroutines.
14097 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14099         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
14100         of the worklist from a vector of gimple stmts to a vector of
14101         stmt_vec_infos.
14102         * tree-vect-stmts.c (vect_mark_relevant, process_use)
14103         (vect_mark_stmts_to_be_vectorized): Likewise
14105 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14107         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
14108         statement before passing it to vect_analyze_stmt.
14109         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
14110         the chain of phi vector definitions.  Track the exit phi via its
14111         stmt_vec_info.
14112         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
14113         STMT_VINFO_REDUC_DEF.
14114         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
14115         stmt_vec_infos to handle the statement chains.
14116         (vect_get_slp_defs): Record the first statement in the node
14117         using a stmt_vec_info.
14118         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
14119         statements here and pass their stmt_vec_info down to subroutines.
14120         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
14121         down to vect_finish_stmt_generation.
14122         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
14123         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
14124         stmt_vec_infos to subroutines.
14125         (vect_remove_stores): Use stmt_vec_infos to handle the statement
14126         chains.
14128 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14130         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
14131         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
14132         (vect_permute_store_chain, vect_setup_realignment)
14133         (vect_permute_load_chain, vect_shift_permute_load_chain)
14134         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
14135         stmts internally, and when passing values to other vectorizer routines.
14136         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14137         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
14138         (vect_analyze_loop_operations, get_initial_def_for_reduction)
14139         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
14140         (vectorizable_reduction, vectorizable_induction)
14141         (vectorizable_live_operation, vect_transform_loop_stmt)
14142         (vect_transform_loop): Likewise.
14143         * tree-vect-patterns.c (vect_reassociating_reduction_p)
14144         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
14145         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
14146         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14147         (vect_slp_analyze_node_operations_1): Likewise.
14148         * tree-vect-stmts.c (vect_mark_relevant, process_use)
14149         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
14150         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
14151         (vect_finish_stmt_generation_1, get_group_load_store_type)
14152         (get_load_store_type, vect_build_gather_load_calls)
14153         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14154         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14155         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14156         (vectorizable_store, vectorizable_load, vectorizable_condition)
14157         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
14158         (supportable_widening_operation): Likewise.
14159         (vect_get_vector_types_for_stmt): Likewise.
14160         * tree-vectorizer.h (vect_dr_behavior): Likewise.
14162 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14164         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14165         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
14166         (vect_permute_store_chain, vect_permute_load_chain)
14167         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
14168         repeated stmt_vec_info lookups.
14169         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14170         (vect_update_ivs_after_vectorizer): Likewise.
14171         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
14172         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14173         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
14174         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14175         (vect_bb_slp_scalar_cost): Likewise.
14176         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
14178 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14180         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
14181         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
14182         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
14183         to get gassigns and gcalls, rather than operating on generc gimple
14184         stmts.
14185         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
14186         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
14187         (vectorizable_load, vect_analyze_stmt): Likewise.
14188         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
14190 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14192         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
14193         return stmt_vec_infos rather than gimple stmts.  Do not accept
14194         null arguments.
14195         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
14196         of a gimple stmt.
14197         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
14198         Update use of get_later_stmt.
14199         (vect_get_constant_vectors): Update call accordingly.
14200         (vect_schedule_slp_instance): Likewise
14201         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
14202         (vect_slp_analyze_instance_dependence): Likewise.
14203         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
14205 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14207         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
14208         (stmt_info_for_cost::stmt_info): ...this new field.
14209         (add_stmt_costs): Update accordingly.
14210         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
14211         (vect_get_known_peeling_cost): Likewise.
14212         (vect_estimate_min_profitable_iters): Likewise.
14213         * tree-vect-stmts.c (record_stmt_cost): Likewise.
14215 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14217         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
14218         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14219         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14220         accordingly.
14221         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
14223 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14225         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
14226         a gimple stmt to a stmt_vec_info.
14227         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
14229 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14231         * tree-vectorizer.h (vec_info::grouped_stores): Change from
14232         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14233         (_loop_vec_info::reduction_chains): Likewise.
14234         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
14235         accordingly.
14236         * tree-vect-slp.c (vect_analyze_slp): Likewise.
14238 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14240         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
14241         a gimple stmt to a stmt_vec_info.
14242         (_stmt_vec_info::next_element): Likewise.
14243         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
14244         (vect_slp_analyze_and_verify_node_alignment)
14245         (vect_analyze_group_access_1, vect_analyze_group_access)
14246         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
14247         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
14248         (vect_supportable_dr_alignment): Update accordingly.
14249         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
14250         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
14251         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
14252         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14253         * tree-vect-slp.c (vect_build_slp_tree_1)
14254         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
14255         (vect_split_slp_store_group, vect_analyze_slp_instance)
14256         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
14257         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
14258         (get_group_load_store_type, get_load_store_type)
14259         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
14260         (vect_transform_stmt, vect_remove_stores): Likewise.
14262 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14264         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
14265         than a gimple stmt.
14266         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14267         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
14268         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
14269         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
14270         (vector_alignment_reachable_p, vect_get_data_access_cost)
14271         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
14272         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
14273         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14274         (vect_analyze_group_access_1, vect_analyze_group_access)
14275         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
14276         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
14277         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
14278         result of vect_dr_stmt and use the stmt_vec_info instead of
14279         the associated gimple stmt.
14280         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14281         (vect_gen_prolog_loop_niters): Likewise.
14282         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14284 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14286         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
14287         to a vec<stmt_vec_info>.
14288         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
14289         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
14290         vec<stmt_vec_info>.
14291         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
14292         to a vec<stmt_vec_info>.
14293         (bst_traits::value_type, bst_traits::value_type): Likewise.
14294         (bst_traits::hash): Update accordingly.
14295         (vect_get_and_check_slp_defs): Change the stmts parameter from
14296         a vec<gimple *> to a vec<stmt_vec_info>.
14297         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
14298         (vect_build_slp_tree): Likewise.
14299         (vect_build_slp_tree_2): Likewise.  Update uses of
14300         SLP_TREE_SCALAR_STMTS.
14301         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
14302         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
14303         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
14304         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
14305         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
14306         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
14307         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
14308         (vect_get_constant_vectors, vect_get_slp_defs)
14309         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
14310         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
14311         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
14312         instead of gimple stmts.
14313         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
14314         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
14315         (vect_slp_analyze_instance_dependence): Update uses of
14316         SLP_TREE_SCALAR_STMTS.
14317         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14318         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
14319         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
14320         (get_initial_defs_for_reduction): Likewise.
14321         (vect_create_epilog_for_reduction): Likewise.
14322         (vectorize_fold_left_reduction): Likewise.
14323         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
14324         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
14325         (can_vectorize_live_stmts): Likewise.
14327 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14329         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
14330         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14331         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
14332         than gimple stmts.
14333         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
14334         of a gimple stmt.
14335         (gather_scalar_reductions): Update after above interface changes.
14336         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
14337         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
14338         than gimple stmts.
14339         (vect_force_simple_reduction): Likewise.
14340         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
14341         LOOP_VINFO_REDUCTIONS.
14342         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14344 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14346         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
14347         a gimple stmt to a stmt_vec_info.
14348         * tree-vect-loop.c (vect_active_double_reduction_p)
14349         (vect_force_simple_reduction, vectorizable_reduction): Update
14350         accordingly.
14352 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14354         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
14355         vec<gimple *> to a vec<stmt_vec_info>.
14356         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
14357         the reduction_phis argument from a vec<gimple *> to a
14358         vec<stmt_vec_info>.
14359         (vectorizable_reduction): Likewise the phis local variable that
14360         is passed to vect_create_epilog_for_reduction.  Update for new type
14361         of SLP_TREE_VEC_STMTS.
14362         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
14363         (vectorizable_live_operation): Likewise.
14364         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
14365         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
14367 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14369         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
14370         a gimple stmt to a stmt_vec_info.
14371         (vectorizable_condition, vectorizable_live_operation)
14372         (vectorizable_reduction, vectorizable_induction): Pass back the
14373         vectorized statement as a stmt_vec_info.
14374         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
14375         use of STMT_VINFO_VEC_STMT.
14376         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
14377         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
14378         as stmt_vec_infos rather than gimple stmts.
14379         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
14380         to a stmt_vec_info.
14381         (vectorizable_live_operation): Likewise.
14382         (vectorizable_reduction, vectorizable_induction): Likewise,
14383         updating use of STMT_VINFO_VEC_STMT.
14384         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
14385         of STMT_VINFO_VEC_STMT.
14386         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
14387         (vectorizable_simd_clone_call, vectorizable_conversion)
14388         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14389         (vectorizable_store, vectorizable_load, vectorizable_condition)
14390         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
14391         from a gimple stmt to a stmt_vec_info.
14392         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
14393         pointer to a stmt_vec_info to the vectorizable_* routines.
14395 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14397         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
14398         a gimple stmt to a stmt_vec_info.
14399         (is_pattern_stmt_p): Update accordingly.
14400         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
14401         (vect_record_grouped_load_vectors): Likewise.
14402         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
14403         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
14404         (vect_model_reduction_cost): Likewise.
14405         (vect_create_epilog_for_reduction): Likewise.
14406         (vectorizable_reduction, vectorizable_induction): Likewise.
14407         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14408         Return the stmt_vec_info for the pattern statement.
14409         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
14410         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
14411         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
14412         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
14413         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
14414         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
14415         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
14416         (free_stmt_vec_info, vect_is_simple_use): Likewise.
14418 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14420         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
14421         (vect_finish_stmt_generation): Likewise.
14422         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
14423         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
14424         (vect_build_gather_load_calls): Use the return value of the above
14425         functions instead of a separate call to vinfo_for_stmt.  Use narrow
14426         scopes for the input gimple stmt and wider scopes for the associated
14427         stmt_vec_info.  Use vec_info::lookup_def when setting these
14428         stmt_vec_infos from an SSA_NAME definition.
14429         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14430         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14431         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14432         (vectorizable_store, vectorizable_load, vectorizable_condition)
14433         (vectorizable_comparison): Likewise.
14434         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
14435         (vectorizable_reduction): Likewise.
14437 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14439         * tree-vectorizer.h (vect_is_simple_use): Add an optional
14440         stmt_vec_info * parameter before the optional gimple **.
14441         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
14442         (process_use, vect_get_vec_def_for_operand_1): Update callers.
14443         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
14444         * tree-vect-loop.c (vectorizable_reduction): Likewise.
14445         (vectorizable_live_operation): Likewise.
14446         * tree-vect-patterns.c (type_conversion_p): Likewise.
14447         (vect_look_through_possible_promotion): Likewise.
14448         (vect_recog_rotate_pattern): Likewise.
14449         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14451 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14453         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
14454         a typedef to a wrapper class.
14455         (NULL_STMT_VEC_INFO): New macro.
14456         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
14457         (stmt_vec_info::operator*): New function.
14458         (stmt_vec_info::operator gimple *): Likewise.
14459         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
14460         (add_stmt_costs): Likewise.
14461         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
14462         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
14463         (vect_get_known_peeling_cost): Likewise.
14464         (vect_estimate_min_profitable_iters): Likewise.
14465         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14466         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
14467         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
14468         (vectorizable_store, free_stmt_vec_infos): Likewise.
14469         (new_stmt_vec_info): Change return type of xcalloc to
14470         _stmt_vec_info *.
14472 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14474         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
14475         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
14476         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
14477         a single_imm_use-based sequence.
14478         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
14480 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14482         * tree-vectorizer.h (vec_info::lookup_def): Declare.
14483         * tree-vectorizer.c (vec_info::lookup_def): New function.
14484         * tree-vect-patterns.c (vect_get_internal_def): Use it.
14485         (vect_widened_op_tree): Likewise.
14486         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
14487         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
14488         (vectorizable_reduction): Likewise.
14489         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
14490         of a gimple *.
14491         (vect_is_slp_reduction): Update calls accordingly.  Use
14492         vec_info::lookup_def.
14493         (vect_is_simple_reduction): Likewise
14494         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
14496 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14498         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
14499         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
14500         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
14501         of vinfo_for_stmt.
14502         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
14503         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
14504         (vect_update_vf_for_slp, vect_analyze_loop_operations)
14505         (vect_is_slp_reduction, vectorizable_induction)
14506         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
14507         * tree-vect-patterns.c (vect_init_pattern_stmt):
14508         (vect_determine_min_output_precision_1, vect_determine_precisions)
14509         (vect_pattern_recog): Likewise.
14510         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
14511         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
14512         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
14513         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
14514         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
14515         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
14516         info field from a loop to a loop_vec_info.
14518 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14520         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
14521         (vec_info::add_stmt): Declare.
14522         * tree-vectorizer.c (vec_info::add_stmt): New function.
14523         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
14524         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
14525         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14526         (vectorizable_induction): Likewise.
14527         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
14528         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
14529         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
14530         (vectorizable_load): Likewise.
14531         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14532         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
14533         (vect_recog_gather_scatter_pattern): Likewise.
14534         (append_pattern_def_seq): Likewise.  Remove a check that is
14535         performed by add_stmt itself.
14537 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14539         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
14540         which make_ssa_name was called with new_stmt before new_stmt
14541         had been created.
14543 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14545         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
14546         split out from...
14547         (vect_is_slp_reduction): ...here...
14548         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
14549         that are already known to be false.
14551 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14553         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
14554         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
14555         STMT_VINFO_NUM_SLP_USES when it's true.
14556         (vect_free_slp_instance): Add a final_p parameter and pass it to
14557         vect_free_slp_tree.
14558         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
14559         (vect_analyze_slp_instance): Likewise.
14560         (vect_slp_analyze_operations): Likewise.
14561         (vect_slp_analyze_bb_1): Likewise.
14562         * tree-vectorizer.c (vec_info): Likewise.
14563         * tree-vect-loop.c (vect_transform_loop): Likewise.
14565 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14567         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14568         function is not called for second and subsequent members of
14569         a reduction group.
14571 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
14573         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
14574         cases for nested loops from here to ...
14575         (vect_create_epilog_for_reduction): ...here.  Only call
14576         vect_is_simple_use for inner-loop reductions.
14578 2018-07-31  Martin Liska  <mliska@suse.cz>
14580         PR gcov-profile/85338
14581         PR gcov-profile/85350
14582         PR gcov-profile/85372
14583         * profile.c (struct location_triplet): New.
14584         (struct location_triplet_hash): Likewise.
14585         (output_location): Do not output a BB that
14586         is already recorded for a line.
14587         (branch_prob): Use streamed_locations.
14589 2018-07-31  Martin Liska  <mliska@suse.cz>
14591         PR gcov-profile/85370
14592         * coverage.c (coverage_begin_function): Do not mark target
14593         clones as artificial functions.
14595 2018-07-31  Martin Liska  <mliska@suse.cz>
14597         PR gcov-profile/83813
14598         PR gcov-profile/84758
14599         PR gcov-profile/85217
14600         PR gcov-profile/85332
14601         * profile.c (branch_prob): Do not record GOTO expressions
14602         for GIMPLE statements which locations are already streamed.
14604 2018-07-31  Olivier Hainque  <hainque@adacore.com>
14606         * gcc.c (handle_spec_function): Accept a soft_matched_part
14607         argument, as do_spec_1.  Pass it down to ...
14608         (eval_spec_function): Accept a soft_matched_part argument,
14609         and pass it down to ...
14610         (do_spec_2): Accept a soft_matched_part argument, and pass
14611         it down to do_spec_1.
14612         (do_spec_1): Pass soft_matched_part to handle_spec_function.
14613         (handle_braces): Update call to handle_spec_function.
14614         (driver::set_up_specs): Update calls to do_spec_2.
14615         (compare_debug_dump_opt_spec_function): Likewise.
14616         (compare_debug_self_opt_spec_function): Likewise.
14618 2018-07-31  Olivier Hainque  <hainque@adacore.com>
14620         * common.opt (nolibc): New option.
14621         * doc/invoke.texi (Link Options): Document it.
14622         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
14623         * config/alpha/linux.h: Likewise.
14624         * config/arc/elf.h: Likewise.
14625         * config/arm/uclinux-elf.h: Likewise.
14626         * config/arm/unknown-elf.h: Likewise.
14627         * config/avr/avrlibc.h: Likewise.
14628         * config/bfin/bfin.h: Likewise.
14629         * config/bfin/linux.h: Likewise.
14630         * config/bfin/uclinux.h: Likewise.
14631         * config/darwin.h: Likewise.
14632         * config/darwin10.h: Likewise.
14633         * config/darwin12.h: Likewise.
14634         * config/gnu-user.h: Likewise.
14635         * config/lm32/uclinux-elf.h: Likewise.
14636         * config/pa/pa-hpux11.h: Likewise.
14637         * config/pa/pa64-hpux.h: Likewise.
14638         * config/sparc/sparc.h: Likewise.
14640 2018-07-31  Olivier Hainque  <hainque@adacore.com>
14642         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
14643         undefined variables.
14645 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
14647         PR target/86640
14648         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
14649         instead of GEN_INT.
14651 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14653         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
14654         terminated string literal.
14656 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
14658         PR rtl-optimization/85160
14659         * combine.c (is_just_move): New function.
14660         (try_combine): Allow combining two instructions into two if neither of
14661         the original instructions was a move.
14663 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
14665         PR target/86673
14666         * doc/extend.texi (Global Register Variables): Discourage use of type
14667         qualifiers.
14668         (Local Register Variables): Likewise.
14670 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
14672         PR tree-optimization/86506
14673         * hwint.h (ceil_log2): Resync with hwint.c implementation.
14675 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
14677         PR target/86547
14678         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
14679         hard_regno, make sure no insn between `from` and `to` clobbers it.
14681 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
14682             Tom de Vries  <tdevries@suse.de>
14684         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
14685         (PTX_DEFAULT_RUNTIME_DIM): ... this.
14686         (nvptx_goacc_validate_dims): Set default worker and gang dims to
14687         PTX_DEFAULT_RUNTIME_DIM.
14688         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
14690 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
14692         * config/pa/pa.c (pa_output_addr_vec): Align address table.
14693         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
14694         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
14696 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
14698         * config/rs6000/constraints.md (wG constraint): Delete, no longer
14699         used.
14700         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
14701         predicate to reflect toc fusion has been deleted.
14702         (toc_fusion_mem_raw): Delete, no longer used.
14703         (toc_fusion_mem_wrapped): Likewise.
14704         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
14705         fusion mask bit.
14706         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
14707         Delete, no longer used.
14708         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
14709         meant to be used for toc fusion.
14710         (rs6000_debug_print_mode): Delete toc fusion debugging.
14711         (rs6000_debug_reg_global): Likewise.
14712         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
14713         fusion and secondary reload support that were never used.
14714         (rs6000_option_override_internal): Delete TOC fusion, that was only
14715         partially defined, and it did not work unless you also used the
14716         -mcmodel= switch.
14717         (rs6000_legitimate_address_p): Delete TOC fusion support.
14718         (rs6000_opt_masks): Likewise.
14719         (fusion_wrap_memory_address): Delete function, no longer used.
14720         (fusion_split_address); Delete TOC fusion support.
14721         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
14722         longer used with toc fusion being deleted.
14723         (TARGET_TOC_FUSION_FP): Likewise.
14724         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
14725         UNSPEC.
14726         (toc fusion spliter): Delete TOC fusion support.
14727         (toc_fusionload_<mode>): Likewise.
14728         (toc_fusionload_di): Likewise.
14729         (fusion_gpr_load_<mode>): Delete generator function, this insn no
14730         longer needs to be named.  Rename predicate to delete TOC fusion.
14731         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14732         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14733         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14734         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14735         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
14737 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14739         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
14740         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
14741         __int128 in built-in function prototypes.
14742         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
14743         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
14745 2018-07-27  Martin Sebor  <msebor@redhat.com>
14747         PR tree-optimization/86696
14748         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
14749         types, including enums.
14750         (handle_char_store): Be prepared for the above function to fail.
14752 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
14754         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
14755         when optimization level is lower than 2 or optimize for size.
14757 2018-07-26  Martin Sebor  <msebor@redhat.com>
14759         PR tree-optimization/86043
14760         PR tree-optimization/86042
14761         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
14762         (get_string_cst_length): Rename...
14763         (get_min_string_length): ...to this.  Add argument.
14764         (handle_char_store): Extend to handle multi-character stores by
14765         MEM_REF.
14766         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
14767         * tree.h (initializer_zerop): Add argument.
14769 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
14771         PR middle-end/86660
14772         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
14773         declare target to variables if they have always,{to,from,tofrom} map
14774         kinds.
14776 2018-07-26  Martin Liska  <mliska@suse.cz>
14778         PR lto/86548
14779         * lto-wrapper.c: Add linker_output as prefix
14780         for ltrans_output_file.
14782 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
14784         PR rtl-optimization/85805
14785         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
14786         value for hard registers if that was written in the same mode.
14788 2018-07-26  Martin Liska  <mliska@suse.cz>
14790         PR gcov-profile/86536
14791         * gcov.c (format_gcov): Use printf format %.*f directly
14792         and do not handle special values.
14794 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14796         * common/config/arc/arc-common.c (arc_option_optimization_table):
14797         Update default optimizations for size.
14799 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14801         * config/arc/arc.md (movsf_insn): Add short instruction selection.
14802         * config/arc/constraints.md (CfZ): New constraint.
14803         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
14804         (subsf3_fpu): Likewise.
14805         (cmpsf_fpu): Likewise.
14806         (cmpsf_fpu_uneq): Likewise.
14808 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14810         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
14811         accesses as well.
14812         (arc_is_uncached_mem_p): uncached applies to both the variable and
14813         the pointer.
14815 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14817         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
14818         register names.
14820 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
14822         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
14823         field "m_scopes" from vec to auto_vec.
14825 2018-07-25  Martin Liska  <mliska@suse.cz>
14827         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
14828         return type.
14830 2018-07-25  Richard Biener  <rguenther@suse.de>
14832         PR debug/86654
14833         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
14834         special wrt context_die late.
14835         (gen_subprogram_die): Re-use DIEs in local scope.
14837 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
14839         PR tree-optimization/86644
14840         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
14842 2018-07-25  Martin Liska  <mliska@suse.cz>
14844         PR middle-end/86645
14845         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
14846         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
14848 2018-07-25  Martin Liska  <mliska@suse.cz>
14850         PR sanitizer/79635
14851         * params.def: Explain ASan abbreviation and provide
14852         a documentation link.
14854 2018-07-24  Martin Sebor  <msebor@redhat.com>
14856         PR tree-optimization/86622
14857         PR tree-optimization/86532
14858         * builtins.h (string_length): Declare.
14859         * builtins.c (c_strlen): Correct handling of non-constant offsets.
14860         (check_access): Be prepared for non-constant length ranges.
14861         (string_length): Make extern.
14862         * expr.c (string_constant): Only handle the minor non-constant
14863         array index.  Use string_constant to compute the length of
14864         a generic string constant.
14866 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
14868         PR tree-optimization/86618
14869         * tree-vect-stmts.c (vectorizable_call): Don't take the address
14870         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
14872 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
14874         PR tree-optimization/86636
14875         * json.cc (json::object::set): Fix comment.  Add assertions.
14876         (json::array::append): Move here from json.h.  Add comment and an
14877         assertion.
14878         (json::string::string): Likewise.
14879         * json.h (json::array::append): Move to json.cc.
14880         (json::string::string): Likewise.
14881         * optinfo-emit-json.cc
14882         (optrecord_json_writer::impl_location_to_json): Assert that we
14883         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
14884         wrapper around it.  Expand the location once, rather than three
14885         times.
14886         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
14887         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
14888         wrappers.
14889         (optrecord_json_writer::optinfo_to_json): Likewise, in four
14890         places.  Fix some overlong lines.
14892 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
14894         * config/aarch64/aarch64-simd.md
14895         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
14896         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
14897         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
14898         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
14899         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
14900         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
14901         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
14902         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
14903         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
14905 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
14907         PR middle-end/86627
14908         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
14909         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
14910         and abs_d == d, do the power of two handling if profitable.
14912 2018-07-24  Richard Biener  <rguenther@suse.de>
14914         * match.pd: Add BIT_FIELD_REF canonicalizations.
14916 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14918         PR c/86617
14919         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
14921 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14923         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
14924         terminated STRING_CST object.
14926 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14928         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
14930 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14932         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
14933         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
14934         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
14935         the elements into a register.
14936         (rs6000_split_v4si_init_di_reg): Delete.
14937         (rs6000_split_v4si_init): Delete.
14938         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
14939         (vsx_init_v4si): Rewrite as a define_expand.
14941 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14943         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
14944         zero_extend argument from memory): New.
14946 2018-07-22  Martin Sebor  <msebor@redhat.com>
14948         PR bootstrap/86621
14949         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
14950         diagnosing calls with unknown arguments unless -Walloca-larger-than
14951         is restricted to less than PTRDIFF_MAX bytes.
14953 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
14955         * doc/gcov.texi (Invoking Gcov): Editorial changes.
14957 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
14959         * pretty-print.c (text_info::set_location): Remove redundant
14960         "line_table" parameter from call to rich_location::set_range.
14962 2018-07-20  Martin Sebor  <msebor@redhat.com>
14964         PR middle-end/82063
14965         * builtins.c (expand_builtin_alloca): Adjust.
14966         * calls.c (alloc_max_size): Simplify.
14967         * cgraphunit.c (cgraph_node::expand): Adjust.
14968         * common.opt (larger_than_size, warn_frame_larger_than): Remove
14969         variables.
14970         (frame_larger_than_size): Same.
14971         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
14972         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
14973         Initialize.
14974         * doc/invoke.texi (GCC Command Options): Document option arguments.
14975         Explain byte-size arguments and suffixes.
14976         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
14977         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
14978         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
14979         * doc/options.texi (UInteger): Expand.
14980         (Host_Wide_Int, ByteSize): Document new properties.
14981         * final.c (final_start_function_1): Include sizes in an error message.
14982         * function.c (frame_offset_overflow): Same.
14983         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
14984         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
14985         Diagnose unbounded alloca calls only for limits of less than
14986         PTRDIFF_MAX.
14987         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
14988         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
14989         for alloca(0).
14990         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
14991         only for limits of less than PTRDIFF_MAX.
14992         * langhooks-def.h (lhd_handle_option): Change function argument
14993         to HOST_WIDE_INT.
14994         * langhooks.c (lhd_handle_option): Same.
14995         * langhooks.h (handle_option): Same.
14996         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
14997         ByteSize flags.
14998         (var_type, var_type_struct): Same.
14999         (var_set): Handle ByteSize flag.
15000         * optc-gen.awk: Add comments to output to ease debugging.  Make
15001         use of HOST_WIDE_INT where appropriate.
15002         * opts-gen-save.awk:  Use %lx to format unsigned long.
15003         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
15004         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
15005         arguments.  Parse bytes-size suffixes.
15006         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
15007         (enum_value_to_arg): Same.
15008         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
15009         (handle_option): Adjust.
15010         (generate_option): Change function argument to HOST_WIDE_INT.
15011         (cmdline_handle_error): Adjust.
15012         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
15013         (set_option): Change function argument to HOST_WIDE_INT.
15014         (option_enabled): Handle cl_host_wide_int.
15015         (get_option_state): Handle CLVC_SIZE.
15016         (control_warning_option): Same.
15017         * opts.c (common_handle_option): Change function argument to
15018         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
15019         OPT_Wvla_larger_than_.
15020         * opts.h (enum cl_var_type): Add an enumerator.
15021         * stor-layout.c (layout_decl): Print a more meaningful warning.
15022         * toplev.c (output_stack_usage): Adjust.
15024 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
15026         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
15027         call to inline_expand_builtin_string_cmp.
15028         (expand_builtin_strcmp): Likewise.
15029         (expand_builtin_strncmp): Likewise.
15030         (inline_string_cmp): Delete the last parameter, change char_type_node
15031         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
15032         two operands.
15033         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
15034         the inlining expansion on target where the type of the call has same or
15035         narrower precision than unsigned char.
15037 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
15039         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
15040         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
15041         * common.opt (fsave-optimization-record): New option.
15042         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
15043         * doc/invoke.texi (-fsave-optimization-record): New option.
15044         * dumpfile.c: Include "optinfo-emit-json.h".
15045         (struct kv_pair): Move to coretypes.h.
15046         (optgroup_options): Make non-static.
15047         (dump_context::end_scope): Call
15048         optimization_records_maybe_pop_dump_scope.
15049         * dumpfile.h (optgroup_options): New decl.
15050         * json.cc: New file.
15051         * json.h: New file.
15052         * optinfo-emit-json.cc: New file.
15053         * optinfo-emit-json.h: New file.
15054         * optinfo.cc: Include "optinfo-emit-json.h".
15055         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
15056         (optinfo_enabled_p): Check optimization_records_enabled_p.
15057         (optinfo_wants_inlining_info_p): Likewise.
15058         * optinfo.h: Update comment.
15059         * profile-count.c (profile_quality_as_string): New function.
15060         * profile-count.h (profile_quality_as_string): New decl.
15061         (profile_count::quality): New accessor.
15062         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
15063         and optinfo_emit_json_cc_tests.
15064         * selftest.h (selftest::json_cc_tests): New decl.
15065         (selftest::optinfo_emit_json_cc_tests): New decl.
15066         * toplev.c: Include "optinfo-emit-json.h".
15067         (compile_file): Call optimization_records_finish.
15068         (do_compile): Call optimization_records_start.
15069         * tree-ssa-live.c: Include optinfo.h.
15070         (remove_unused_scope_block_p): Retain inlining information if
15071         optinfo_wants_inlining_info_p returns true.
15073 2018-07-20  Richard Biener  <rguenther@suse.de>
15075         PR debug/86585
15076         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
15077         to cover -flto-partition=none.
15079 2018-07-20  Martin Liska  <mliska@suse.cz>
15081         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
15082         (get_decl_source_range): Remove unused function.
15084 2018-07-20  Richard Biener  <rguenther@suse.de>
15086         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
15087         (struct vn_phi_s): Likewise.
15088         (struct vn_reference_s): Likewise.
15089         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
15090         for searching the slot of an entry known to be in the hash itself.
15091         (vn_phi_hasher::equal): Likewise.
15092         (vn_reference_hasher::equal): Likewise.
15093         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
15094         globals.
15095         (optimistic_info, current_info): Remove, keeping only valid_info.
15096         (vn_reference_lookup_1): Remove fallback lookup.
15097         (vn_reference_lookup_2): Likewise.
15098         (vn_nary_op_lookup_1): Likewise.
15099         (vn_phi_lookup): Likewise.
15100         (vn_nary_build_or_lookup_1): Make sure to not chain the built
15101         hash element.
15102         (vn_reference_insert): Adjust, chain the inserted hash element
15103         at last_inserted_ref.
15104         (vn_reference_insert_pieces): Likewise.
15105         (visit_reference_op_call): Likewise.
15106         (vn_nary_op_insert_into): Chain the inserted hash element at
15107         last_inserted_nary.
15108         (vn_nary_op_insert_pieces): Adjust.
15109         (vn_nary_op_insert): Likewise.
15110         (vn_nary_op_insert_stmt): Likewise.
15111         (vn_phi_insert): Adjust, chain the inserted hash element at
15112         last_inserted_phi.
15113         (process_scc): Remove clearing and copying the optimistic
15114         table.  Instead remove elements inserted during an optimistic
15115         iteration from the single table we maintain.
15116         (init_scc_vn): Adjust.
15117         (free_scc_vn): Likewise.
15118         (sccvn_dom_walker::record_cond): Likewise.
15119         (sccvn_dom_walker::after_dom_children): Likewise.
15121 2018-07-19  Martin Sebor  <msebor@redhat.com>
15123         PR tree-optimization/84047
15124         PR tree-optimization/83776
15125         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
15126         (check_array_bounds): Call it.
15128 2018-07-19  Martin Sebor  <msebor@redhat.com>
15130         * align.h (align_flags): Use member initialization.
15132 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
15134         * Makefile.in (OBJS): Add optinfo.o.
15135         * coretypes.h (class symtab_node): New forward decl.
15136         (struct cgraph_node): New forward decl.
15137         (class varpool_node): New forward decl.
15138         * dump-context.h: New file.
15139         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
15140         "tree-pass.h".
15141         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
15142         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
15143         (set_alt_dump_file): Likewise.
15144         (dump_context::~dump_context): New dtor.
15145         (dump_gimple_stmt): Move implementation to...
15146         (dump_context::dump_gimple_stmt): ...this new member function.
15147         Add the stmt to any pending optinfo, creating one if need be.
15148         (dump_gimple_stmt_loc): Move implementation to...
15149         (dump_context::dump_gimple_stmt_loc): ...this new member function.
15150         Start a new optinfo and add the stmt to it.
15151         (dump_gimple_expr): Move implementation to...
15152         (dump_context::dump_gimple_expr): ...this new member function.
15153         Add the stmt to any pending optinfo, creating one if need be.
15154         (dump_gimple_expr_loc): Move implementation to...
15155         (dump_context::dump_gimple_expr_loc): ...this new member function.
15156         Start a new optinfo and add the stmt to it.
15157         (dump_generic_expr): Move implementation to...
15158         (dump_context::dump_generic_expr): ...this new member function.
15159         Add the tree to any pending optinfo, creating one if need be.
15160         (dump_generic_expr_loc): Move implementation to...
15161         (dump_context::dump_generic_expr_loc): ...this new member
15162         function.  Add the tree to any pending optinfo, creating one if
15163         need be.
15164         (dump_printf): Move implementation to...
15165         (dump_context::dump_printf_va): ...this new member function.  Add
15166         the text to any pending optinfo, creating one if need be.
15167         (dump_printf_loc): Move implementation to...
15168         (dump_context::dump_printf_loc_va): ...this new member function.
15169         Start a new optinfo and add the stmt to it.
15170         (dump_dec): Move implementation to...
15171         (dump_context::dump_dec): ...this new member function.  Add the
15172         value to any pending optinfo, creating one if need be.
15173         (dump_context::dump_symtab_node): New member function.
15174         (dump_context::get_scope_depth): New member function.
15175         (dump_context::begin_scope): New member function.
15176         (dump_context::end_scope): New member function.
15177         (dump_context::ensure_pending_optinfo): New member function.
15178         (dump_context::begin_next_optinfo): New member function.
15179         (dump_context::end_any_optinfo): New member function.
15180         (dump_context::s_current): New global.
15181         (dump_context::s_default): New global.
15182         (dump_scope_depth): Delete global.
15183         (dumpfile_ensure_any_optinfo_are_flushed): New function.
15184         (dump_symtab_node): New function.
15185         (get_dump_scope_depth): Reimplement in terms of dump_context.
15186         (dump_begin_scope): Likewise.
15187         (dump_end_scope): Likewise.
15188         (selftest::temp_dump_context::temp_dump_context): New ctor.
15189         (selftest::temp_dump_context::~temp_dump_context): New dtor.
15190         (selftest::verify_item): New function.
15191         (ASSERT_IS_TEXT): New macro.
15192         (ASSERT_IS_TREE): New macro.
15193         (ASSERT_IS_GIMPLE): New macro.
15194         (selftest::test_capture_of_dump_calls): New test.
15195         (selftest::dumpfile_c_tests): Call it.
15196         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
15197         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
15198         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
15199         descriptive comment.
15200         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
15201         (dump_node, dump_bb): Move these unrelated decls.
15202         (class dump_manager): Add leading comment.
15203         * optinfo.cc: New file.
15204         * optinfo.h: New file.
15206 2018-07-19  Michael Collison  <michael.collison@arm.com>
15207             Richard Henderson <rth@redhat.com>
15209         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
15210         (subti): Handle op1 zero.
15211         (subvti4, usub4ti4): New.
15212         (*sub<GPI>3_compare1_imm): New.
15213         (sub<GPI>3_carryinCV): New.
15214         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
15215         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
15217 2018-07-19  Michael Collison  <michael.collison@arm.com>
15218             Richard Henderson <rth@redhat.com>
15220         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
15221         (addti3): Create simpler code if low part is already known to be 0.
15222         (addvti4, uaddvti4): New.
15223         (*add<GPI>3_compareC_cconly_imm): New.
15224         (*add<GPI>3_compareC_cconly): New.
15225         (*add<GPI>3_compareC_imm): New.
15226         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
15227         handle constants within this pattern..
15228         (*add<GPI>3_compareV_cconly_imm): New.
15229         (*add<GPI>3_compareV_cconly): New.
15230         (*add<GPI>3_compareV_imm): New.
15231         (add<GPI>3_compareV): New.
15232         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
15233         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
15234         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
15235         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
15236         with 'comparison' operator.
15237         (*add<GPI>3_compareV_cconly_imm): Ditto.
15238         (*add<GPI>3_compareV_cconly): Ditto.
15239         (*add<GPI>3_compareV_imm): Ditto.
15240         (add<GPI>3_compareV): Ditto.
15241         (add<mode>3_carryinC): Ditto.
15242         (*add<mode>3_carryinC_zero): Ditto.
15243         (*add<mode>3_carryinC): Ditto.
15244         (add<mode>3_carryinV): Ditto.
15245         (*add<mode>3_carryinV_zero): Ditto.
15246         (*add<mode>3_carryinV): Ditto.
15248 2018-07-19  Michael Collison  <michael.collison@arm.com>
15249             Richard Henderson <rth@redhat.com>
15251         * config/aarch64/aarch64-modes.def (CC_V): New.
15252         * config/aarch64/aarch64-protos.h
15253         (aarch64_addti_scratch_regs): Declare
15254         (aarch64_subvti_scratch_regs): Declare.
15255         (aarch64_expand_subvti): Declare.
15256         (aarch64_gen_unlikely_cbranch): Declare
15257         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
15258         for signed overflow using CC_Vmode.
15259         (aarch64_get_condition_code_1): Handle CC_Vmode.
15260         (aarch64_gen_unlikely_cbranch): New function.
15261         (aarch64_addti_scratch_regs): New function.
15262         (aarch64_subvti_scratch_regs): New function.
15263         (aarch64_expand_subvti): New function.
15265 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15267         * config/aarch64/aarch64-option-extensions.def: New entry for profile
15268         extension.
15269         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
15270         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
15271         extension.
15273 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15275         PR target/83009
15276         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15277         address check not strict.
15279 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15281         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
15282         Umq with Umn.
15283         (store_pair_lanes<mode>): Likewise.
15284         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
15285         enum value 'ADDR_QUERY_LDP_STP_N'.
15286         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
15287         (aarch64_print_address_internal): Add declaration.
15288         (aarch64_print_ldpstp_address): Remove.
15289         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
15290         (aarch64_print_operand): Change printing of 'y'.
15291         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
15292         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
15293         'true' rather than '1'.
15294         * config/aarch64/constraints.md (Uml): Likewise.
15295         (Uml): Rename to Umn.
15296         (Umq): Remove.
15298 2018-07-19  Richard Biener  <rguenther@suse.de>
15300         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
15301         a trailing array.
15302         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
15303         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
15304         (vn_reference_hasher): Likewise.
15305         (struct vn_tables_s): Remove obstack and alloc-pool members.
15306         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
15307         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
15308         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
15309         (vn_reference_insert_pieces): Likewise.
15310         (alloc_vn_nary_op_noinit): Adjust.
15311         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
15312         (vn_phi_eq): Adjust.
15313         (shared_lookup_phiargs): Remove.
15314         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
15315         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
15316         (visit_reference_op_call): Likewise.
15317         (copy_nary, copy_phi, copy_reference): Remove.
15318         (process_scc): Rewind the obstack when iterating.  Do not
15319         copy the elements to valid_info but just move them from one
15320         hashtable to the other.
15321         (allocate_vn_table): Adjust.
15322         (free_vn_table): Likewise.
15323         (init_scc_vn): Likewise.
15324         (free_scc_vn): Likewise.
15326 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
15328         PR target/86560
15329         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
15330         indirect_return as function type attribute.
15331         (ix86_attribute_table): Change indirect_return to function
15332         type attribute.
15333         * doc/extend.texi: Update indirect_return attribute.
15335 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
15337         * wide-int.h (widest2_int): New.
15338         * gimple-fold.c (arith_overflowed_p): Use it.
15339         * tree.h (widest2_int_cst): New.
15340         * tree-vrp.c (wide_int_binop_overflow): Rename from
15341         vrp_int_const_binop.
15342         Rewrite to work on trees.
15343         (extract_range_from_multiplicative_op_1): Abstract code to...
15344         (wide_int_range_min_max): ...here.
15345         (wide_int_range_cross_product): ...and here.
15346         (extract_range_from_binary_expr_1): Abstract overflow code to...
15347         (wide_int_range_mult_wrapping): ...here.
15348         * tree-vrp.h (wide_int_range_cross_product): New.
15349         (wide_int_range_mult_wrapping): New.
15351 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
15352             Julia Koval  <julia.koval@intel.com>
15354         * config/i386/x86-tune-costs.h (skylake_memcpy,
15355         skylake_memset): Replace rep_prefix with unrolling for size 512.
15357 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
15359         PR middle-end/86544
15360         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
15361         comparision with EQ_EXPR in last stmt.
15363 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15365         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
15366         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
15367         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
15368         previously known as "PowerPC AltiVec Built-in Functions".  Move
15369         some material to new subsubsections "PowerPC AltiVec Built-in
15370         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
15371         ISA 2.07".
15372         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
15373         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
15374         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
15375         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
15377 2018-07-18  Richard Biener  <rguenther@suse.de>
15379         PR tree-optimization/86557
15380         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
15381         EXACT_DIV_EXPR.
15383 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
15385         * config/s390/s390.c (s390_function_profiler): Generate CFI.
15387 2018-07-17  Jeff Law  <law@redhat.com>
15389         * config/arm/arm.c (get_label_padding): Update for recent
15390         changes to label_to_alignment.
15392         PR tree-optimization/86010
15393         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
15395         * config/mips/mips.c (vr4130_align_insns): Update for recent
15396         changes to label_to_alignment.
15398         * config/frv/frv.c (frv_label_align): Update for recent changes
15399         to label_to_alignment.
15401         * config/nios2/nios2.c (nios2_label_align): Update for recent
15402         changes which dropped ALIGN_LABELS_LOG.
15404 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
15406         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
15407         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
15409 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
15411         * config/arc/arc.c (arc_label_align): Use align_labels instead of
15412         deprecated align_labels_log.
15414 2018-07-17  Richard Biener  <rguenther@suse.de>
15416         PR lto/86456
15417         * dwarf2out.c (init_sections_and_labels): Always generate
15418         a debug_line_str_section for early LTO debug.
15419         (dwarf2out_finish): Reset debug_line_str_hash output early.
15420         Bump counter for extra dwarf5 .debug_loc labels to not conflict
15421         with fat LTO part.
15422         (dwarf2out_early_finish): Output debug_line_str.
15424 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
15426         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
15427         index register on z196 or later.
15429 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
15431         * config/s390/s390.c (s390_default_align): Set default function
15432         alignment to 16.
15433         (s390_override_options_after_change): Call s390_default align.
15434         (s390_option_override_internal): Call s390_default_align.
15435         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
15437 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
15439         PR middle-end/86542
15440         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
15441         to _looptemp_ clauses, other than the first two.
15443 2018-07-17  Martin Liska  <mliska@suse.cz>
15445         * opts.c: Do not enable OPT_falign_* for -Os.
15447 2018-07-17  Martin Liska  <mliska@suse.cz>
15449         * align.h (MAX_CODE_ALIGN): New.
15450         (MAX_CODE_ALIGN_VALUE): New.
15451         * common/config/i386/i386-common.c (ix86_handle_option):
15452         (MAX_CODE_ALIGN): Moved to align.h.
15453         * final.c (MAX_CODE_ALIGN): Likewise.
15454         * opts.c (parse_and_check_align_values):
15455         (MAX_CODE_ALIGN): Likewise.
15456         (MAX_CODE_ALIGN_VALUE): Likewise.
15458 2018-07-17  Martin Liska  <mliska@suse.cz>
15460         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
15461         in order to fulfil coding style.
15462         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
15463         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15464         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15465         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
15466         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
15467         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
15468         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
15469         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
15470         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15472 2018-07-17  Martin Liska  <mliska@suse.cz>
15474         * align.h: New file.
15475         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
15476         directly.
15477         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
15478         align_flags of label_to_alignment.
15479         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
15480         align_flags class.
15481         * config/m68k/m68k.c: Do not use removed align_labels_value and
15482         align_loops_value.
15483         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
15484         (LOOP_ALIGN): Likewise.
15485         (LABEL_ALIGN): Likewise.
15486         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
15487         Remove not used macro.
15488         (rs6000_loop_align): Change return type to align_flags.
15489         (rs6000_loop_align_max_skip): Remove.
15490         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
15491         Change return type to align_flags.
15492         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
15493         Remove not used macro.
15494         (rs6000_loop_align):  Change return type to align_flags.
15495         (rs6000_loop_align_max_skip): Remove.
15496         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
15497         * config/rx/rx-protos.h (rx_align_for_label): Make it
15498         static function.
15499         * config/rx/rx.c (rx_align_for_label): Change return type
15500         to align_flags.
15501         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
15502         macro definitions.
15503         into align_flags class.
15504         (LABEL_ALIGN): Likewise.
15505         (LOOP_ALIGN): Likewise.
15506         * config/s390/s390.c (s390_label_align): Use align_flags
15507         class member.
15508         (s390_asm_output_function_label): Likewise.
15509         * config/sh/sh.c (sh_override_options_after_change):
15510         Use align_flags class directly without macros.
15511         (find_barrier): Likewise.
15512         (barrier_align): Likewise.
15513         (sh_loop_align): Likewise.
15514         * config/spu/spu.c (spu_option_override):
15515         Use align_flags_tuple::get_value instead of removed macros.
15516         (spu_sched_init): Likewise.
15517         * config/spu/spu.h (GTY): Likewise.
15518         * config/visium/visium.c (visium_option_override):
15519         Set "8" as default secondary alignment.
15520         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
15521         in order to guarantee secondary alignment of 8.
15522         * coretypes.h: Include align.h header file.
15523         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
15524         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
15525         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
15526         * doc/tm.texi.in: Likewise.
15527         * final.c (struct label_alignment): Remove not used structure.
15528         (LABEL_ALIGN): Change type to align_flags.
15529         (LOOP_ALIGN): Likewise.
15530         (JUMP_ALIGN): Likewise.
15531         (default_loop_align_max_skip): Remove.
15532         (default_label_align_max_skip): Likewise.
15533         (default_jump_align_max_skip): Likewise.
15534         (default_label_align_after_barrier_max_skip):
15535         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
15536         (LABEL_TO_MAX_SKIP): Remove.
15537         (label_to_alignment): Return align_flags type instead of integer.
15538         (label_to_max_skip): Remove.
15539         (align_fuzz): Use align_flags type.
15540         (compute_alignments): Use align_flags type and use align_flags::max
15541         to combine multiple alignments.
15542         (grow_label_align): Grow vec instead of C array.
15543         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
15544         (shorten_branches):  Use align_flags type and use align_flags::max
15545         to combine multiple alignments.
15546         (final_scan_insn_1): Remove usage of secondary alignment that comes
15547         from label alignment, but instead use proper secondary alignment
15548         which is computed in grow_label_align.
15549         * flags.h (struct align_flags_tuple): Move to align.h.
15550         (struct align_flags): Likewise.
15551         (state_align_loops): Rename to align_loops.
15552         (state_align_jumps): Rename to align_jumps.
15553         (state_align_labels): Rename to align_labels.
15554         (state_align_functions): Rename to align_functions.
15555         (align_loops_log): Remove.
15556         (align_jumps_log): Remove.
15557         (align_labels_log): Remove.
15558         (align_functions_log): Remove.
15559         (align_loops_max_skip): Remove.
15560         (align_jumps_max_skip): Remove.
15561         (align_labels_max_skip): Remove.
15562         (align_functions_max_skip): Remove.
15563         (align_loops_value): Remove.
15564         (align_jumps_value): Remove.
15565         (align_labels_value): Remove.
15566         (align_functions_value): Remove.
15567         * output.h (label_to_alignment): Change return type to align_flags.
15568         (label_to_max_skip): Remove.
15569         * target.def: Remove loop_align_max_skip, label_align_max_skip,
15570         jump_align_max_skip macros.
15571         * targhooks.h (default_loop_align_max_skip): Remove.
15572         (default_label_align_max_skip): Likewise.
15573         (default_jump_align_max_skip): Likewise.
15574         (default_label_align_after_barrier_max_skip): Remove.
15575         * toplev.c (read_log_maxskip): Use ::normalize function.
15576         (parse_N_M): Remove not used argument and also call ::normalize.
15577         (parse_alignment_opts): Do not pass unused arguments.
15578         * varasm.c (assemble_start_function): Use directly align_functions
15579         instead of removed macros.
15580         * system.h: Do not poison removed macros.
15582 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
15584         PR middle-end/86539
15585         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
15586         and cond temporaries don't have reference type if iterator has
15587         pointer type.  For init use &for_pre_body instead of pre_p if
15588         for_pre_body is non-empty.
15590 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
15592         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
15593         double-double modes to SFmode directly directly.
15594         (trunc<mode>sf2_fprs): Delete.
15596 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
15598         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
15599         for conversions between IFmode and the decimal floating point modes.
15600         (init_float128_ieee): Use the correct names for conversions between
15601         KFmode and the decimal floating point modes.
15603 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
15605         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
15606         for the conversions between TDmode and IFmode.
15607         (init_float128_ieee): Use more correct names for the conversions
15608         between TDmode and KFmode.
15610 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
15612         PR tree-optimization/86526
15613         * builtins.c (expand_builtin_memcmp): Formatting fixes.
15614         (inline_expand_builtin_string_cmp): Likewise.
15615         (inline_string_cmp): Likewise.  Use c_readstr instead of
15616         builtin_memcpy_read_str.  Add unit_mode temporary.
15618 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15620         PR middle-end/86528
15621         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
15622         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
15624 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15626         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
15627         Alphabetize prototypes of built-in functions, separating out
15628         built-in functions that are listed in this section but should be
15629         described elsewhere.
15631 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
15633         PR target/86511
15634         * expmed.c (emit_store_flag): Do not emit setcc followed by a
15635         conditional move when trapping comparison was split to a
15636         non-trapping one (and vice versa).
15638 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
15640         * config/s390/s390.c (s390_function_profiler): Generate nops
15641         instead of profiler call sequences.
15642         * config/s390/s390.opt: Add the new option.
15644 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
15646         * config/s390/s390.c (s390_function_profiler): Generate
15647         __mcount_loc section.
15648         * config/s390/s390.opt: Add the new option.
15650 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
15652         * common.opt: Add the new warning.
15653         * config/s390/s390.c (s390_function_profiler): Emit "brasl
15654         %r0,__fentry__" when -mfentry is specified.
15655         (s390_option_override_internal): Disallow -mfentry for 31-bit
15656         CPUs.
15657         * config/s390/s390.opt: Add the new option.
15659 2018-07-16  Richard Biener  <rguenther@suse.de>
15661         PR lto/86523
15662         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
15663         for function-local FUNCTION_DECL and RESULT_DECL immediately.
15665 2018-07-16  Martin Liska  <mliska@suse.cz>
15667         PR ipa/86529
15668         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
15669         to ::get_create.
15671 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
15673         * config/arc/arcHS.md: Update ARCHS scheduling rules.
15675 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
15677         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
15678         for ARCHS4x.
15679         * config/arc/arc-cpus.def (hs4x): New cpu.
15680         (hs4xd): Likewise.
15681         * config/arc/arc-tables.opt: Regenerate.
15682         * config/arc/arc.c (arc_sched_issue_rate): New function.
15683         (TARGET_SCHED_ISSUE_RATE): Define.
15684         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
15685         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
15686         fpu_cvt.
15687         (attr tune): Add ARCHS4x tune values.
15688         (attr tune_dspmpy): Define.
15689         (*tst): Correct instruction type.
15690         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
15691         * config/arc/arcHS4x.md: New file.
15692         * config/arc/fpu.md: Update instruction type attributes.
15693         * config/arc/t-multilib: Regenerate.
15695 2018-07-16  Tom de Vries  <tdevries@suse.de>
15697         PR debug/86455
15698         * var-tracking.c (vt_initialize): Fix pre_dec handling.
15700 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
15702         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
15703         early clobber.
15705 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
15707         PR tree-optimization/86514
15708         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
15709         conversion to a boolean type from a type with greater precision.
15711 2018-07-16  Tom de Vries  <tdevries@suse.de>
15713         * var-tracking.c (vt_initialize): Print adjusted insn slim if
15714         dump_flags request TDF_SLIM.
15716 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
15718         * fold-const.c (int_const_binop_1): Abstract...
15719         (wide_int_binop): ...wide int code here.
15720         (poly_int_binop): ...poly int code here.
15721         Abstract the rest of int_const_binop_1 into int_const_binop.
15722         * fold-const.h (wide_int_binop): New.
15723         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
15724         Remove useless PLUS/MINUS_EXPR case.
15725         (zero_nonzero_bits_from_vr): Move wide int code...
15726         (zero_nonzero_bits_from_bounds): ...here.
15727         (extract_range_from_binary_expr_1): Move mask optimization code...
15728         (range_easy_mask_min_max): ...here.
15729         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
15730         (range_easy_mask_min_max): New.
15732 2018-07-15  Jeff Law  <law@redhat.com>
15734         PR target/85993
15735         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
15736         block.
15738 2018-07-14  Jim Wilson  <jimw@sifive.com>
15740         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
15742 2018-07-14  Paul Koning  <ni1d@arrl.net>
15744         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
15746 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
15748         * lto-streamer-out.c (copy_function_or_variable): Dump info about
15749         copying section.
15751 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
15752             Steve Munroe  <munroesj52@gmail.com>
15754         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
15755         (_mm_andnot_si128): Likewise.
15756         (_mm_or_si128): Likewise.
15757         (_mm_xor_si128): Likewise.
15759 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
15761         PR middle-end/78809
15762         * builtins.c (expand_builtin_memcmp): Inline the calls first
15763         when result_eq is false.
15764         (expand_builtin_strcmp): Inline the calls first.
15765         (expand_builtin_strncmp): Likewise.
15766         (inline_string_cmp): New routine. Expand a string compare
15767         call by using a sequence of char comparison.
15768         (inline_expand_builtin_string_cmp): New routine. Inline expansion
15769         a call to str(n)cmp/memcmp.
15770         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
15771         New option.
15772         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
15774 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
15776         * config/arm/driver-arm.c: Include arm-native.h.
15777         (host_detect_local_cpu): Use auto-generated data tables.
15778         (vendors, arm_cpu_table): Delete.  Move part information to ...
15779         * config/arm/arm-cpus.in: ... here.
15780         * config/arm/parsecpu.awk (gen_native): New function.
15781         (vendor, part): New CPU fields.
15782         (END): Add support for building the native CPU detection tables.
15783         * config/arm/t-arm (arm-native.h): Add build rule.
15784         (driver-arm.o): Add dependency on arm-native.h.
15786 2018-07-13  Richard Biener  <rguenther@suse.de>
15788         PR middle-end/85974
15789         * match.pd (addr1 - addr2): Allow either of the operand to
15790         have a conversion.
15792 2018-07-13  Tom de Vries  <tdevries@suse.de>
15794         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
15795         in remap_ssa_name.
15797 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
15799         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
15800         arrays instead of numbered variables.
15802 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
15804         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
15805         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
15807 2018-07-13  Richard Biener  <rguenther@suse.de>
15809         PR debug/86452
15810         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
15811         instead of get_context_die.
15813 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
15814             Richard Biener  <rguenther@suse.de>
15816         PR middle-end/86489
15817         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
15818         that the loop latch destination where phi is defined.
15820 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
15822         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
15823         (riscv_expand_epilogue): Add assertion to check interrupt mode.
15824         (riscv_set_current_function): Extract getting interrupt type to new
15825         function.
15826         (riscv_get_interrupt_type): New function.
15827         (riscv_merge_decl_attributes): New function, checking interrupt type is
15828         same.
15829         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
15831 2018-07-12  Paul Koning  <ni1d@arrl.net>
15833         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
15834         directive.
15836 2018-07-12  Paul Koning  <ni1d@arrl.net>
15838         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
15839         zero reference, add doloop_end instead.
15840         * doc/md.texi (decrement_and_branch_until_zero): Remove.
15841         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
15842         detail for doloop_end.
15844 2018-07-12  Martin Sebor  <msebor@redhat.com>
15846         PR c/86453
15847         * attribs.c (decl_attributes): Reject conflicting attributes before
15848         calling attribute handlers.
15850 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15852         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
15853          parameter.
15854         (gcc::dump_manager::get_dump_file_name): likewise.
15855         (dump_begin): Likewise.
15856         * dumpfile.h (dump_begin): Update prototype.
15857         (gcc::dump_manager::get_dump_file_name,
15858         gcc::dump_manager::get_dump_file_name): Update prototype.
15860 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15862         * internal-fn.h (vectorizable_internal_fn_p): New function.
15863         * tree-vect-slp.c (compatible_calls_p): Likewise.
15864         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
15865         to internal functions.
15866         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
15868 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15870         * fold-const.h (inverse_conditions_p): Declare.
15871         * fold-const.c (inverse_conditions_p): New function.
15872         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
15873         that test the inverse condition of a conditional internal function.
15874         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
15875         * internal-fn.c (internal_fn_mask_index): Handle conditional
15876         internal functions.
15877         (vectorized_internal_fn_supported_p): New function.
15878         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
15879         (any_pred_load_store): Replace with...
15880         (need_to_predicate): ...this new variable.
15881         (redundant_ssa_names): New variable.
15882         (ifcvt_can_use_mask_load_store): Move initial checks to...
15883         (ifcvt_can_predicate): ...this new function.  Handle tree codes
15884         for which a conditional internal function exists.
15885         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
15886         instead of ifcvt_can_use_mask_load_store.  Update after variable
15887         name change.
15888         (predicate_load_or_store): New function, split out from
15889         predicate_mem_writes.
15890         (check_redundant_cond_expr): New function.
15891         (value_available_p): Likewise.
15892         (predicate_rhs_code): Likewise.
15893         (predicate_mem_writes): Rename to...
15894         (predicate_statements): ...this.  Use predicate_load_or_store
15895         and predicate_rhs_code.
15896         (combine_blocks, tree_if_conversion): Update after above name changes.
15897         (ifcvt_local_dce): Handle redundant_ssa_names.
15898         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
15899         general conditional functions.
15900         * tree-vect-stmts.c (vectorizable_call): Likewise.
15902 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15903             Alan Hayward  <alan.hayward@arm.com>
15904             David Sherwood  <david.sherwood@arm.com>
15906         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
15907         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
15908         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
15909         plus and minus and convert them into IFN_COND_FMA-based sequences.
15910         (convert_mult_to_fma): Handle conditional plus and minus.
15912 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15914         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
15915         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
15916         (cond_fnms_optab): New optabs.
15917         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
15918         internal functions.
15919         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
15920         * internal-fn.h (get_conditional_internal_fn): Declare.
15921         (get_unconditional_internal_fn): Likewise.
15922         * internal-fn.c (cond_ternary_direct): New macro.
15923         (expand_cond_ternary_optab_fn): Likewise.
15924         (direct_cond_ternary_optab_supported_p): Likewise.
15925         (FOR_EACH_COND_FN_PAIR): Likewise.
15926         (get_conditional_internal_fn): New function.
15927         (get_unconditional_internal_fn): Likewise.
15928         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
15929         (gimple_match_op::gimple_match_op): Add a new overload for 5
15930         operands.
15931         (gimple_match_op::set_op): Likewise.
15932         (gimple_resimplify5): Declare.
15933         * genmatch.c (decision_tree::gen): Generate simplifications for
15934         5 operands.
15935         * gimple-match-head.c (gimple_simplify): Define an overload for
15936         5 operands.  Handle calls with 5 arguments in the top-level overload.
15937         (convert_conditional_op): Handle conversions from unconditional
15938         internal functions to conditional ones.
15939         (gimple_resimplify5): New function.
15940         (build_call_internal): Pass a fifth operand.
15941         (maybe_push_res_to_seq): Likewise.
15942         (try_conditional_simplification): Try converting conditional
15943         internal functions to unconditional internal functions.
15944         Handle 3-operand unconditional forms.
15945         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
15946         Define ternary equivalents of the current rules for binary conditional
15947         internal functions.
15948         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
15949         ternary operations.
15950         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
15951         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
15952         (optab): Handle them.
15953         (SVE_COND_FP_TERNARY): New int iterator.
15954         (sve_fmla_op, sve_fmad_op): New int attributes.
15955         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
15956         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
15957         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
15959 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15961         * target.def (preferred_else_value): New target hook.
15962         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
15963         * doc/tm.texi: Regenerate.
15964         * targhooks.h (default_preferred_else_value): Declare.
15965         * targhooks.c (default_preferred_else_value): New function.
15966         * internal-fn.h (conditional_internal_fn_code): Declare.
15967         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
15968         (get_conditional_internal_fn): Use it.
15969         (conditional_internal_fn_code): New function.
15970         * gimple-match.h (gimple_match_cond): New struct.
15971         (gimple_match_op): Add a cond member function.
15972         (gimple_match_op::gimple_match_op): Update all forms to take a
15973         gimple_match_cond.
15974         * genmatch.c (expr::gen_transform): Use the same condition as res_op
15975         for the suboperation, but don't specify a particular else_value.
15976         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
15977         (visit_nary_op, visit_reference_op_load): Pass
15978         gimple_match_cond::UNCOND to the gimple_match_op constructor.
15979         * gimple-match-head.c: Include tree-eh.h
15980         (convert_conditional_op): New function.
15981         (maybe_resimplify_conditional_op): Likewise.
15982         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
15983         (gimple_resimplify2): Likewise.
15984         (gimple_resimplify3): Likewise.
15985         (gimple_resimplify4): Likewise.
15986         (maybe_push_res_to_seq): Return null for conditional operations.
15987         (try_conditional_simplification): New function.
15988         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
15989         constructor.
15990         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
15991         IFN_COND_* call.
15992         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
15993         function.
15994         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
15996 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15998         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15999         DECL_FCONTEXT
16000         (hash_tree): Do not hash DECL_FCONTEXT
16001         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
16002         Do not stream DECL_FCONTEXT.
16003         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
16004         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
16006 2018-07-12  Richard Biener  <rguenther@suse.de>
16008         PR debug/86462
16009         * dwarf2out.c (gen_block_die): Only output blocks when they have
16010         at least one !DECL_IGNORED_P variable.
16012 2018-07-12  Richard Biener  <rguenther@suse.de>
16014         PR target/84829
16015         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
16016         Remove -mieee-fp handling.
16018 2018-07-12  Richard Biener  <rguenther@suse.de>
16020         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
16021         left-over from last patch.
16023 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
16025         PR tree-optimization/86492
16026         * gimple-ssa-store-merging.c
16027         (imm_store_chain_info::coalesce_immediate_stores): Call
16028         check_no_overlap even for the merge_overlapping case.  Formatting fix.
16030 2018-07-12  Richard Biener  <rguenther@suse.de>
16032         PR middle-end/86479
16033         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
16034         move possibly trapping operations into the conditional.
16036 2018-07-12  Richard Biener  <rguenther@suse.de>
16038         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
16039         (vn_lookup_simplify_result): Remove recursion limit applied
16040         here.
16041         (vn_nary_build_or_lookup_1): Adjust.
16042         (try_to_simplify): Likewise.
16043         * gimple-match-head.c (gimple_resimplify1): Instead apply one
16044         here.
16045         (gimple_resimplify2): Likewise.
16046         (gimple_resimplify3): Likewise.
16047         (gimple_resimplify4): Likewise.
16049 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
16051         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
16052         Use __mmask64 type instead of __mmask8 for __M argument.
16053         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
16054         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
16055         __U argument.
16056         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
16057         __mmask16 for __M argument.
16058         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
16059         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
16060         to __mmask16 instead of __mmask8.
16061         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
16062         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
16063         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
16064         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
16065         instead of __mmask16 for __U argument.
16066         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
16067         __mmask16 instead of __mmask8 for __U argument.
16068         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
16069         __U argument.
16070         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
16071         __mmask16.
16072         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
16073         argument.
16074         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
16075         __U argument.
16076         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
16077         __mmask16.
16078         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
16079         of __mmask16.
16080         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
16081         __U argument.
16082         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
16083         __U argument.
16084         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
16085         __U argument.
16086         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
16087         __U argument.
16088         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
16089         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
16090         return type as well as __M argument type and all casts from __mmask8
16091         to __mmask32.
16092         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
16093         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
16094         return type as well as __M argument type and all casts from __mmask8
16095         to __mmask16.
16096         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
16097         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
16098         return type as well as __M argument type and all casts from __mmask8
16099         to __mmask32.
16100         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
16101         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
16102         return type as well as __M argument type and all casts from __mmask8
16103         to __mmask16.
16104         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
16105         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
16106         __mmask16.
16108 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
16110         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
16111         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
16112         for __U argument.
16114 2018-07-11  Paul Koning  <ni1d@arrl.net>
16116         * doc/md.texi (define_subst): Document how multiple occurrences of
16117         the same argument in the replacement pattern are handled.
16119 2018-07-11  Paul Koning  <ni1d@arrl.net>
16121         * doc/extend.texi (Common Variable Attributes): Move "mode" into
16122         alphabetical order.
16123         (Common Type Attributes): Add "mode" attribute.
16125 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
16127         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
16128         stream DECL_ORIGINAL_TYPE.
16129         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
16130         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
16131         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
16132         Do not walk original type.
16133         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
16134         external decls.
16135         (write_ts_decl_non_common_tree_pointers): Do not stream
16136         DECL_ORIGINAL_TYPE
16137         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
16138         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
16140 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
16142         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
16143         thread twice from the same starting edge.
16145 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
16147         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
16148         * gimple.c (gimple_call_nonnull_result_p): ...here...
16149         (gimple_call_nonnull_arg): ...and here.
16150         * gimple.h (gimple_call_nonnull_result_p): New.
16151         (gimple_call_nonnull_arg): New.
16153 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
16155         * config/arm/arm-cpus.in: Move information from fpu field of each
16156         cpu definition to the isa field.
16157         * config/arm/parsecpu.awk (fpu): Delete match rule.
16158         (gen_comm_data): Don't add bits from the CPU's FPU entry.
16160 2018-07-11  Richard Biener  <rguenther@suse.de>
16162         PR debug/86457
16163         * dwarf2out.c (init_sections_and_labels): Use
16164         output_asm_line_debug_info consistently.
16165         (dwarf2out_early_finish): Likewise.
16166         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
16167         type units.
16169 2018-07-11  Richard Biener  <rguenther@suse.de>
16171         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
16172         Rework father_bb setting in a way to avoid propagating constants
16173         multiple times on a loop body.
16175 2018-07-10  Mark Wielaard  <mark@klomp.org>
16177         PR debug/86459
16178         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
16180 2018-07-10  Richard Biener  <rguenther@suse.de>
16182         * hash-map.h (hash_map::iterator::operator*): Return
16183         references to key and value.
16185 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
16187         PR c++/86443
16188         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
16189         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
16190         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
16191         (gimplify_omp_for): For composite loops, move outer
16192         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
16193         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
16194         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
16195         TREE_LIST for both the original class iterator and the "last" helper
16196         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
16197         loop, remember has_decl_expr from outer composite loops for the
16198         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
16200 2018-07-09  Martin Sebor  <msebor@redhat.com>
16202         PR middle-end/77357
16203         PR middle-end/86428
16204         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
16205         accessing implicitly initialized array elements.
16206         * expr.c (string_constant): Handle string initializers of
16207         character arrays within aggregates.
16208         * gimple-fold.c (fold_array_ctor_reference): Add argument.
16209         Store element offset.  As a special case, handle zero size.
16210         (fold_nonarray_ctor_reference): Same.
16211         (fold_ctor_reference): Add argument.  Store subobject offset.
16212         * gimple-fold.h (fold_ctor_reference): Add argument.
16214 2018-07-09  Paul Koning  <ni1d@arrl.net>
16216         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
16217         (pdp11_insn_cost): New function.
16218         (pdp11_md_asm_adjust): New function.
16219         (TARGET_INVALID_WITHIN_DOLOOP): Define.
16220         (pdp11_rtx_costs): Update to match machine better.
16221         (output_addr_const_pdp11): Correct format mismatch warnings.
16222         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
16223         * config/pdp11/pdp11.md: General change to add base_cost and/or
16224         length attributes for use by new pdp11_insn_cost function.
16225         (MIN_BRANCH): Correct definition.
16226         (MIN_SOB): Ditto.
16227         (doloop_end): Use standard pattern name for looping pattern.
16228         (doloop_end_nocc): New.
16229         (movsf): Add another constraint alternative.
16230         (zero_extendqihi2): Add constraint alternatives for not in place
16231         extend.
16232         (zero_extendhisi2): Remove.
16233         (shift patterns): Add CC handling variants.
16234         (bswaphi2): New.
16235         (bswapsi2): New.
16236         (rothi3): New.
16237         (define_peephole2): New peephole to recognize mov that sets CC for
16238         subsequent test.
16240 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16242         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
16243         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
16244         wi::add.
16246 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
16248         PR c/86420
16249         * real.c (real_nextafter): Return true if result is denormal.
16251 2018-07-09  Martin Liska  <mliska@suse.cz>
16253         * common.opt: Add back wrongly removed attribute.
16255 2018-07-09  Richard Biener  <rguenther@suse.de>
16257         PR debug/86413
16258         * dwarf2out.c (gen_block_die): For an early generated DIE
16259         always output high/low PC attributes.
16261 2018-07-09  Tom de Vries  <tdevries@suse.de>
16263         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
16264         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
16265         onto VAR_DECL with abstract origin.
16267 2018-07-07  Jim Wilson  <jimw@sifive.com>
16269         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
16271 2018-07-07  Tom de Vries  <tdevries@suse.de>
16273         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
16275 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
16277         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
16278         overflow_type.
16279         (combine_bound): Use wide-int overflow calculation instead of
16280         rolling our own.
16281         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
16282         overflow_type.
16283         * fold-const.c (int_const_binop_2): Same.
16284         (extract_muldiv_1): Same.
16285         (fold_div_compare): Same.
16286         (fold_abs_const): Same.
16287         * match.pd: Same.
16288         * poly-int.h (add): Same.
16289         (sub): Same.
16290         (neg): Same.
16291         (mul): Same.
16292         * predict.c (predict_iv_comparison): Same.
16293         * profile-count.c (slow_safe_scale_64bit): Same.
16294         * simplify-rtx.c (simplify_const_binary_operation): Same.
16295         * tree-chrec.c (tree_fold_binomial): Same.
16296         * tree-data-ref.c (split_constant_offset_1): Same.
16297         * tree-if-conv.c (idx_within_array_bound): Same.
16298         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
16299         * tree-ssa-phiopt.c (minmax_replacement): Same.
16300         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
16301         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
16302         * vr-values.c (vr_values::adjust_range_with_scev): Same.
16303         * wide-int.cc (wi::add_large): Same.
16304         (wi::mul_internal): Same.
16305         (wi::sub_large): Same.
16306         (wi::divmod_internal): Same.
16307         * wide-int.h: Change overflow type to overflow_type for neg, add,
16308         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
16309         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
16310         mul_internal, divmod_internal.
16311         (overflow_type): New enum.
16312         (accumulate_overflow): New.
16314 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
16316         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
16317         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
16319 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
16321         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
16322         argument is checked for zero before entering loop, avoid checking again.
16324 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
16326         * gimplify.h (generic_expr_could_trap_p): Set as global function.
16327         * gimplify.h (generic_expr_could_trap_p): Likwise.
16328         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
16330 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
16332         PR tree-optimization/86401
16333         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
16334         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
16335         (fold_bit_and_mask): ... here.  New helper function for match.pd.
16336         * fold-const.h (fold_bit_and_mask): Declare.
16337         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
16339 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
16341         PR target/86324
16342         * target.def (translate_mode_attribute): New hook.
16343         * targhooks.h (default_translate_mode_attribute): Declare.
16344         * targhooks.c (default_translate_mode_attribute): New function.
16345         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
16346         * doc/tm.texi: Regenerate.
16347         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
16348         (rs6000_translate_mode_attribute): New function.
16350 2018-07-06  Paul Koning  <ni1d@arrl.net>
16352         * doc/md.texi (define_split): Document DONE and FAIL.
16353         (define_peephole2): Ditto.
16355 2018-07-05  Jeff Law  <law@redhat.com>
16357         PR tree-optimization/86010
16358         * tree-ssa-dse.c (compute_trims): More aggressively trim at
16359         both the head and tail of mem* and str* calls.
16361 2018-07-05  Jim Wilson  <jimw@sifive.com>
16363         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
16365 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
16367         * config/aarch64/aarch64-simd.md: correct flags text for
16368         MIN_EXPR replacement.
16370 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
16372         * configure: Regenerated.
16374 2018-07-05  Carl Love  <cel@us.ibm.com>
16376         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
16377         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
16378         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
16379         VSX_BUILTIN_DOUBLEL_V4SF.
16381 2018-07-05  Martin Sebor  <msebor@redhat.com>
16383         PR c++/86400
16384         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
16385         than its domain to compute its the upper bound of a char array.
16387 2018-07-05  Nathan Sidwell  <nathan@acm.org>
16389         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
16390         * doc/cpp.texi: Update comment.
16391         * doc/tm.texi: Rebuilt.
16392         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
16393         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
16394         * doc/extend.texi (Backwards Compatibility): Clarify it is system
16395         headers affected by extern "C".
16396         * system.h: Poison NO_IMPLICIT_EXTERN_C.
16397         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
16398         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
16399         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
16400         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
16401         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
16402         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
16403         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
16404         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
16405         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
16406         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
16407         config/sparc/sp64-elf.h, config/spu/spu.h,
16408         config/stormy16/stormy16.h, config/v850/v850.h,
16409         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
16410         define NO_IMPLICIT_EXTERN_C.
16411         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
16413 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
16415         PR target/84711
16416         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16417         instead of GET_MODE_SIZE when comparing Units.
16419 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
16421         PR target/84711
16422         * rtlanal.c (set_noop_p): Constrain on mode change,
16423         include hard-reg-set.h
16425 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
16427         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
16429 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
16431         Revert
16432         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
16434         PR sanitizer/84250
16435         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
16436         libasan.
16437         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
16439 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
16441         PR sanitizer/84250
16442         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
16443         libasan.
16444         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
16446 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
16448         PR middle-end/86380
16449         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
16451 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
16453         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
16454         neg_*_op* variables.
16456 2018-07-04  Martin Liska  <mliska@suse.cz>
16458         * tree-switch-conversion.c: Define
16459         max_ratio_for_speed and max_ratio_for_size constants.
16461 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
16462             Martin Liska  <mliska@suse.cz>
16464         PR middle-end/66240
16465         PR target/45996
16466         PR c/84100
16467         * common.opt: Rename align options with 'str_' prefix.
16468         * common/config/i386/i386-common.c (set_malign_value): New
16469         function.
16470         (ix86_handle_option): Use it to set -falign-* options/
16471         * config/aarch64/aarch64-protos.h (struct tune_params): Change
16472         type from int to string.
16473         * config/aarch64/aarch64.c: Update default values from int
16474         to string.
16475         * config/alpha/alpha.c (alpha_override_options_after_change):
16476         Likewise.
16477         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
16478         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16479         max skip conditionally.
16480         * config/i386/freebsd.h (SUBALIGN_LOG): New.
16481         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16482         max skip conditionally.
16483         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16484         max skip conditionally.
16485         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
16486         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16487         max skip conditionally.
16488         * config/i386/i386.c (struct ptt): Change type from int to
16489         string.
16490         (ix86_default_align): Set default values.
16491         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
16492         max skip conditionally.
16493         * config/i386/iamcu.h (SUBALIGN_LOG): New.
16494         (ASM_OUTPUT_MAX_SKIP_ALIGN):
16495         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
16496         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16497         max skip conditionally.
16498         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
16499         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
16500         * config/i386/x86-64.h (SUBALIGN_LOG): New.
16501         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16502         max skip conditionally.
16503         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
16504         * config/ia64/ia64.c (ia64_option_override): Set default values
16505         for alignment options.
16506         * config/m68k/m68k.c: Handle new str_align_* options.
16507         * config/mips/mips.c (mips_set_compression_mode): Change
16508         type of constants.
16509         (mips_option_override): Set default values for options.
16510         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
16511         Likewise.
16512         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16513         Likewise.
16514         * config/rx/rx.c (rx_option_override): Likewise.
16515         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
16516         (LABEL_ALIGN): Use align_labels_log.
16517         (LOOP_ALIGN): Use align_loops_align.
16518         * config/s390/s390.c (s390_asm_output_function_label): Use new
16519         macros.
16520         * config/sh/sh.c (sh_override_options_after_change):
16521         Change type of constants.
16522         * config/spu/spu.c (spu_sched_init): Likewise.
16523         * config/sparc/sparc.c (sparc_option_override): Set default
16524         values for options.
16525         * config/visium/visium.c (visium_option_override): Likewise.
16526         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
16527         emit p2align format with last argument if it's not needed.
16528         * doc/invoke.texi: Document extended format of -falign-*.
16529         * final.c: Use align_labels alignment.
16530         * flags.h (struct target_flag_state): Change type to use
16531         align_flags.
16532         (struct align_flags_tuple): New.
16533         (struct align_flags): Likewise.
16534         (align_loops_log): Redefine macro to use new types.
16535         (align_loops_max_skip): Redefine macro to use new types.
16536         (align_jumps_log): Redefine macro to use new types.
16537         (align_jumps_max_skip): Redefine macro to use new types.
16538         (align_labels_log): Redefine macro to use new types.
16539         (align_labels_max_skip): Redefine macro to use new types.
16540         (align_functions_log): Redefine macro to use new types.
16541         (align_loops): Redefine macro to use new types.
16542         (align_jumps): Redefine macro to use new types.
16543         (align_labels): Redefine macro to use new types.
16544         (align_functions): Redefine macro to use new types.
16545         (align_functions_max_skip): Redefine macro to use new types.
16546         (align_loops_value): New macro.
16547         (align_jumps_value): New macro.
16548         (align_labels_value): New macro.
16549         (align_functions_value): New macro.
16550         * function.c (invoke_set_current_function_hook): Propagate
16551         alignment values from flags to global variables default in
16552         topleev.h.
16553         * ipa-icf.c (sem_function::equals_wpa): Use
16554         cl_optimization_option_eq instead of memcmp.
16555         * lto-streamer.h (cl_optimization_stream_out): Support streaming
16556         of string types.
16557         (cl_optimization_stream_in): Likewise.
16558         * optc-save-gen.awk: Support strings in cl_optimization.
16559         * opth-gen.awk: Likewise.
16560         * opts.c (finish_options): Remove error checking of invalid
16561         value ranges.
16562         (MAX_CODE_ALIGN): Remove.
16563         (MAX_CODE_ALIGN_VALUE): Likewise.
16564         (parse_and_check_align_values): New function.
16565         (check_alignment_argument): Likewise.
16566         (common_handle_option): Use check_alignment_argument.
16567         * opts.h (parse_and_check_align_values): Declare.
16568         * toplev.c (init_alignments): Remove.
16569         (read_log_maxskip): New.
16570         (parse_N_M): Likewise.
16571         (parse_alignment_opts): Likewise.
16572         (backend_init_target): Remove usage of init_alignments.
16573         * toplev.h (parse_alignment_opts): Declare.
16574         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
16575         argument.
16576         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
16577         * tree.c (cl_option_hasher::equal): New.
16578         * varasm.c: Use new global macros.
16580 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
16582         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
16583         Use a simpler align directive also if MAXSKIP = ALIGN-1.
16584         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16585         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16586         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16587         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
16588         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
16589         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
16590         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
16591         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16592         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16593         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16594         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16596 2018-07-04  Martin Liska  <mliska@suse.cz>
16597             Jonathan Wakely  <jwakely@redhat.com>
16599         * coverage.c: Use correct type.
16600         * doc/invoke.texi: Language correction.
16602 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
16604         PR target/85620
16605         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
16606         ENDBRANCH for non-tail call which may return via indirect branch.
16607         * doc/extend.texi: Document indirect_return attribute.
16609 2018-07-03  Martin Sebor  <msebor@redhat.com>
16611         PR tree-optimization/86274
16612         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
16613         precondition.
16614         (format_floating): Correct handling of infinities and NaNs.
16616 2018-07-03  Martin Sebor  <msebor@redhat.com>
16618         * print-tree.c (print_real_cst): New function.
16619         (print_node_brief): Call it.
16620         (print_node): Ditto.
16622 2018-07-03  Jeff Law  <law@redhat.com>
16624         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
16625         into a single pattern.
16627         * config/h8300/h8300.md (ors code_iterator): New.
16628         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
16629         a single pattern and single splitter.
16630         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
16631         (iorqi3_1, xorqi3_1): Likewise.
16632         (iorqi3, xorqi3 expanders): Similarly.
16634         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
16635         (movmd_internal) into a single pattern using the P mode iterator.
16636         (movmd splitters): Similarly.
16637         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
16638         (movsd splitters): Similarly.
16640         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
16641         ADDB, ADDW and ADDL into a single ADD attribute which selects the
16642         right table based on the size of the operand.
16643         * config/h8300/h8300.md (length_table): Corresponding changes. All
16644         references to "addb", "addw" and "addl" changed to "add".
16645         (btst patterns): Merge two variants into a single pattern.
16646         (tstqi, tsthi): Likewise.
16647         (addhi3_incdec, addsi3_incdec): Likewise.
16648         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
16649         (mulhi3, mulsi3): Likewise.
16650         (udivhi3, udivsi3): Likewise.
16651         (divhi3, divsi3): Likewise.
16652         (andorqi3, andorhi3, andorsi3): Likewise.
16654 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
16656         PR target/85694
16657         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
16658         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
16660 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16662         PR tree-optimization/85694
16663         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
16664         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
16665         UNSPEC_URHADD.
16666         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
16667         (<u>avg<mode>3_ceil): New patterns.
16669 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
16671         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
16672         scan-tree-dump directive.
16673         * gcc.dg/vect/slp-perm-2.c: Likewise.
16674         * gcc.dg/vect/slp-perm-3.c: Likewise.
16675         * gcc.dg/vect/slp-perm-5.c: Likewise.
16676         * gcc.dg/vect/slp-perm-6.c: Likewise.
16677         * gcc.dg/vect/slp-perm-7.c: Likewise.
16678         * gcc.dg/vect/slp-perm-8.c: Likewise.
16680 2018-07-03  Marek Polacek  <polacek@redhat.com>
16682         PR middle-end/86202
16683         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
16685 2018-07-03  Richard Biener  <rguenther@suse.de>
16687         PR ipa/86389
16688         * tree-ssa-structalias.c (find_func_clobbers): Properly
16689         handle indirect calls.
16691 2018-07-03  Jeff Law  <law@redhat.com>
16693         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
16694         (shifts): New code iterator.
16695         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
16696         expander.  Fix HImode handling on H8/SX.
16697         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
16698         (subqi3, subhi3, subsi3 expanders): Likewise.
16699         (andqi3, andhi3, andsi3 expanders): Likewise.
16700         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
16701         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
16702         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
16703         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
16704         (zero_extendqihi2, zero_extendqisi2): Likewise.
16705         (extendqihi2, extendqisi2): Likewise.
16706         (rotlqi3, rotlhi3, rotlsi3): Likewise.
16707         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
16708         (rotlqi3_1, rotlhi3_1): Likewise.
16709         (logicalhi3_sn, logicalsi3_sn): Likewise.
16710         (logicalhi3, logicalsi3): Likewise.
16712 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16714         * tree-vect-patterns.c (vect_recog_rotate_pattern)
16715         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
16716         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
16717         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
16718         type to append_pattern_def_seq instead of creating a stmt_vec_info
16719         directly.
16720         (build_mask_conversion): Likewise.  Remove vinfo argument.
16721         (vect_add_conversion_to_patterm): Likewise, renaming to...
16722         (vect_add_conversion_to_pattern): ...this.
16723         (vect_recog_mask_conversion_pattern): Update call to
16724         build_mask_conversion.  Pass the vector type to
16725         append_pattern_def_seq here too.
16726         (vect_recog_gather_scatter_pattern): Update call to
16727         vect_add_conversion_to_pattern.
16729 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16731         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
16732         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16733         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
16734         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
16735         STMT_VINFO_PATTERN_DEF_SEQ to null here.
16736         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
16737         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
16738         append_pattern_def_seq instead of new_pattern_def_seq.
16739         (vect_recog_divmod_pattern): Do both of the above.
16740         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
16741         is null.
16743 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16745         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
16746         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
16747         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
16748         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
16749         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
16750         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
16751         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
16752         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16753         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16754         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
16755         parameter with a single stmt_vec_info.
16756         (vect_recog_func_ptr): Likewise.
16757         (vect_recog_gather_scatter_pattern): Likewise, folding in...
16758         (vect_try_gather_scatter_pattern): ...this.
16759         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
16760         the stmt_vec_info of the statement to be matched.  Don't clear
16761         STMT_VINFO_RELATED_STMT.
16762         (vect_pattern_recog): Update call accordingly.
16764 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16766         PR tree-optimization/85694
16767         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
16768         (uavgM3_ceil): Document new optabs.
16769         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
16770         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
16771         functions.
16772         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
16773         (savg_ceil_optab): New optabs.
16774         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
16775         (vect_vect_recog_func_ptrs): Add it.
16776         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
16777         constant directly from the associated lhs.
16779 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16781         * tree-vect-patterns.c (vect_split_statement): New function.
16782         (vect_convert_input): Use it to try to split an existing cast.
16784 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16786         * poly-int.h (print_hex): New function.
16787         * dumpfile.h (dump_dec, dump_hex): Declare.
16788         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
16789         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
16790         min_input_precision, operation_precision and operation_sign.
16791         * tree-vect-patterns.c (vect_get_range_info): New function.
16792         (vect_same_loop_or_bb_p, vect_single_imm_use)
16793         (vect_operation_fits_smaller_type): Delete.
16794         (vect_look_through_possible_promotion): Add an optional
16795         single_use_p parameter.
16796         (vect_recog_over_widening_pattern): Rewrite to use new
16797         stmt_vec_info infomration.  Handle one operation at a time.
16798         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
16799         (vect_truncatable_operation_p, vect_set_operation_type)
16800         (vect_set_min_input_precision): New functions.
16801         (vect_determine_min_output_precision_1): Likewise.
16802         (vect_determine_min_output_precision): Likewise.
16803         (vect_determine_precisions_from_range): Likewise.
16804         (vect_determine_precisions_from_users): Likewise.
16805         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
16806         (vect_vect_recog_func_ptrs): Put over_widening first.
16807         Add cast_forwprop.
16808         (vect_pattern_recog): Call vect_determine_precisions.
16810 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16812         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
16813         statements that have been replaced by further pattern statements.
16814         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
16816 2018-07-03  Richard Biener  <rguenther@suse.de>
16818         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
16819         always set *dt.  Dump vectype in vectype overload.
16820         * dumpfile.h (dump_gimple_expr): New function.
16821         (dump_gimple_expr_loc): Likewise.
16822         * dumpfile.c (dump_gimple_expr): New function.
16823         (dump_gimple_expr_loc): Likewise.
16825 2018-07-02  Jeff Law  <law@redhat.com>
16827         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
16828         the H8/300, H8/300H and H8/S variants into a single pattern.
16829         (movhi_h8300, movqi_h8300hs): Similarly.
16830         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
16831         (QHI mode iterator): New.
16833         * config/h8300/h8300.md: Remove trailing whitespace.
16835 2018-07-02  Jim Wilson  <jimw@sifive.com>
16837         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
16838         instead of emit_insn for interrupt returns.
16839         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
16840         (riscv_sret, riscv_uret): Likewise.
16842 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16844         * pretty-print.c (selftest::test_pp_format): Move save and restore
16845         of quotes to class auto_fix_quotes, and add an instance.
16846         * selftest.c: Include "intl.h".
16847         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
16848         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
16849         * selftest.h (selftest::auto_fix_quotes): New class.
16851 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
16853         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
16854         (aarch64_sve_prepare_conditional_op): Remove.
16855         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
16856         Allow aarch64_simd_reg_or_zero as select operand; remove
16857         the aarch64_sve_prepare_conditional_op call.
16858         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
16859         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16860         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
16861         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
16862         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
16863         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
16864         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
16865         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
16866         and a splitters to match all of the *_any patterns.
16867         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
16869         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
16870         (SVE_COND_FP_BINARY_REV): Remove.
16871         (sve_int_op_rev, sve_fp_op_rev): New.
16872         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
16873         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
16874         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
16875         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
16876         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
16877         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
16878         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
16879         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
16880         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
16882         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
16883         Remove match_dup 1 from the inner unspec.
16884         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16886         * config/aarch64/aarch64.md (movprfx): New attr.
16887         (length): Default movprfx to 8.
16888         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
16889         (*madd<SVE_I>, *msub<SVE_I): Likewise.
16890         (*<su>mul<SVE_I>3_highpart): Likewise.
16891         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
16892         (*v<ASHIFT><SVE_I>3): Likewise.
16893         (*<su><MAXMIN><SVE_I>3): Likewise.
16894         (*<su><MAXMIN><SVE_F>3): Likewise.
16895         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
16896         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
16897         (*div<SVE_F>4): Likewise.
16899 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
16901         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
16902         in dump string.
16904 2018-07-02  Richard Biener  <rguenther@suse.de>
16906         PR tree-optimization/86363
16907         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
16908         memset argument refers to a non-variable address.
16910 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16912         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
16913         {PLUS,MINUS}_EXPR code to...
16914         (adjust_symbolic_bound): ...here,
16915         (combine_bound): ...here,
16916         (set_value_range_with_overflow): ...and here.
16918 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16920         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
16921         code...
16922         (extract_range_from_abs_expr): ...here.
16924 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
16926         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
16927         -fno-omit-frame-pointer when not optimizing.
16929 2018-07-02  Martin Liska  <mliska@suse.cz>
16931         PR ipa/86279
16932         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
16933         (propagate_nothrow): Likewise.
16935 2018-07-02  Martin Liska  <mliska@suse.cz>
16937         PR ipa/86323
16938         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
16940 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16942         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
16943         function in r262149, changing "loc" param from source_location to
16944         const dump_location_t &.
16945         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
16946         declaration, as above.
16948 2018-07-01  Paul Koning  <ni1d@arrl.net>
16950         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
16951         -munit-asm, -mgnu-asm, -mdec-asm.
16952         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
16953         (pdp11_output_labelref): New.
16954         (pdp11_output_def): New.
16955         (pdp11_output_addr_vec_elt): New.
16956         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
16957         %# and %@ format codes.
16958         (pdp11_option_override): New.
16959         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
16960         (pdp11_output_ident): New.
16961         (pdp11_asm_named_section): New.
16962         (pdp11_asm_init_sections): New.
16963         (pdp11_file_start): New.
16964         (pdp11_file_end): New.
16965         (output_ascii): Use .ascii/.asciz for -mdec-asm.
16966         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
16967         %o, like %c but octal.
16968         (pdp11_option_override): New.
16969         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
16970         -mdec-asm.
16971         (DATA_SECTION_ASM_OP): Ditto.
16972         (READONLY_DATA_SECTION_ASM_OP): New.
16973         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
16974         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
16975         (ASM_OUTPUT_LABELREF): Ditto.
16976         (ASM_OUTPUT_DEF): Ditto.
16977         (ASM_OUTPUT_EXTERNAL): New.
16978         (ASM_OUTPUT_SOURCE_FILENAME): New.
16979         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
16980         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
16981         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
16982         %# and %@ format codes.
16983         * config/pdp11/pdp11.opt (mgnu-asm): New.
16984         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
16985         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
16986         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
16988 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
16990         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
16991         dereferencing path[] beyond its length.
16992         (debug_path): New.
16993         (debug_all_paths): New.
16994         (rewire_first_differing_edge): New.
16995         (adjust_paths_after_duplication): New.
16996         (duplicate_thread_path): Call adjust_paths_after_duplication.
16997         Add new argument.
16998         (thread_through_all_blocks): Add new argument to
16999         duplicate_thread_path.
17001 2018-06-30  Jim Wilson  <jimw@sifive.com>
17003         * config/riscv/predicates.md (p2m1_shift_operand): New.
17004         (high_mask_shift_operand): New.
17005         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
17006         pattern using p2m1_shift_operand.
17007         (lshsi3_zero_extend_3+2): New combiner pattern using
17008         high_mask_shift_operand.
17010 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17012         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
17013         split out from...
17014         (vect_recog_rotate_pattern): ...here.
17015         (vect_convert_input): Try to insert casts of invariants in the
17016         preheader.
17017         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
17018         preheader to be empty.
17020 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17022         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
17023         vector type.  If given, install it in the new statement's
17024         STMT_VINFO_VECTYPE.
17025         (vect_element_precision): New function.
17026         (vect_unpromoted_value): New struct.
17027         (vect_unpromoted_value::vect_unpromoted_value): New function.
17028         (vect_unpromoted_value::set_op): Likewise.
17029         (vect_look_through_possible_promotion): Likewise.
17030         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
17031         (vect_widened_op_tree, vect_convert_input): Likewise.
17032         (vect_convert_inputs, vect_convert_output): Likewise.
17033         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
17034         to handle the optional cast of the multiplication result and
17035         vect_widened_op_tree to detect the widened multiplication itself.
17036         Do not require the input and output of promotion casts to have
17037         the same sign, but base the signedness of the operation on the
17038         input rather than the result.  If the pattern includes two
17039         promotions, check that those promotions have the same sign.
17040         Do not restrict the MULT_EXPR handling to a double-width result;
17041         handle quadruple-width results and wider.  Use vect_convert_inputs
17042         to convert the inputs to the common type.
17043         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
17044         to handle the optional cast of the ABS result.  Also allow a sign
17045         change or a sign extension between the ABS and MINUS.
17046         Use vect_widened_op_tree to detect the widened subtraction and use
17047         vect_convert_inputs to convert the inputs to the common type.
17048         (vect_handle_widen_op_by_const): Delete.
17049         (vect_recog_widen_op_pattern): New function.
17050         (vect_recog_widen_mult_pattern): Use it.
17051         (vect_recog_widen_shift_pattern): Likewise.
17052         (vect_recog_widen_sum_pattern): Use
17053         vect_look_through_possible_promotion to handle the promoted
17054         PLUS_EXPR operand.
17056 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17058         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
17059         the containing gimple_seq *.
17060         * gimple-iterator.h (gsi_for_stmt): Declare it.
17061         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
17062         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
17063         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17064         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17065         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
17066         checks.
17067         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
17068         split out from...
17069         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
17070         statement being replaced is part of an existing pattern
17071         definition sequence, inserting the new pattern statements before
17072         the original one.
17073         (vect_pattern_recog_1): Don't return a bool.  If the statement
17074         is already part of a pattern, instead apply pattern matching
17075         to the pattern definition statements.  Don't clear the
17076         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
17077         (vect_pattern_recog): Don't break after the first match;
17078         continue processing the pattern definition statements instead.
17079         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
17081 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17083         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17084         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17085         (vect_recog_widen_sum_pattern): Use it.
17087 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17089         * tree-vect-loop.c (vectorizable_reduction): Assert that the
17090         phi is not a pattern statement and has not been replaced by
17091         a pattern statement.
17092         * tree-vect-patterns.c (type_conversion_p): Don't check
17093         STMT_VINFO_IN_PATTERN_P.
17094         (vect_recog_vector_vector_shift_pattern): Likewise.
17095         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
17096         the pattern statement rather than the original statement; check
17097         directly for a WIDEN_MULT_EXPR here.
17098         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
17099         vect_is_simple_use to return the pattern statement rather
17100         than the original statement; use is_pattern_stmt_p to check
17101         for such a pattern statement.
17102         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
17103         to return the pattern statement rather than the original statement;
17104         don't do the same transformation here.
17105         (vect_is_simple_use): If the defining statement has been replaced
17106         by a pattern statement, return the pattern statement instead.
17107         Remove the corresponding (local) transformation from the vectype
17108         overload.
17110 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
17112         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
17113         end and default to null.
17114         * tree-vect-loop.c (vect_create_epilog_for_reduction)
17115         (vectorizable_reduction): Update calls accordingly, dropping the
17116         gimple ** argument if the passed-back statement isn't needed.
17117         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
17118         (vect_recog_rotate_pattern): Likewise.
17119         (vect_recog_mask_conversion_pattern): Likewise.
17120         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
17121         (vect_mask_constant_operand_p): Likewise.
17122         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
17123         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
17124         (get_group_load_store_type, get_load_store_type): Likewise.
17125         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
17126         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
17127         (vectorizable_conversion, vectorizable_assignment): Likewise.
17128         (vectorizable_shift, vectorizable_operation): Likewise.
17129         (vectorizable_store, vect_is_simple_cond): Likewise.
17130         (vectorizable_condition, vectorizable_comparison): Likewise.
17131         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
17132         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
17133         and move it to the end.  Cope with null def_stmt_outs.
17135 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17137         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
17139 2018-06-29  Jeff Law  <law@redhat.com>
17141         * config/v850/v850.c (v850_legitimate_address_p): Handle large
17142         displacements for TARGET_V850E2V3 and newer.
17143         (TARGET_LRA_P): Remove.  Defaults to LRA now.
17144         * config/v850/v850.md (sign23byte_load): Remove.
17145         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
17146         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
17148 2018-06-29  Martin Liska  <mliska@suse.cz>
17150         PR lto/85759
17151         * coverage.c (coverage_init): Mangle full path name.
17152         * doc/invoke.texi: Document the change.
17153         * gcov-io.c (mangle_path): New.
17154         * gcov-io.h (mangle_path): Likewise.
17155         * gcov.c (mangle_name): Use mangle_path for path mangling.
17157 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17159         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
17160         if starting source register is not even.
17162 2018-06-29  Martin Liska  <mliska@suse.cz>
17164         PR tree-optimization/86263
17165         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
17166         Make edge redirection.
17168 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
17170         * dumpfile.c (dump_loc): Add indentation based on scope depth.
17171         (dump_scope_depth): New variable.
17172         (get_dump_scope_depth): New function.
17173         (dump_begin_scope): New function.
17174         (dump_end_scope): New function.
17175         * dumpfile.h (get_dump_scope_depth): New declaration.
17176         (dump_begin_scope): New declaration.
17177         (dump_end_scope): New declaration.
17178         (class auto_dump_scope): New class.
17179         (AUTO_DUMP_SCOPE): New macro.
17180         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
17181         AUTO_DUMP_SCOPE.
17183 2018-06-29  Richard Biener  <rguenther@suse.de>
17185         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
17186         compute_all_dependences succeeds.
17187         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
17188         exceed --param loop-max-datarefs-for-datadeps.
17190 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
17192         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
17194 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
17196         PR target/86348
17197         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
17198         alternative 0 in preferred_for_speed attribute.
17200 2018-06-28  Paul Koning  <ni1d@arrl.net>
17202         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
17203         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
17204         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
17205         * config/pdp11/pdp11.md: Correct "length" attribute calculation
17206         for shift insn patterns.
17208 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
17210         * cgraph.c (cgraph_node::get_body): Replace assignments to
17211         "dump_file" with calls to set_dump_file.
17212         * dumpfile.c (alt_dump_file): Make static, and group with...
17213         (alt_flags): ...this definition.
17214         (dumps_are_enabled): New variable.
17215         (refresh_dumps_are_enabled): New function.
17216         (set_dump_file): New function.
17217         (set_alt_dump_file): New function.
17218         (gcc::dump_manager::dump_start): Replace assignments to
17219         "dump_file" and "alt_dump_file" with calls to set_dump_file and
17220         set_alt_dump_file.
17221         (gcc::dump_manager::dump_finish): Likewise.
17222         * dumpfile.h (alt_dump_file): Delete decl.
17223         (dumps_are_enabled): New variable decl.
17224         (set_dump_file): New function decl.
17225         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
17226         global.
17227         * tree-nested.c (lower_nested_functions): Replace assignments to
17228         "dump_file" with calls to set_dump_file.
17230 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
17232         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
17233         goto_locus of each outgoing edge of each basic block.
17235 2018-06-28  Richard Biener  <rguenther@suse.de>
17237         * dwarf2out.c (decl_scope_table): Remove.
17238         (push_decl_scope): Likewise.
17239         (pop_decl_scope): Likewise.
17240         (gen_type_die_for_member): Do not call push/pop_decl_scope.
17241         (gen_struct_or_union_type_die): Likewise.
17242         (gen_tagged_type_die): Likewise.
17243         (dwarf2out_init): Do not initialize decl_scope_table.
17244         (dwarf2out_c_finalize): Do not free it.
17246 2018-06-28  Richard Biener  <rguenther@suse.de>
17248         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
17249         deciding whether to not re-use a DIE.
17251 2018-06-28  Richard Biener  <rguenther@suse.de>
17253         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
17254         DW_AT_abstract_origin attribute.
17256 2018-06-28  Martin Liska  <mliska@suse.cz>
17258         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
17259         Use newly introduced constants.
17260         * tree-switch-conversion.h (struct jump_table_cluster):
17261         Define max_ratio_for_size and max_ratio_for_speed.
17263 2018-06-28  Martin Liska  <mliska@suse.cz>
17265         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17266         Add new checking assert to catch invalid state.
17267         (jump_table_cluster::can_be_handled): Handle single case
17268         clusters.
17269         (jump_table_cluster::is_beneficial): Bail out for such case.
17270         (bit_test_cluster::find_bit_tests):
17271         Add new checking assert to catch invalid state.
17272         (bit_test_cluster::can_be_handled): Handle single case
17273         clusters.
17274         (bit_test_cluster::is_beneficial): Bail out for such case.
17275         (switch_decision_tree::analyze_switch_statement):
17276         Fix comment.
17278 2018-06-28  Martin Liska  <mliska@suse.cz>
17280         * common.opt: Introduce -completion option.
17281         * gcc.c (driver_handle_option): Handle it.
17282         (driver::main): Print completions if completion
17283         is set.
17284         * opt-suggestions.c (option_proposer::get_completions):
17285         New function.
17286         (option_proposer::suggest_completion): Likewise.
17287         (option_proposer::find_param_completions): Likewise.
17288         (verify_autocompletions): Likewise.
17289         (test_completion_valid_options): Likewise.
17290         (test_completion_valid_params): Likewise.
17291         (in_completion_p): Likewise.
17292         (empty_completion_p): Likewise.
17293         (test_completion_partial_match): Likewise.
17294         (test_completion_garbage): Likewise.
17295         (opt_proposer_c_tests): Likewise.
17296         * opt-suggestions.h: Declare new functions.
17297         * opts.c (common_handle_option): Handle OPT__completion_.
17298         * selftest-run-tests.c (selftest::run_tests): Add
17299         opt_proposer_c_tests.
17300         * selftest.c (assert_str_startswith): New.
17301         * selftest.h (assert_str_startswith): Likewise.
17302         (opt_proposer_c_tests): New.
17303         (ASSERT_STR_STARTSWITH): Likewise.
17305 2018-06-28  Martin Liska  <mliska@suse.cz>
17307         * Makefile.in: Add opt-suggestions.o.
17308         * gcc-main.c: Include opt-suggestions.h.
17309         * gcc.c (driver::driver): Likewise.
17310         (driver::~driver): Remove m_option_suggestions.
17311         (driver::build_option_suggestions): Moved to option_proposer.
17312         (driver::suggest_option): Likewise.
17313         (driver::handle_unrecognized_options): Use option_proposer.
17314         * gcc.h (class driver): Add new memver m_option_proposer.
17315         * opt-suggestions.c: New file.
17316         * opt-suggestions.h: New file.
17318 2018-06-28  Martin Liska  <mliska@suse.cz>
17320         * vec.h (class auto_string_vec): New (moved from auto_argvec).
17321         (auto_string_vec::~auto_string_vec): Likewise.
17323 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
17325         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
17326         prevent_decl_creation_for_types fields up and add reset_location field.
17327         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
17328         statement if id->reset_location is true.
17329         (copy_edges_for_bb): Do not set goto_locus on the new edges if
17330         id->reset_location is true.
17331         (copy_phis_for_bb): Force input_location on the arguments if
17332         id->reset_location is true.
17333         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
17334         is set on the function to be inlined.
17336 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
17338         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
17340 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
17342         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
17343         registers for Pmode.
17344         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
17345         hard registers for the clobbered pseudo.
17347 2018-06-27  Paul Koning  <ni1d@arrl.net>
17349         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17350         mutually exclusive options.
17351         * config/pdp11/constraints.md (h): New constraint.
17352         (O): Update definition to match shift code generation.
17353         (D): New constraint.
17354         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
17355         (CCFP): Remove.
17356         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
17357         function.
17358         (output_jump): Change arguments.
17359         (pdp11_fixed_cc_regs): New function.
17360         (pdp11_cc_mode): Ditto.
17361         (pdp11_expand_shift): Ditto.
17362         (pdp11_assemble_shift): Ditto.
17363         (pdp11_small_shift): Ditto.
17364         (pdp11_branch_cost): Remove.
17365         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
17366         from output.
17367         (pdp11_register_move_cost): Update for CC registers.
17368         (pdp11_rtx_costs): Add case for LSHIFTRT.
17369         (pdp11_output_jump): Add CCNZ mode conditional branches.
17370         (notice_update_cc_on_set): Remove.
17371         (pdp11_cc_mode): New function.
17372         (simple_memory_operand): Correct pre/post decrement case.
17373         (no_side_effect_operand): New function.
17374         (pdp11_regno_reg_class): Add CC_REGS class.
17375         (pdp11_fixed_cc_regs): New function.
17376         (pdp11_small_shift): New function.
17377         (pdp11_expand_shift): New function to expand shift insns.
17378         (pdp11_assemble_shift): New function to output shifts.
17379         (pdp11_branch_cost): Remove.
17380         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
17381         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
17382         (WCHAR_TYPE): Ditto.
17383         (PTRDIFF_TYPE): Ditto.
17384         (ADJUST_INSN_LENGTH): New macro.
17385         (FIXED_REGISTERS): Add CC registers.
17386         (CALL_USED_REGISTERS): Ditto.
17387         (reg_class): Ditto.
17388         (REG_CLASS_NAMES): Ditto.
17389         (REG_CLASS_CONTENTS): Ditto.
17390         (SELECT_CC_MODE): Use new function.
17391         (TARGET_FLAGS_REGNUM): New macro.
17392         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
17393         (cc0_reg_rtx): Remove.
17394         (CC_STATUS_MDEP): Remove.
17395         (CC_STATUS_MDEFP_INIT): Remove.
17396         (CC_IN_FPU): Remove.
17397         (NOTICE_UPDATE_CC): Remove.
17398         (REGISTER_NAMES): Add CC registers.
17399         (BRANCH_COST): Change to constant 1.
17400         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
17401         handling.
17402         * config/pdp11/pdp11.opt (mbcopy): Remove.
17403         (mbcopy-builtin): Remove.
17404         (mbranch-cheap): Remove.
17405         (mbranch-expensive): Remove.
17406         * config/pdp11/predicates.md (expand_shift_operand): Update to
17407         match shift code generation.
17408         (ccnz_operator): New predicate.
17409         * doc/invoke.texi (PDP-11 Options): Remove deleted options
17410         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
17411         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
17412         exclusive options.
17413         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
17414         description of O constraint.
17416 2018-06-27  Jeff Law  <law@redhat.com>
17417             Austin Law  <austinklaw@gmail.com>
17419         * config/v850/v850.md (addsi3_set_flags): New pattern.
17420         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
17421         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
17422         (zero_extendhisi2_v850_set_flags): Likewise.
17423         (zero_extendqisi2_v850_set_flags): Likewise.
17424         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
17425         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
17426         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
17428         * config/v850/v850-protos.h (notice_update_cc): Remove.
17429         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
17430         (v850_print_operand): Handle 'D' and "d".
17431         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
17432         Add handling of arithmetic/logical operations compared against zero.
17433         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
17434         Do not look at v850_compare_op, instead get mode from last argument.
17435         (v850_gen_compare): Remove
17436         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
17437         after reload for prologue insns.
17438         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
17439         patterns.
17440         (construct_save_jarl): Likewise.
17441         (TARGET_FLAGS_REGNUM): Define.
17442         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
17443         (NOTICE_UPDATE_CC): Remove.
17444         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
17445         than cc0.  Conditionalize on reload_completed.
17446         (cmpsi_insn, setfcc_insn): Likewise.
17447         (tst1 splitter): Turn into define_and_split which sets the flags
17448         after reload.
17449         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
17450         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
17451         (cstoresf4, cstoredf4): Clobber the flags.
17452         (cmpsi, cmpsf, cmpdf): Remove expanders.
17453         (setf_insn): Remove pattern.
17454         (addsi3): Turn into define_and_split which clobbers the flags after
17455         reload and a suitable pattern (addsi3_clobber_flags) for use after
17456         reload.
17457         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
17458         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
17459         (ashrsi3, ashrsi3_v850e2): Likewise.
17460         (bins): Clobber the flags.
17461         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
17462         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
17463         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
17464         (call_value_internal_short, call_value_internal_long): Likewise.
17465         (callt_save_interrupt, callt_return_interrupt): Likewise.
17466         (save_interrupt, return_interrupt): Likewise.
17467         (callt_save_all_interrupt, save_all_interrupt): Likewise.
17468         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
17469         (restore_all_interrupt, _restore_all_interrupt): Likewise.
17470         (All FP comparisons): Only allow after reload has completed.
17471         (trfsr): Likewise.
17472         (divh, divhu): Tweak output template.
17473         (branch_z_normal, branch_z_invert): Remove
17474         (branch_nz_normal, branch_nz_invert): Likewise.
17475         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
17477         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
17478         * config/v850/v850.c (notice_update_cc): Remove.
17479         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
17480         (CC_NO_CARRY): Likewise.
17481         (NOTICE_UPDATE_CC): Define to nothing.
17482         * config/v850/v850.md: Remove block comment on cc0 handling
17483         Remove "cc" attribute from all patterns.  Remove cc_status handling
17484         from all patterns.  Minor formatting fixes.
17486 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17488         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
17489         (cortex-a76.cortex-a55): Likewise.
17490         * config/aarch64/aarch64-tune.md: Regenerate.
17491         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
17492         cortex-a76.cortex-a55.
17494 2018-06-27  Jeff Law  <law@redhat.com>
17496         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
17497         (MULTILIB_DIRNAMES): Similarly.
17499 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
17501         * gimple.h (gimple_return_retbnd): Delete.
17502         (gimple_return_set_retbnd): Likewise.
17503         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
17504         gimple_return_set_retbnd.
17505         * gimple-pretty-print.c (dump_gimple_return): Remove call to
17506         gimple_return_retbnd and adjust.
17507         * tree-inline.h (struct copy_body_data): Remove retbnd field.
17508         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
17509         Explicitly return NULL in a couple more cases.  Move assertion
17510         on debug statements and remove unreachable code.
17511         (reset_debug_binding): Do not test id->retbnd.
17512         (expand_call_inline): Do not set it.
17514 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
17516         * configure.ac: Add --disable-gcov option.
17517         * configure: Regenerate.
17518         * Makefile.in: Honour @enable_gcov@.
17519         * doc/install.texi: Document --disable-gcov.
17521 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17523         * config/arm/arm-cpus.in (cortex-a76): New entry.
17524         (cortex-a76.cortex-a55): Likewise.
17525         * config/arm/arm-tables.opt: Regenerate.
17526         * config/arm/arm-tune.md: Likewise.
17527         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
17528         * doc/invoke.texi (ARM Options): Document cortex-a76 and
17529         cortex-a76.cortex-a55.
17531 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
17533         PR target/85769
17534         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
17536 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
17538         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
17539         comment.
17540         (EPILOGUE_USES): Likewise.
17542 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
17544         * tree-inline.c (remap_location): New function extracted from...
17545         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
17546         (copy_phis_for_bb): ...here.  Call remap_location.
17547         (copy_cfg_body): Adjust call to copy_edges_for_bb.
17549 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
17551         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
17552         unaligned vsx for 16B memset.
17554 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
17556         PR target/86285
17557         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
17558         ieee128_float_type_node to long_double_type_node unless
17559         TARGET_LONG_DOUBLE_128 is set.
17561 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
17563         * cfgloop.c (get_loop_location): Convert return type from
17564         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
17565         by implicit construction from rtx_insn *, and using
17566         dump_user_location_t::from_function_decl for the fallback case.
17567         * cfgloop.h (get_loop_location): Convert return type from
17568         location_t to dump_user_location_t.
17569         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
17570         dump_printf_loc to pass in a dump_location_t rather than a
17571         location_t, via the gimple stmt.
17572         * coverage.c (get_coverage_counts): Update calls to
17573         dump_printf_loc to pass in dump_location_t rather than a
17574         location_t.
17575         * doc/optinfo.texi (Dump types): Convert example of
17576         dump_printf_loc from taking "locus" to taking "insn".  Update
17577         description of the "_loc" calls to cover dump_location_t.
17578         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
17579         "selftest.h".
17580         (dump_user_location_t::dump_user_location_t): New constructors,
17581         from gimple *stmt and rtx_insn *.
17582         (dump_user_location_t::from_function_decl): New function.
17583         (dump_loc): Make static.
17584         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
17585         const dump_location_t &.
17586         (dump_generic_expr_loc): Delete.
17587         (dump_printf_loc): Convert param "loc" from location_t to
17588         const dump_location_t &.
17589         (selftest::test_impl_location): New function.
17590         (selftest::dumpfile_c_tests): New function.
17591         * dumpfile.h: Include "profile-count.h".
17592         (class dump_user_location_t): New class.
17593         (struct dump_impl_location_t): New struct.
17594         (class dump_location_t): New class.
17595         (dump_printf_loc): Convert 2nd param from source_location to
17596         const dump_location_t &.
17597         (dump_generic_expr_loc): Delete.
17598         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
17599         const dump_location_t &.
17600         * gimple-fold.c (fold_gimple_assign): Update call to
17601         dump_printf_loc to pass in a dump_location_t rather than a
17602         location_t, via the gimple stmt.
17603         (gimple_fold_call): Likewise.
17604         * gimple-loop-interchange.cc
17605         (loop_cand::analyze_iloop_reduction_var): Update for change to
17606         check_reduction_path.
17607         (tree_loop_interchange::interchange): Update for change to
17608         find_loop_location.
17609         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
17610         change in return-type of find_loop_location.
17611         (graphite_regenerate_ast_isl): Likewise.
17612         * graphite-optimize-isl.c (optimize_isl): Likewise.
17613         * graphite.c (graphite_transform_loops): Likewise.
17614         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
17615         pass in a dump_location_t rather than a location_t, via the
17616         gimple stmt.
17617         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
17618         * ipa.c (walk_polymorphic_call_targets): Likewise.
17619         * loop-unroll.c (report_unroll): Convert "locus" param from
17620         location_t to dump_location_t.
17621         (decide_unrolling): Update for change to get_loop_location's
17622         return type.
17623         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
17624         location_t to dump_user_location_t.
17625         (grid_find_single_omp_among_assignments_1): Updates calls to
17626         dump_printf_loc to pass in a dump_location_t rather than a
17627         location_t, via the gimple stmt.
17628         (grid_parallel_clauses_gridifiable): Convert "tloc" from
17629         location_t to dump_location_t.  Updates calls to dump_printf_loc
17630         to pass in a dump_location_t rather than a location_t, via the
17631         gimple stmt.
17632         (grid_inner_loop_gridifiable_p): Likewise.
17633         (grid_dist_follows_simple_pattern): Likewise.
17634         (grid_gfor_follows_tiling_pattern): Likewise.
17635         (grid_target_follows_gridifiable_pattern): Likewise.
17636         (grid_attempt_target_gridification): Convert initialization
17637         of local "grid" from memset to zero-initialization; FIXME: does
17638         this require C++11?  Update call to dump_printf_loc to pass in a
17639         optinfo_location rather than a location_t, via the gimple stmt.
17640         * profile.c (read_profile_edge_counts): Updates call to
17641         dump_printf_loc to pass in a dump_location_t rather than a
17642         location_t
17643         (compute_branch_probabilities): Likewise.
17644         * selftest-run-tests.c (selftest::run_tests): Call
17645         dumpfile_c_tests.
17646         * selftest.h (dumpfile_c_tests): New decl.
17647         * tree-loop-distribution.c (pass_loop_distribution::execute):
17648         Update for change in return type of find_loop_location.
17649         * tree-parloops.c (parallelize_loops): Likewise.
17650         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
17651         "locus" from location_t to dump_user_location_t.
17652         (canonicalize_loop_induction_variables): Likewise.
17653         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
17654         for change in return type of find_loop_location.
17655         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
17656         to dump_printf_loc to pass in a dump_location_t rather than a
17657         location_t, via the stmt.
17658         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
17659         Likewise.
17660         * tree-vect-loop-manip.c (find_loop_location): Convert return
17661         type from source_location to dump_user_location_t.
17662         (vect_do_peeling): Update for above change.
17663         (vect_loop_versioning): Update for change in type of
17664         vect_location.
17665         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
17666         from location_t to dump_user_location_t.
17667         (vect_estimate_min_profitable_iters): Update for change in type
17668         of vect_location.
17669         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
17670         location_t to dump_location_t.
17671         (vect_slp_bb): Update for change in type of vect_location.
17672         * tree-vectorizer.c (vect_location): Convert from source_location
17673         to dump_user_location_t.
17674         (try_vectorize_loop_1): Update for change in vect_location's type.
17675         (vectorize_loops): Likewise.
17676         (increase_alignment): Likewise.
17677         * tree-vectorizer.h (vect_location): Convert from source_location
17678         to dump_user_location_t.
17679         (find_loop_location): Convert return type from source_location to
17680         dump_user_location_t.
17681         (check_reduction_path): Convert 1st param from location_t to
17682         dump_user_location_t.
17683         * value-prof.c (check_counter): Update call to dump_printf_loc to
17684         pass in a dump_user_location_t rather than a location_t; update
17685         call to error_at for change in type of "locus".
17686         (check_ic_target): Update call to dump_printf_loc to
17687         pass in a dump_user_location_t rather than a location_t, via the
17688         call_stmt.
17690 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
17692         * config/s390/s390.h (enum processor_flags): Do not use
17693         default tune parameter when -march was specified.
17695 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
17697         PR target/86314
17698         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
17699         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
17700         operands.
17702 2018-06-26  Richard Biener  <rguenther@suse.de>
17704         PR tree-optimization/86287
17705         PR bootstrap/86316
17706         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
17707         (vect_analyze_loop): Initialize n_stmts.
17709 2018-06-26  Richard Biener  <rguenther@suse.de>
17711         PR middle-end/86271
17712         * fold-const.c (fold_convertible_p): Pointer extension
17713         isn't valid.
17715 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
17717         PR debug/86064
17718         * dwarf2out.c (loc_list_has_views): Adjust comments.
17719         (dw_loc_list): Split single cross-partition range with
17720         nonzero locview.
17722 2018-06-25  Jeff Law  <law@redhat.com>
17724         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
17725         on -mbig-switch by default.
17727         * config/v850/predicates.md (const_float_1_operand): Fix match_code
17728         test.
17729         (const_float_0_operand): Remove unused predicate.
17730         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
17731         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
17732         (recipsf2): New expander.  Original pattern now called
17733         (recipsf2_insn).
17734         (recipdf2, recipdf2_insn): Similarly.
17735         (rsqrtsf2, rsqrtsf2_insn): Similarly
17736         (rsqrtdf2, rsqrtdf2_insn): Similarly
17738 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
17740         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
17741         Simplify logic for FreeBSD (twice).
17743 2018-06-25  Martin Sebor  <msebor@redhat.com>
17745         PR tree-optimization/86204
17746         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
17747         a strnlen result if it's less than the length of the string.
17749 2018-06-25  Martin Sebor  <msebor@redhat.com>
17751         PR tree-optimization/85700
17752         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
17753         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
17754         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
17756 2018-06-25  Martin Sebor  <msebor@redhat.com>
17758         * doc/extend.texi (Zero-length arrays): Update and clarify.
17760 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
17762         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
17763         added IEEE/IBM long double multilib support on PowerPC little
17764         endian Linux systems.
17765         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
17766         (MULTILIB_DEFAULTS): Likewise.
17767         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17768         Likewise.
17769         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
17770         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
17771         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
17773 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
17775         PR middle-end/86311
17776         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
17777         (REORDER_45): Likewise.
17779 2018-06-25  Jeff Law  <law@redhat.com>
17781         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
17782         dividend to 32 bits.  Adjust length.
17783         (udivmodhi4): Cleanup output template.  Fix length.
17785 2018-06-25  Carl Love  <cel@us.ibm.com>
17787         * config/rs6000/vsx.md: Change word selector to prefered location.
17789 2018-06-25  Richard Biener  <rguenther@suse.de>
17791         PR tree-optimization/86304
17792         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
17793         epilogue-if-converted loops as well.
17795 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
17797         * lto-section-out.c (lto_begin_section): Do not print section
17798         name for noaddr and unnumbered dumps.
17800 2018-06-25  Richard Biener  <rguenther@suse.de>
17802         * tree-vectorizer.h (struct vec_info_shared): New structure
17803         with parts split out from struct vec_info and loop_nest from
17804         struct _loop_vec_info.
17805         (struct vec_info): Adjust accordingly.
17806         (struct _loop_vec_info): Likewise.
17807         (LOOP_VINFO_LOOP_NEST): Adjust.
17808         (LOOP_VINFO_DATAREFS): Likewise.
17809         (LOOP_VINFO_DDRS): Likewise.
17810         (struct _bb_vec_info): Likewise.
17811         (BB_VINFO_DATAREFS): Likewise.
17812         (BB_VINFO_DDRS): Likewise.
17813         (struct _stmt_vec_info): Add dr_aux member.
17814         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
17815         (DR_MISALIGNMENT_UNINITIALIZED): New.
17816         (set_dr_misalignment): Adjust.
17817         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
17818         (vect_analyze_loop): Adjust prototype.
17819         (vect_analyze_loop_form): Likewise.
17820         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17821         Compute dependences lazily.
17822         (vect_record_base_alignments): Use shared datarefs/ddrs.
17823         (vect_verify_datarefs_alignment): Likewise.
17824         (vect_analyze_data_refs_alignment): Likewise.
17825         (vect_analyze_data_ref_accesses): Likewise.
17826         (vect_analyze_data_refs): Likewise.
17827         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
17828         constructor parameter for shared part.
17829         (vect_analyze_loop_form): Pass in shared part and adjust.
17830         (vect_analyze_loop_2): Pass in storage for the number of
17831         stmts.  Move loop nest finding to the caller.  Compute
17832         datarefs lazily.
17833         (vect_analyze_loop): Pass in shared part.
17834         (vect_transform_loop): Verify shared datarefs are unchanged.
17835         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
17836         constructor parameter for shared part.
17837         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
17838         (vect_slp_bb): Verify shared datarefs are unchanged before
17839         transform.
17840         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
17841         change.
17842         (new_stmt_vec_info): Initialize DR_AUX misalignment to
17843         DR_MISALIGNMENT_UNINITIALIZED.
17844         * tree-vectorizer.c (vec_info::vec_info): Add constructor
17845         parameter for shared part.
17846         (vec_info::~vec_info): Adjust.
17847         (vec_info_shared::vec_info_shared): New.
17848         (vec_info_shared::~vec_info_shared): Likewise.
17849         (vec_info_shared::save_datarefs): Likewise.
17850         (vec_info_shared::check_datarefs): Likewise.
17851         (try_vectorize_loop_1): Construct shared part live for analyses
17852         of a single loop for multiple vector sizes.
17853         * tree-parloops.c (gather_scalar_reductions): Adjust.
17855 2018-06-25  Richard Biener  <rguenther@suse.de>
17857         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
17858         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
17859         (vect_analyze_data_refs): Remove similar code from here and
17860         simplify accordingly.
17862 2018-06-25  Richard Biener  <rguenther@suse.de>
17864         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
17865         for reverse storage order accesses rather than asserting
17866         they cannot happen here.
17868 2018-06-25  Tom de Vries  <tdevries@suse.de>
17870         PR debug/86257
17871         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
17872         Use data16 instead of .byte for insn prefix.
17874 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
17876         PR C++/86082
17877         * parser.c (make_char_string_pack): Pass this literal chars
17878         through cpp_interpret_string.
17879         (cp_parser_userdef_numeric_literal): Check the result of
17880         make_char_string_pack.
17882 2018-06-24  Maya Rashish  <coypu@sdf.org>
17884         * ginclude/stddef.h: Simplify conditions around avoiding
17885         re-definition of __size_t.
17887 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17889         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
17890         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
17892 2018-06-22  Maya Rashish  <coypu@sdf.org>
17894         * doc/invoke.texi (mno-fancy-math-387): Update for changes
17895         made to OpenBSD and NetBSD through the years.
17897 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17899         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
17900         behavior of vec_pack (vector double, vector double) to match
17901         behavior of vec_float2 (vector double, vector double).
17903 2018-06-22  Olivier Hainque  <hainque@adacore.com>
17905         * gimplify.c (gimplify_function_tree): Prevent creation
17906         of a trampoline for the address of the current function
17907         passed to entry/exit instrumentation hooks.
17909 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
17911         PR target/86222
17912         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
17913         correctly.
17915 2018-06-22  Martin Liska  <mliska@suse.cz>
17917         PR tree-optimization/86263
17918         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17919         Bail out if is_enabled is false.
17920         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
17921         New declaration.
17922         (jump_table_cluster::is_enabled): New function.
17924 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17926         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
17927         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
17928         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
17929         (lto_input_ts_binfo_tree_pointers): Likewise.
17930         * tree-streamer-out.c (streamer_write_tree_bitfields,
17931         write_ts_binfo_tree_pointers): Likewise.
17932         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
17934 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17936         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
17938 2018-06-22  Martin Liska  <mliska@suse.cz>
17940         * symbol-summary.h (get): Make it pure and inline move
17941         functionality from ::get function.
17942         (get): Remove and inline into ::get and ::get_create.
17943         (get_create): Move code from ::get function.
17945 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17947         PR target/85994
17948         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
17949         -x assembler-with-cpp.
17951 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17953         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
17954         _FILE_OFFSET_BITS=64 for C++.
17956 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
17958         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
17959         conversion insn that shows up when pr85657-3.c is compiled using
17960         IEEE 128-bit long double.
17961         (neg<mode>2_internal): Use the correct mode to check whether the
17962         mode is IBM extended.
17963         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
17964         multiply and divide external functions from being created more
17965         than once.
17967 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17969         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
17970         functions.
17971         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
17972         the edge can be forwarded.
17973         (cfg_layout_merge_blocks): Likewise.
17975 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17977         * except.c (finish_eh_generation): Commit edge insertions only after
17978         the EH edges have been redirected from post-landing to landing pads.
17980 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17982         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
17983         create_tmp_var_for to create the FRAME decl.
17984         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
17986 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17988         * tree-inline.c (copy_edges_for_bb): Minor tweak.
17989         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
17990         debug statement when resetting its value.
17991         (expand_call_inline): Copy the locus of the call onto the assignment
17992         of the return value, if any.  Use local variable in more cases.
17994 2018-06-21  Martin Liska  <mliska@suse.cz>
17996         * ipa-pure-const.c (propagate_nothrow): Use
17997         funct_state_summaries->get.
17998         (dump_malloc_lattice): Likewise.
17999         (propagate_malloc): Likewise.
18001 2018-06-21  Richard Biener  <rguenther@suse.de>
18003         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
18004         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
18005         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
18006         comment.
18007         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
18008         BLOCK_ABSTRACT_ORIGIN unconditionally.
18010 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
18012         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
18013         deleting it.
18014         * ipa-reference.c (ipa_reference_c_finalize): Delete
18015         ipa_ref_opt_sum_summaries and set it to NULL.
18017 2018-06-21  Tom de Vries  <tdevries@suse.de>
18019         PR tree-optimization/85859
18020         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
18021         test with comment from bb_no_side_effects_p.
18023 2018-06-21  Richard Biener  <rguenther@suse.de>
18025         PR tree-optimization/86232
18026         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
18027         max for constant niter.
18029 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18031         * config/aarch64/aarch64-simd.md
18032         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
18034 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18036         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
18037         Make opernads of the unspec commutative.
18039 2018-06-21  Richard Biener  <rguenther@suse.de>
18041         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
18042         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
18043         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
18044         (vect_analyze_data_ref_dependence): Re-order checks to deal with
18045         NULL DR_STEP.
18046         (vect_record_base_alignments): Do not record base alignment
18047         for gathers or scatters.
18048         (vect_compute_data_ref_alignment): Drop return value that is always
18049         true.  Bail out early for gathers or scatters.
18050         (vect_enhance_data_refs_alignment): Bail out early for gathers
18051         or scatters.
18052         (vect_find_same_alignment_drs): Likewise.
18053         (vect_analyze_data_refs_alignment): Remove dead code.
18054         (vect_slp_analyze_and_verify_node_alignment): Likewise.
18055         (vect_analyze_data_refs): For possible gathers or scatters do
18056         not create an alternate DR, just check their possible validity
18057         and mark them.  Adjust DECL_NONALIASED handling to not rely
18058         on DR_BASE_ADDRESS.
18059         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
18060         update inits of gathers or scatters.
18061         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
18062         Also copy gather/scatter flag to pattern vinfo.
18064 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18066         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18067         behavior of vec_packsu (vector unsigned long long, vector unsigned
18068         long long) to match behavior of vec_packs with same signature.
18070 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
18071             Thomas Schwinge <thomas@codesourcery.com>
18072             Cesar Philippidis  <cesar@codesourcery.com>
18074         * gimplify.c (gimplify_scan_omp_clauses): Add support for
18075         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18076         (gimplify_adjust_omp_clauses): Likewise.
18077         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
18078         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
18079         (gimplify_omp_target_update): Update handling of acc update and
18080         enter/exit data.
18081         * omp-low.c (install_var_field): Remove unused parameter
18082         base_pointers_restrict.
18083         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
18084         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
18085         FINALIZE}
18086         (omp_target_base_pointers_restrict_p): Delete.
18087         (scan_omp_target): Update call to scan_sharing_clauses.
18088         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
18089         FINALIZE}.
18090         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
18091         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18092         (convert_local_omp_clauses): Likewise.
18093         * tree-pretty-print.c (dump_omp_clause): Likewise.
18094         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
18095         FINALIZE}.
18096         (omp_clause_code_name): Likewise.
18098 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
18100         PR debug/86194
18101         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
18102         be narrowed.
18104         PR tree-optimization/86231
18105         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
18106         anti-range don't overwrite *vr0min before using it to compute *vr0max.
18108 2018-06-20  Tom de Vries  <tdevries@suse.de>
18110         PR tree-optimization/86097
18111         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
18112         iv type if signedness of iv type is not the same as that of *nit.
18114 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
18116         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
18117         EDGE_EH edges, verify they are all EDGE_EH.
18119 2018-06-20  Maya Rashish  <coypu@sdf.org>
18121         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
18123 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18125         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
18126         * config/aarch64/aarch64.c (xgene1_tunings): Add
18127         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
18128         (aarch64_mode_valid_for_sched_fusion_p):
18129         Allow 16-byte modes.
18130         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
18131         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
18132         128-bit modes.
18133         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
18134         New pattern.
18135         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
18136         * config/aarch64/iterators.md (VQ2): New mode iterator.
18138 2018-06-20  Martin Liska  <mliska@suse.cz>
18140         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
18141         Change default ratio from 10 to 8.
18143 2018-06-20  Martin Liska  <mliska@suse.cz>
18145         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18146         New.
18147         (bit_test_cluster::find_bit_tests): Likewise.
18148         (switch_decision_tree::analyze_switch_statement): Find clusters.
18149         * tree-switch-conversion.h (struct jump_table_cluster): Document
18150         hierarchy.
18152 2018-06-20  Martin Liska  <mliska@suse.cz>
18154         * tree-switch-conversion.c (switch_conversion::collect):
18155         Record m_uniq property.
18156         (switch_conversion::expand): Bail out for special conditions.
18157         (group_cluster::~group_cluster): New.
18158         (group_cluster::group_cluster): Likewise.
18159         (group_cluster::dump): Likewise.
18160         (jump_table_cluster::emit): New.
18161         (switch_decision_tree::fix_phi_operands_for_edges): New.
18162         (struct case_node): Remove struct.
18163         (jump_table_cluster::can_be_handled): New.
18164         (case_values_threshold): Moved to header.
18165         (reset_out_edges_aux): Likewise.
18166         (jump_table_cluster::is_beneficial): New.
18167         (bit_test_cluster::can_be_handled): Likewise.
18168         (add_case_node): Remove.
18169         (bit_test_cluster::is_beneficial): New.
18170         (case_bit_test::cmp): New.
18171         (bit_test_cluster::emit): New.
18172         (expand_switch_as_decision_tree_p): Remove.
18173         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
18174         (fix_phi_operands_for_edge): Likewise.
18175         (switch_decision_tree::analyze_switch_statement): New.
18176         (compute_cases_per_edge): Move ...
18177         (switch_decision_tree::compute_cases_per_edge): ... here.
18178         (try_switch_expansion): Likewise.
18179         (switch_decision_tree::try_switch_expansion): Likewise.
18180         (record_phi_operand_mapping): Likewise.
18181         (switch_decision_tree::record_phi_operand_mapping): Likewise.
18182         (emit_case_decision_tree): Likewise.
18183         (switch_decision_tree::emit): Likewise.
18184         (balance_case_nodes): Likewise.
18185         (switch_decision_tree::balance_case_nodes): Likewise.
18186         (dump_case_nodes): Likewise.
18187         (switch_decision_tree::dump_case_nodes): Likewise.
18188         (emit_jump): Likewise.
18189         (switch_decision_tree::emit_jump): Likewise.
18190         (emit_cmp_and_jump_insns): Likewise.
18191         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
18192         (emit_case_nodes): Likewise.
18193         (switch_decision_tree::emit_case_nodes): Likewise.
18194         (conditional_probability): Remove.
18195         * tree-switch-conversion.h (enum cluster_type): New.
18196         (PRINT_CASE): New.
18197         (struct cluster): Likewise.
18198         (cluster::cluster): Likewise.
18199         (struct simple_cluster): Likewise.
18200         (simple_cluster::simple_cluster): Likewise.
18201         (struct group_cluster): Likewise.
18202         (struct jump_table_cluster): Likewise.
18203         (struct bit_test_cluster): Likewise.
18204         (struct min_cluster_item): Likewise.
18205         (struct case_tree_node): Likewise.
18206         (case_tree_node::case_tree_node): Likewise.
18207         (jump_table_cluster::case_values_threshold): Likewise.
18208         (struct case_bit_test): Likewise.
18209         (struct switch_decision_tree): Likewise.
18210         (struct switch_conversion): Likewise.
18211         (switch_decision_tree::reset_out_edges_aux): Likewise.
18213 2018-06-20  Martin Liska  <mliska@suse.cz>
18215         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
18216         (hoist_edge_and_branch_if_true): Likewise.
18217         (expand_switch_using_bit_tests_p): Likewise.
18218         (struct case_bit_test): Likewise.
18219         (case_bit_test_cmp): Likewise.
18220         (emit_case_bit_tests): Likewise.
18221         (switch_conversion::switch_conversion): New class.
18222         (struct switch_conv_info): Remove old struct.
18223         (collect_switch_conv_info): More to ...
18224         (switch_conversion::collect): ... this.
18225         (check_range): Likewise.
18226         (switch_conversion::check_range): Likewise.
18227         (check_all_empty_except_final): Likewise.
18228         (switch_conversion::check_all_empty_except_final): Likewise.
18229         (check_final_bb): Likewise.
18230         (switch_conversion::check_final_bb): Likewise.
18231         (create_temp_arrays): Likewise.
18232         (switch_conversion::create_temp_arrays): Likewise.
18233         (free_temp_arrays): Likewise.
18234         (gather_default_values): Likewise.
18235         (switch_conversion::gather_default_values): Likewise.
18236         (build_constructors): Likewise.
18237         (switch_conversion::build_constructors): Likewise.
18238         (constructor_contains_same_values_p): Likewise.
18239         (switch_conversion::contains_same_values_p): Likewise.
18240         (array_value_type): Likewise.
18241         (switch_conversion::array_value_type): Likewise.
18242         (build_one_array): Likewise.
18243         (switch_conversion::build_one_array): Likewise.
18244         (build_arrays): Likewise.
18245         (switch_conversion::build_arrays): Likewise.
18246         (gen_def_assigns): Likewise.
18247         (switch_conversion::gen_def_assigns): Likewise.
18248         (prune_bbs): Likewise.
18249         (switch_conversion::prune_bbs): Likewise.
18250         (fix_phi_nodes): Likewise.
18251         (switch_conversion::fix_phi_nodes): Likewise.
18252         (gen_inbound_check): Likewise.
18253         (switch_conversion::gen_inbound_check): Likewise.
18254         (process_switch): Use the newly created class.
18255         (switch_conversion::expand): New.
18256         (switch_conversion::~switch_conversion): New.
18257         * tree-switch-conversion.h: New file.
18259 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18261         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
18262         tree-vect-patterns.c.
18263         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
18264         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
18265         (vect_recog_sad_pattern): Likewise.
18266         (vect_recog_widen_sum_pattern): Likewise.
18267         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
18268         (vect_recog_widen_shift_pattern): Remove the type_in argument.
18269         (vect_recog_rotate_pattern): Likewise.
18270         (vect_recog_mult_pattern): Likewise.
18271         (vect_recog_vector_vector_shift_pattern): Likewise.
18272         (vect_recog_divmod_pattern): Likewise.
18273         (vect_recog_mixed_size_cond_pattern): Likewise.
18274         (vect_recog_bool_pattern): Likewise.
18275         (vect_recog_mask_conversion_pattern): Likewise.
18276         (vect_try_gather_scatter_pattern): Likewise.
18277         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
18278         (vect_recog_over_widening_pattern): Likewise.
18279         (vect_recog_gather_scatter_pattern): Likewise.
18280         (vect_recog_func_ptr): Move from tree-vectorizer.h
18281         (vect_vect_recog_func_ptrs): Move further down the file.
18282         (vect_recog_func): Likewise.  Remove the third argument.
18283         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
18284         (vect_pattern_recog_1): Expect the pattern function to do any
18285         necessary target tests.  Also expect it to provide a vector type.
18286         Remove the type_in handling.
18288 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18290         * tree-vect-patterns.c (vect_pattern_detected): New function.
18291         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
18292         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
18293         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
18294         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
18295         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
18296         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
18297         (vect_recog_mask_conversion_pattern)
18298         (vect_try_gather_scatter_pattern): Likewise.
18300 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18302         * tree-vect-patterns.c (vect_get_internal_def): New function.
18303         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
18304         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
18305         (search_type_for_mask_1): Use it.
18307 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18309         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18310         redundant WIDEN_SUM_EXPR handling.
18311         (vect_recog_sad_pattern): Likewise.
18313 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18315         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18316         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
18317         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
18318         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
18319         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
18321 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18323         * tree-vect-stmts.c (vectorizable_call): Make sure that we
18324         use the stmt_vec_info of the original bb statement for the
18325         new zero assignment, even if the call is part of a pattern.
18327 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18329         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
18330         that the sequence is attached to the original statement rather
18331         than the pattern statement.
18332         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
18333         PATTERN_DEF_SEQ from the original statement rather than
18334         the main pattern statement.
18335         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
18336         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
18337         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
18339 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
18341         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
18342         definition statements before the early exit for statements that aren't
18343         live or relevant.
18344         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
18345         split out from...
18346         (vect_transform_loop): ...here.  Process pattern definition
18347         statements without first checking whether the main pattern
18348         statement is live or relevant.
18350 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
18352         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
18353         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
18355 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
18357         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
18358         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
18359         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
18360         (expand_block_compare): Change select_block_compare_mode call.
18361         (expand_strncmp_align_check): Use new functions, fix comment.
18362         (emit_final_str_compare_gpr): New function.
18363         (expand_strn_compare): Refactor and clean up code.
18364         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
18366 2018-06-19  Tony Reix  <tony.reix@atos.com>
18367             Damien Bergamini  <damien.bergamini@atos.com>
18368             David Edelsohn  <dje.gcc@gmail.com>
18370         * collect2.c (static_obj): New variable.
18371         (static_libs): New variable.
18372         (is_in_list): Uncomment declaration.
18373         (main): Track AIX libraries linked statically.
18374         (is_in_list): Uncomment definition.
18375         (scan_prog_file): Don't add AIX shared libraries initializer
18376         to constructor list if linking statically.
18378 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
18380         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
18381         constant.
18382         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
18384 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
18386         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
18387         blocks.
18389 2018-06-19  Martin Liska  <mliska@suse.cz>
18391         * config/i386/i386.c (ix86_can_inline_p): Do not use
18392         ipa_fn_summaries::get_create.
18393         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
18394         get.
18395         (devirtualization_time_bonus): Likewise.
18396         (ipcp_propagate_stage): Likewise.
18397         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18398         (edge_set_predicate): Likewise.
18399         (evaluate_conditions_for_known_args): Likewise.
18400         (evaluate_properties_for_edge): Likewise.
18401         (ipa_call_summary::reset): Tranform to ...
18402         (ipa_call_summary::~ipa_call_summary): ... this.
18403         (ipa_fn_summary::reset): Transform to ...
18404         (ipa_fn_summary::~ipa_fn_summary): ... this.
18405         (ipa_fn_summary_t::remove): Rename to ...
18406         (ipa_fn_summary_t::remove_callees): ... this.
18407         (ipa_fn_summary_t::duplicate): Use placement new
18408         instead of memory copy.
18409         (ipa_call_summary_t::duplicate): Likewise.
18410         (ipa_call_summary_t::remove): Remove.
18411         (dump_ipa_call_summary): Change get_create to get.
18412         (ipa_dump_fn_summary): Dump only when summary exists.
18413         (analyze_function_body): Use symbol_summary::get instead
18414         of get_create.
18415         (compute_fn_summary): Likewise.
18416         (estimate_edge_devirt_benefit): Likewise.
18417         (estimate_edge_size_and_time): Likewise.
18418         (inline_update_callee_summaries): Likewise.
18419         (remap_edge_change_prob): Likewise.
18420         (remap_edge_summaries): Likewise.
18421         (ipa_merge_fn_summary_after_inlining): Likewise.
18422         (write_ipa_call_summary): Likewise.
18423         (ipa_fn_summary_write): Likewise.
18424         (ipa_free_fn_summary): Likewise.
18425         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
18426         (struct ipa_call_summary): Likewise.
18427         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
18428         of get_create.
18429         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
18430         (estimate_size_after_inlining): Likewise.
18431         (estimate_growth): Likewise.
18432         (growth_likely_positive): Likewise.
18433         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
18434         (inline_call): Likewise.
18435         * ipa-inline.c (caller_growth_limits): Likewise.
18436         (can_inline_edge_p): Likewise.
18437         (can_inline_edge_by_limits_p): Likewise.
18438         (compute_uninlined_call_time): Likewise.
18439         (compute_inlined_call_time): Likewise.
18440         (want_inline_small_function_p): Likewise.
18441         (edge_badness): Likewise.
18442         (update_caller_keys): Likewise.
18443         (update_callee_keys): Likewise.
18444         (inline_small_functions): Likewise.
18445         (inline_to_all_callers_1): Likewise.
18446         (dump_overall_stats): Likewise.
18447         (early_inline_small_functions): Likewise.
18448         (early_inliner): Likewise.
18449         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
18450         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18451         * ipa-pure-const.c (malloc_candidate_p): Likewise.
18452         * ipa-split.c (execute_split_functions): Likewise.
18453         * symbol-summary.h: Likewise.
18454         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
18456 2018-06-19  Richard Biener  <rguenther@suse.de>
18458         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
18459         (vectorize_loops): ... here.  Fix dbgcnt handling.
18460         (try_vectorize_loop): Wrap try_vectorize_loop_1.
18462 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
18464         PR target/86197
18465         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
18466         ieee128 argument takes up only one (vector) register, not two (floating
18467         point) registers.
18469 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
18471         * gimplify.c (gimplify_init_constructor): Really never clear for an
18472         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
18474 2018-06-19  Richard Biener  <rguenther@suse.de>
18476         PR tree-optimization/86179
18477         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
18478         after failed recognition.
18480 2018-06-18  Martin Sebor  <msebor@redhat.com>
18482         PR middle-end/85602
18483         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
18484         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
18485         Handle integer subtraction.
18486         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
18487         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
18489 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
18491         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
18492         param from rtx to rtx_insn *.
18493         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
18494         param.
18495         (frv_ifcvt_modify_insn): Likwise.
18496         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
18497         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
18498         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
18499         as_a <rtx_insn *> cast to local "unprotected_region" once
18500         it's been established that it's not NULL or pc_rtx.
18501         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
18502         param "sethi" from rtx to rtx_insn *.
18503         (nds32_group_float_insns): Likewise for param "insn".
18504         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
18505         param.
18506         (vax_output_int_subtract): Likewise.
18507         * config/vax/vax.c (vax_output_int_add): Likewise for param
18508         "insn".
18509         (vax_output_int_subtract): Likewise.
18510         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
18511         (emit_pattern_after): Likewise for param "after".
18512         (emit_insn_after): Likewise.
18513         (emit_jump_insn_after): Likewise.
18514         (emit_call_insn_after): Likewise.
18515         (emit_debug_insn_after): Likewise.
18516         (emit_pattern_before): Likewise for param "before".
18517         (emit_insn_before): Likewise.
18518         (emit_jump_insn_before): Likewise.
18519         * final.c (get_insn_template): Likewise for param "insn", removing
18520         a cast.
18521         * output.h (get_insn_template): Likewise for 2nd param.
18522         * rtl.h (emit_insn_before): Likewise.
18523         (emit_jump_insn_before): Likewise.
18524         (emit_debug_insn_before_noloc): Likewise.
18525         (emit_insn_after): Likewise.
18526         (emit_jump_insn_after): Likewise.
18527         (emit_call_insn_after): Likewise.
18528         (emit_debug_insn_after): Likewise.
18529         (set_insn_deleted): Likewise for param.
18531 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
18533         PR target/85358
18534         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
18535         floating point modes, so that IFmode is numerically greater than
18536         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
18537         to declare the ordering.  This prevents IFmode from being
18538         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
18539         machine.  Include rs6000-modes.h to share the fractional values
18540         between genmodes* and the rest of the compiler.
18541         (IFmode): Likewise.
18542         (KFmode): Likewise.
18543         (TFmode): Likewise.
18544         * config/rs6000/rs6000-modes.h: New file.
18545         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
18546         meaning of rs6000_long_double_size so that 126..128 selects an
18547         appropriate 128-bit floating point type.
18548         (rs6000_option_override_internal): Likewise.
18549         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
18550         (TARGET_LONG_DOUBLE_128): Change the meaning of
18551         rs6000_long_double_size so that 126..128 selects an appropriate
18552         128-bit floating point type.
18553         (LONG_DOUBLE_TYPE_SIZE): Update comment.
18554         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
18555         source and destination to match the standard usage.
18556         (truncifkf2): Likewise.
18557         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
18558         ISA 2.07 to use an explicit clobber, instead of passing in a
18559         temporary.
18560         (copysign<mode>3_soft): Likewise.
18562 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
18564         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
18565         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
18566         (vect_slp_analyze_instance_dependence): Likewise.
18567         (vect_enhance_data_refs_alignment): Likewise.
18568         (vect_analyze_data_refs_alignment): Likewise.
18569         (vect_slp_analyze_and_verify_instance_alignment
18570         (vect_analyze_data_ref_accesses): Likewise.
18571         (vect_prune_runtime_alias_test_list): Likewise.
18572         (vect_analyze_data_refs): Likewise.
18573         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
18574         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18575         (vect_analyze_scalar_cycles_1): Likewise.
18576         (vect_get_loop_niters): Likewise.
18577         (vect_analyze_loop_form_1): Likewise.
18578         (vect_update_vf_for_slp): Likewise.
18579         (vect_analyze_loop_operations): Likewise.
18580         (vect_analyze_loop): Likewise.
18581         (vectorizable_induction): Likewise.
18582         (vect_transform_loop): Likewise.
18583         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
18584         * tree-vect-slp.c (vect_analyze_slp): Likewise.
18585         (vect_make_slp_decision): Likewise.
18586         (vect_detect_hybrid_slp): Likewise.
18587         (vect_slp_analyze_operations): Likewise.
18588         (vect_slp_bb): Likewise.
18589         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
18590         (vectorizable_bswap): Likewise.
18591         (vectorizable_call): Likewise.
18592         (vectorizable_simd_clone_call): Likewise.
18593         (vectorizable_conversion): Likewise.
18594         (vectorizable_assignment): Likewise.
18595         (vectorizable_shift): Likewise.
18596         (vectorizable_operation): Likewise.
18597         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
18599 2018-06-18  Martin Sebor  <msebor@redhat.com>
18601         PR tree-optimization/81384
18602         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
18603         * builtins.c (expand_builtin_strnlen): New function.
18604         (expand_builtin): Call it.
18605         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
18606         * builtins.def (BUILT_IN_STRNLEN): New.
18607         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
18608         Warn for bounds in excess of maximum object size.
18609         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
18610         single-value ranges.  Handle strnlen.
18611         (handle_builtin_strlen): Handle strnlen.
18612         (strlen_check_and_optimize_stmt): Same.
18613         * doc/extend.texi (Other Builtins): Document strnlen.
18615 2018-06-18  Maya Rashish  <coypu@sdf.org>
18617         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
18618         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
18619         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
18621         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
18622         here to ...
18623         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
18625 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18627         * tree.c (escaped_string::escape): Replace cast to char * by
18628         const_cast<char *> (unescaped).
18630 2018-06-18  Nick Clifton  <nickc@redhat.com>
18632         PR 84195
18633         * tree.c (escaped_string): New class.  Converts an unescaped
18634         string into its escaped equivalent.
18635         (warn_deprecated_use): Use the new class to convert the
18636         deprecation message, if present.
18637         (test_escaped_strings): New self test.
18638         (test_c_tests): Add test_escaped_strings.
18639         * doc/extend.texi (deprecated): Add a note that the
18640         deprecation message is affected by the -fmessage-length
18641         option, and that control characters will be escaped.
18642         (#pragma GCC error): Document this pragma.
18643         (#pragma GCC warning): Likewise.
18644         * doc/invoke.texi (-fmessage-length): Document this option's
18645         effect on the #warning and #error preprocessor directives and
18646         the deprecated attribute.
18648 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
18650         * tree.c (decl_value_expr_lookup): Revert latest change.
18651         (decl_value_expr_insert): Likewise.
18653 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
18655         * gimplify.c (nonlocal_vlas): Delete.
18656         (nonlocal_vla_vars): Likewise.
18657         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
18658         referenced VLAs.
18659         (gimplify_body): Do not create and destroy nonlocal_vlas.
18660         * tree-nested.c: Include diagnostic.h.
18661         (use_pointer_in_frame): Tweak.
18662         (lookup_field_for_decl): Add assertion and declare the transformation.
18663         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
18664         internal error when the reference is in a wrong context.  Do not
18665         create a debug decl by default.
18666         (note_nonlocal_block_vlas): Delete.
18667         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
18668         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
18669         create a debug decl by default.
18670         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
18671         call is in a wrong context.
18672         (fixup_vla_decls): New function.
18673         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
18674         debug variables were created.
18675         * tree.c (decl_value_expr_lookup): Add checking assertion.
18676         (decl_value_expr_insert): Likewise.
18678 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18680         PR middle-end/82479
18681         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
18682         * tree-scalar-evolution.c (interpret_expr): Likewise.
18683         (expression_expensive_p): Likewise.
18684         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
18685         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
18686         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
18687         (ssa_defined_by_minus_one_stmt_p): New.
18689 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18691         PR middle-end/64946
18692         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
18693         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
18694         * dojump.c (do_jump): Likewise.
18695         * expr.c (expand_expr_real_2): Check operand type's sign.
18696         * fold-const.c (const_unop): Handle ABSU_EXPR.
18697         (fold_abs_const): Likewise.
18698         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
18699         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
18700         (strip_sign_op_1): Likesise.
18701         * match.pd: Add new pattern to generate ABSU_EXPR.
18702         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
18703         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18704         * tree-eh.c (operation_could_trap_helper_p): Likewise.
18705         * tree-inline.c (estimate_operator_cost): Likewise.
18706         * tree-pretty-print.c (dump_generic_node): Likewise.
18707         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
18708         * tree.def (ABSU_EXPR): New.
18710 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
18712         PR middle-end/86095
18713         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
18714         documented as preserved for backward compatibility only.
18715         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
18717         PR rtl-optimization/86108
18718         * bb-reorder.c (create_forwarder_block): Renamed to ...
18719         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
18720         jump from new landing pad to the second part.
18721         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
18722         Adjust callers.
18724 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
18726         PR middle-end/85878
18727         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
18728         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
18729         Only call store_expr for halves if the mode is the same.
18731         PR middle-end/86123
18732         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
18733         Fix up comment formatting.
18735 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18737         * typed-splay-tree.h (typed_splay_tree::remove): New function.
18738         (typed_splay_tree::closure,
18739         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
18740         (typed_splay_tree::typed_splay_tree,
18741         typed_splay_tree::operator =): Declared private.
18742         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
18743         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
18744         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
18745         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
18746         typed_splay_tree::splay_tree_splay,
18747         typed_splay_tree::splay_tree_foreach_helper,
18748         typed_splay_tree::splay_tree_insert,
18749         typed_splay_tree::splay_tree_remove,
18750         typed_splay_tree::splay_tree_lookup,
18751         typed_splay_tree::splay_tree_predecessor,
18752         typed_splay_tree::splay_tree_successor,
18753         typed_splay_tree::splay_tree_min,
18754         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
18755         (typed_splay_tree::root, typed_splay_tree::comp,
18756         typed_splay_tree::delete_key,
18757         typed_splay_tree::delete_value): New data members.
18758         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
18759         typed_splay_tree::remove.
18761 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
18763         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
18764         -mginv and -mno-ginv to the assembler.
18765         * config/mips/mips.opt (-mcrc): New option.
18766         (-mginv): Likewise.
18767         * doc/invoke.text (-mcrc): Document.
18768         (-mginv): Likewise.
18770 2018-06-15  Nick Clifton  <nickc@redhat.com>
18772         PR 84195
18773         * tree.c (escaped_string): New class.  Converts an unescaped
18774         string into its escaped equivalent.
18775         (warn_deprecated_use): Use the new class to convert the
18776         deprecation message, if present.
18777         (test_escaped_strings): New self test.
18778         (test_c_tests): Add test_escaped_strings.
18779         * doc/extend.texi (deprecated): Add a note that the
18780         deprecation message is affected by the -fmessage-length
18781         option, and that control characters will be escaped.
18782         (#pragma GCC error): Document this pragma.
18783         (#pragma GCC warning): Likewise.
18784         * doc/invoke.texi (-fmessage-length): Document this option's
18785         effect on the #warning and #error preprocessor directives and
18786         the deprecated attribute.
18788 2018-06-15  Richard Biener  <rguenther@suse.de>
18790         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
18791         here, also noting vector size used.
18792         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
18793         size used in MSG_OPTIMIZED_LOCATIONS dump.
18794         (pass_slp_vectorize::execute): Adjust.
18796 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
18798         PR target/85968
18799         * config/arc/arc.c (arc_return_address_register): Fix
18800         if-condition.
18802 2018-06-15  Richard Biener  <rguenther@suse.de>
18804         PR middle-end/86159
18805         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
18806         leave useless conversion stripping to force_gimple_operand_gsi.
18807         (gimplify_build2): Likewise.
18808         (gimplify_build1): Likewise.
18810 2018-06-15  Richard Biener  <rguenther@suse.de>
18812         PR middle-end/86076
18813         * tree-cfg.c (move_stmt_op): unshare invariant addresses
18814         before adjusting their block.
18816 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18818         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
18819         multilibs for *-*-rtems*.
18820         * config/riscv/t-rtems: New file.
18822 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
18824         PR middle-end/86122
18825         * match.pd ((A +- CST1) +- CST2): Punt if last resort
18826         unsigned_type_for returns NULL.
18828         PR target/85945
18829         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
18830         subregs of multi-word pseudos unless the float mode has word size.
18832 2018-06-14  Richard Biener  <rguenther@suse.de>
18834         PR middle-end/86139
18835         * tree-vect-generic.c (build_word_mode_vector_type): Remove
18836         duplicate and harmful type_hash_canon.
18837         * tree.c (type_hash_canon): Assert we didn't find ourselves.
18839 2018-06-14  Richard Biener  <rguenther@suse.de>
18841         PR ipa/86124
18842         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
18843         NULL cgraph_node.
18845 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18847         * config/rtems.h (STDINT_LONG32): Define.
18849 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
18850             Prachi Godbole  <prachi.godbole@imgtec.com>
18852         * config/mips/mips-cpus.def: Define P6600.
18853         * config/mips/mips-tables.opt: Regenerate.
18854         * config/mips/mips.c (mips_ucbranch_type): New enum.
18855         (mips_rtx_cost_data): Add support for P6600.
18856         (mips_issue_rate): Likewise.
18857         (mips_multipass_dfa_lookahead): Likewise.
18858         (mips_avoid_hazard): Likewise.
18859         (mips_reorg_process_insns): Likewise.
18860         (mips_classify_branch_p6600): New function.
18861         * config/mips/mips.h (TUNE_P6600): New define.
18862         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
18863         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
18864         * config/mips/mips.md: Include p6600.md.
18865         (processor): Add p6600.
18866         * config/mips/p6600.md: New file.
18867         * doc/invoke.texi: Add p6600 to supported architectures.
18869 2018-06-13  Martin Sebor  <msebor@redhat.com>
18871         PR tree-optimization/86114
18872         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
18873         of integer types.
18874         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
18876 2018-06-13  Richard Biener  <rguenther@suse.de>
18878         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
18879         Properly set vector type of the intermediate stmt.
18880         * tree-vect-stmts.c (vectorizable_operation): The destination
18881         var always has vectype_out type.
18883 2018-06-13  Jeff Law  <law@redhat.com>
18885         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
18886         integer 0 for argument to print_rtl_with_bb.
18887         (rl78_reorg): Likewise.
18889 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
18891         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
18892         from rtx to rtx_insn *.
18893         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
18894         "label".
18895         (add_sched_insns_for_speculation): Likewise for local "target",
18896         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18897         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
18898         from rtx_insn ** to rtx_code_label **.
18899         (reorg_emit_nops): Likewise.
18900         (c6x_reorg): Likewise for local "call_labels".
18901         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
18902         rtx to rtx_insn *.
18903         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
18904         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
18905         the loops over LABEL_REFS.
18906         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
18907         braf_label.
18908         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18909         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
18910         (split_branches): Strengthen local "olabel" from rtx to
18911         rtx_insn *, adding a safe_as_a cast.
18912         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
18913         to "rtx_insn *".
18914         (add_insn_after): Likewise for first two params.
18915         (add_insn_before): Likewise.
18916         (remove_insn): Likewise for param.
18917         (emit_pattern_before_noloc): Likewise for second and third params.
18918         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
18919         (emit_call_insn_before_noloc): Likewise.
18920         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
18921         to "rtx_insn *".
18922         (emit_barrier_before): Likewise.
18923         (emit_label_before): Strengthen "label" param from "rtx" to
18924         "rtx_code_label *".  Strengthen "before" param from "rtx" to
18925         "rtx_insn *".
18926         (emit_insn_after_1): Strengthen "after" param from "rtx" to
18927         "rtx_insn *".
18928         (emit_pattern_after_noloc): Likewise.
18929         (emit_insn_after_noloc): Likewise.
18930         (emit_jump_insn_after_noloc): Likewise.
18931         (emit_call_insn_after_noloc): Likewise.
18932         (emit_debug_insn_after_noloc): Likewise.
18933         (emit_barrier_after): Likewise.
18934         (emit_label_after): Likewise for both params.
18935         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
18936         "loc" param from "int" to "location_t".
18937         (emit_insn_after_setloc): Likewise.
18938         (emit_jump_insn_after_setloc): Likewise.
18939         (emit_call_insn_after_setloc): Likewise.
18940         (emit_debug_insn_after_setloc): Likewise.
18941         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
18942         "loc" param from "int" to "location_t".
18943         (emit_pattern_before): Convert NULL_RTX to NULL.
18944         (emit_insn_before_setloc): Convert "loc" param from "int" to
18945         "location_t".
18946         (emit_jump_insn_before_setloc): Likewise.
18947         (emit_call_insn_before_setloc): Likewise.
18948         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
18949         rtx_insn *.  Convert "loc" param from "int" to "location_t".
18950         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
18951         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
18952         Convert 3rd param from "int" to "location_t".
18953         (emit_barrier_before, emit_barrier_after, next_real_insn):
18954         Strengthen param from rtx to rtx_insn *.
18955         (emit_label_before): Strengthen 1st param from "rtx" to
18956         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
18957         "rtx_insn *".
18958         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
18959         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
18960         Strengthen 2nd param from "rtx" to "rtx_insn *".
18961         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
18962         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
18963         Likewise. Convert 3rd param from "int" to "location_t".
18964         (emit_label_after): Strengthen 1st param from "rtx" to
18965         "rtx_code_label *".
18966         (next_real_insn, remove_insn): Strengthen param from "rtx" to
18967         "rtx_insn *".
18968         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
18969         from "rtx" to "rtx_insn *".
18971 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
18973         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
18974         bodies streamed in with -Q.
18975         * dumpfile.c (dump_files): Add lto-stream-out dump file.
18976         * dumpfile.h (tree_dump_index): Add lto_stream_out.
18977         * gimple-streamer-out.c: Include gimple-pretty-print.h
18978         (output_bb): Dump stmts streamed.
18979         * lto-section-out.c: Include print-tree.h
18980         (lto_begin_section): Dump sections created.
18981         (lto_output_decl_index): Dump decl encoded.
18982         * lto-streamer-out.c: Include print-tree.h
18983         (create_output_block): Dump output block created.
18984         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
18985         (output_function): Dump function output.
18986         (output_constructor): Dump constructor streamed.
18987         (write_global_stream): Output indexes encoded.
18988         (produce_asm_for_decls): Dump streams encoded.
18989         * lto-streamer.c (streamer_dump_file): New global var.
18990         * lto-streamer.h (streamer_dump_file): Declare.
18991         * passes.c (ipa_write_summaries): Initialize streamer dump.
18992         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
18993         in.
18995 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
18997         PR target/86048
18998         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
18999         offsets for register save directives.  Emit a second batch of save
19000         directives, if need be, when the function accesses prior frames.
19002 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19004         * config/arc/fpu.md (fmasf4): Force operand to register.
19005         (fnmasf4): Likewise.
19007 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19009         * config/arc/arc-protos.h (arc_pad_return): Remove.
19010         * config/arc/arc.c (machine_function): Remove force_short_suffix
19011         and size_reason.
19012         (arc_print_operand): Adjust printing of '&'.
19013         (arc_verify_short): Remove conditional printing of short suffix.
19014         (arc_final_prescan_insn): Remove reference to size_reason.
19015         (pad_return): New function.
19016         (arc_reorg): Call pad_return.
19017         (arc_pad_return): Remove.
19018         (arc_init_machine_status): Remove reference to force_short_suffix.
19019         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
19020         (attr length): When attribute iscompact is true force to 2
19021         regardless; in the case of maybe check if we want to force the
19022         instruction to have 4 bytes length.
19023         (nopv): Change it to generate 4 byte long nop as well.
19024         (blockage): New pattern.
19025         (simple_return): Remove call to arc_pad_return.
19026         (p_return_i): Likewise.
19028 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19030         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
19032 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19034         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
19035         ARC cores.
19037 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19039         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
19040         for ARC700 and ARCv2.
19042 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
19044         PR target/86076
19045         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
19046         operands[2] instead of operands[1].
19049 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
19051         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
19052         case, check whether the outer register overlaps an unallocatable
19053         register, not just whether it fits the required class.
19055 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
19057         * poly-int.h (can_div_trunc_p): Add new overload in which all values
19058         are poly_ints.
19059         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
19060         (memrefs_conflict_p): Likewise.
19061         (init_alias_analysis): Likewise.
19062         * cfgexpand.c (expand_debug_expr): Likewise.
19063         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
19064         * cse.c (fold_rtx): Likewise.
19065         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
19066         * expr.c (emit_block_move_hints): Likewise.
19067         (clear_storage_hints, push_block, emit_push_insn): Likewise.
19068         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
19069         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
19070         (emit_group_store): Likewise.
19071         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
19072         to read the PRE/POST_MODIFY increment.
19073         * calls.c (store_one_arg): Use strip_offset.
19074         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
19075         poly_int_rtx_p.
19076         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
19077         by a VEC_SELECT.
19078         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
19079         (simplify_binary_operation_1): Extend CONST_INT handling to
19080         poly_int_rtx_p.
19081         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
19082         than a HOST_WIDE_INT.
19083         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
19084         poly_int64.
19085         (adjust_mems, add_stores): Update accodingly.
19086         (vt_canonicalize_addr): Track polynomial offsets.
19087         (emit_note_insn_var_location): Likewise.
19088         (vt_add_function_parameter): Likewise.
19089         (vt_initialize): Likewise.
19091 2018-06-12  Jeff Law  <law@redhat.com>
19093         * config.gcc (alpha*-*-freebsd*): Remove.
19094         * config/alpha/freebsd.h: Remove.
19096 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
19098         PR other/69968
19099         * spellcheck-tree.c (levenshtein_distance): Rename to...
19100         (get_edit_distance): ...this, and update for underlying renaming.
19101         * spellcheck-tree.h (levenshtein_distance): Rename to...
19102         (get_edit_distance): ...this.
19103         * spellcheck.c (levenshtein_distance): Rename to...
19104         (get_edit_distance): ...this.  Convert from Levenshtein distance
19105         to Damerau-Levenshtein distance by supporting transpositions of
19106         adjacent characters.  Rename "v1" to "v_next" and "v0" to
19107         "v_one_ago".
19108         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
19109         (selftest::test_edit_distance_unit_test_oneway): ...this, and
19110         update for underlying renaming.
19111         (selftest::levenshtein_distance_unit_test): Rename to...
19112         (selftest::test_get_edit_distance_unit): ...this, and update for
19113         underlying renaming.
19114         (selftest::test_find_closest_string): Add example from PR 69968
19115         where transposition helps
19116         (selftest::test_metric_conditions): Update for renaming.
19117         (selftest::test_metric_conditions): Likewise.
19118         (selftest::spellcheck_c_tests): Likewise.
19119         * spellcheck.h (levenshtein_distance): Rename both overloads to...
19120         (get_edit_distance): ...this.
19121         (best_match::consider): Update for renaming.
19123 2018-06-12  Martin Sebor  <msebor@redhat.com>
19125         PR tree-optimization/85259
19126         * builtins.c (compute_objsize): Handle constant offsets.
19127         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
19128         true iff a warning has been issued.
19129         * gimple.h (gimple_nonartificial_location): New function.
19130         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
19131         gimple_nonartificial_location and handle -Wno-system-headers.
19132         (handle_builtin_stxncpy): Same.
19134 2018-06-12  Martin Sebor  <msebor@redhat.com>
19136         PR c/85931
19137         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
19139 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
19141         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
19142         BUILTIN_VEC_XST entries for pointer to double and long long.
19144 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
19146         PR target/85990
19147         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19148         Update comments.
19149         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19150         Likewise.
19152 2018-06-12  Martin Liska  <mliska@suse.cz>
19154         * doc/options.texi: Document IntegerRange.
19156 2018-06-12  Martin Liska  <mliska@suse.cz>
19158         * config/i386/i386.opt: Make MPX-related options as Deprecated.
19159         * opt-functions.awk: Handle Deprecated flag.
19160         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
19161         and report error.
19162         (read_cmdline_option): Report warning for a deprecated option.
19163         * opts.h (struct cl_option): Add new field cl_deprecated.
19164         (CL_ERR_DEPRECATED): New.
19166 2018-06-12  Martin Liska  <mliska@suse.cz>
19168         * doc/options.texi: Document Deprecated option flag.
19170 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19172         * config/arc/arc-arch.h (arc_extras): New enum.
19173         (arc_cpu_t):Add field extra.
19174         (arc_cpu_types): Consider the extras.
19175         * config/arc/arc-cpus.def: Add extras info.
19176         * config/arc/arc-opts.h (processor_type): Consider extra field.
19177         * config/arc/arc.c (arc_override_options): Handle extra field.
19179 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
19181         * config/arc/arc-arch.h: Update ARC_OPTX macro.
19182         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
19183         field.
19184         * config/arc/arc.c (arc_init): Update pic warning.
19185         (irq_range): Update irq range parsing warnings.
19186         (arc_override_options): Update various warning messages.
19187         (arc_handle_aux_attribute): Likewise.
19189 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
19191         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
19193 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19195         * doc/sourcebuild.texi: Document usage of line number 0 in verify
19196         compiler messages directives.
19198 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
19200         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
19201         * config/mips/mips-tables.opt: Regenerate.
19202         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
19203         mips64r6.
19204         * doc/invoke.texi: Document -march=i6500.
19206 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
19208         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
19209         (i6400_gpmul): Add cpu_unit.
19210         (i6400_gpdiv): Likewise.
19211         (i6400_msa_add_d): Update reservations.
19212         (i6400_msa_int_add) Likewise.
19213         (i6400_msa_short_logic3) Likewise.
19214         (i6400_msa_short_logic2) Likewise.
19215         (i6400_msa_short_logic) Likewise.
19216         (i6400_msa_move) Likewise.
19217         (i6400_msa_cmp) Likewise.
19218         (i6400_msa_short_float2) Likewise.
19219         (i6400_msa_div_d) Likewise.
19220         (i6400_msa_long_logic1) Likewise.
19221         (i6400_msa_long_logic2) Likewise.
19222         (i6400_msa_mult) Likewise.
19223         (i6400_msa_long_float2) Likewise.
19224         (i6400_msa_long_float4) Likewise.
19225         (i6400_msa_long_float5) Likewise.
19226         (i6400_msa_long_float8) Likewise.
19227         (i6400_fpu_fadd): Include frint type.
19228         (i6400_fpu_store): New define_insn_reservation.
19229         (i6400_fpu_load): Likewise.
19230         (i6400_fpu_move): Likewise.
19231         (i6400_fpu_fcmp): Likewise.
19232         (i6400_fpu_fmadd): Likewise.
19233         (i6400_int_mult): Include imul3nc type and update reservation.
19234         (i6400_int_div): Include idiv3 type and update reservation.
19235         (i6400_int_load): Update to check type not move_type.
19236         (i6400_int_store): Likewise.
19237         (i6400_int_prefetch): Set zero latency.
19239 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
19241         * gcc.c: Document new %@{...} sequence.
19242         (LINK_COMMAND_SPEC): Use it for the -L switches.
19243         (cpp_unique_options): Use it for the -I switches.
19244         (at_file_argbuf): New global variable.
19245         (in_at_file): Likewise.
19246         (alloc_args): Create at_file_argbuf.
19247         (clear_args): Truncate at_file_argbuf.
19248         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
19249         (open_at_file): New function.
19250         (close_at_file): Likewise.
19251         (create_at_file): Delete.
19252         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
19253         <'o'>: Likewise.
19254         <'@'>: New case.
19255         (validate_switches_from_spec): Deal with %@{...} sequence.
19256         (validate_switches): Likewise.
19257         (driver::finalize): Call clear_args.
19259 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
19261         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
19263 2018-06-11  Martin Sebor  <msebor@redhat.com>
19265         * doc/invoke.texi (-Wall): List -Wc++17-compat.
19266         (Wno-class-memaccess): Add @opindex.
19267         (Wno-templates, Wno-multiple-inheritance): Same.
19268         (Wno-virtual-inheritance, Wno-namespaces): Same.
19269         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
19270         (Wno-format-overflow, Wno-format-truncation): Same.
19271         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
19272         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
19273         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
19274         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
19275         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
19276         (Wno-misspelled-isr): Same.
19278 2018-06-11  Martin Sebor  <msebor@redhat.com>
19280         * PR tree-optimization/86083
19281         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
19283 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
19285         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
19287 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
19289         PR target/85755
19290         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
19291         on the correct operand.
19292         (*movdi_internal64): Ditto.
19294 2018-06-11  Martin Liska  <mliska@suse.cz>
19296         PR tree-optimization/86089
19297         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
19299 2018-06-11  Julia Koval  <julia.koval@intel.com>
19301         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
19302         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
19303         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
19305 2018-06-11  Olivier Hainque  <hainque@adacore.com>
19307         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
19308         for Ada with strict dwarf2.
19310 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
19312         PR target/85755
19313         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
19314         addresses.
19316 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
19318         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
19320 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
19322         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
19323         TARGET_ELF.
19325 2018-06-08  Martin Liska  <mliska@suse.cz>
19327         * tree-cfg.h (debug_function): Fix argument type to match
19328         implementation.
19330 2018-06-08  Martin Liska  <mliska@suse.cz>
19332         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
19333         Remove usage of MPX-related (and removed) fields.
19334         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
19336 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
19338         * cfg.c (debug): Use TDF_NONE rather than 0.
19339         * cfghooks.c (debug): Likewise.
19340         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
19341         (struct dump_option_value_info): Convert to...
19342         (struct kv_pair): ...this template type.
19343         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
19344         rather than 0.
19345         (optinfo_verbosity_options): Likewise.
19346         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
19347         OPTGROUP_NONE.
19348         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
19349         than int for "optgroup_flags" param.
19350         (dump_generic_expr_loc): Use dump_flags_t rather than int for
19351         "dump_kind" param.
19352         (dump_dec): Likewise.
19353         (dump_finish): Use TDF_NONE rather than 0.
19354         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
19355         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
19356         than 0.  Update for change to option_ptr.
19357         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
19358         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
19359         0.  Update for changes to optinfo_verbosity_options and
19360         optgroup_options.
19361         (opt_info_switch_p): Convert optgroup_flags from int to
19362         optgroup_flags_t.
19363         (dump_basic_block): Use dump_flags_t rather than int
19364         for "dump_kind" param.
19365         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
19366         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
19367         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
19368         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
19369         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
19370         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
19371         TDF_NONE): Convert from macros to...
19372         (enum dump_flag): ...this new enum.
19373         (dump_flags_t): Update to use enum.
19374         (operator|, operator&, operator~, operator|=, operator&=):
19375         Implement for dump_flags_t.
19376         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
19377         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
19378         Convert from macros to...
19379         (enum optgroup_flag): ...this new enum.
19380         (optgroup_flags_t): New typedef.
19381         (operator|, operator|=): Implement for optgroup_flags_t.
19382         (struct dump_file_info): Convert field "alt_flags" to
19383         dump_flags_t.  Convert field "optgroup_flags" to
19384         optgroup_flags_t.
19385         (dump_basic_block): Use dump_flags_t rather than int for param.
19386         (dump_generic_expr_loc): Likewise.
19387         (dump_dec): Likewise.
19388         (dump_register): Convert param "optgroup_flags" to
19389         optgroup_flags_t.
19390         (opt_info_enable_passes): Likewise.
19391         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
19392         than 0.
19393         * gimple-pretty-print.c (debug): Likewise.
19394         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
19395         (merged_store_group::apply_stores): Likewise.
19396         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
19397         * gimple.c (verify_gimple_pp): Likewise.
19398         * graphite-poly.c (print_pbb_body): Likewise.
19399         * passes.c (pass_manager::register_one_dump_file): Convert
19400         local "optgroup_flags" to optgroup_flags_t.
19401         * print-tree.c (print_node): Use TDF_NONE rather than 0.
19402         (debug): Likewise.
19403         (debug_body): Likewise.
19404         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
19405         to optgroup_flags_t.
19406         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
19407         than 0.
19408         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
19409         (convert_mult_to_fma): Likewise.
19410         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
19411         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
19412         * tree-vect-data-refs.c (dump_lower_bound): Convert param
19413         "dump_kind" to dump_flags_t.
19415 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
19417         * config/rs6000/rs6000.c (min, max): Delete.
19419 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
19421         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
19422         -mabi=spe and -mabi=no-spe.
19424 2018-06-08  Martin Liska  <mliska@suse.cz>
19426         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
19427         where we expect an existing summary.
19429 2018-06-08  Martin Liska  <mliska@suse.cz>
19431         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
19432         * ipa-inline.h (estimate_edge_growth): Likewise.
19434 2018-06-08  Martin Liska  <mliska@suse.cz>
19436         * cgraph.c (function_version_hasher::hash): Use
19437         cgraph_node::get_uid ().
19438         (function_version_hasher::equal):
19439         * cgraph.h (cgraph_node::get_uid): New method.
19440         * ipa-inline.c (update_caller_keys): Use
19441         cgraph_node::get_uid ().
19442         (update_callee_keys): Likewise.
19443         * ipa-utils.c (searchc): Likewise.
19444         (ipa_reduced_postorder): Likewise.
19445         * lto-cgraph.c (input_node): Likewise.
19446         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
19447         * symbol-summary.h (symtab_insertion): Likewise.
19448         (symtab_removal): Likewise.
19449         (symtab_duplication): Likewise.
19450         * tree-pretty-print.c (dump_function_header): Likewise.
19451         * tree-sra.c (convert_callers_for_node): Likewise.
19453 2018-06-08  Martin Liska  <mliska@suse.cz>
19455         * cgraph.c (symbol_table::create_edge): Always assign a new
19456         unique number.
19457         (symbol_table::free_edge): Do not recycle numbers.
19458         * cgraph.h (cgraph_edge::get): New method.
19459         * symbol-summary.h (symtab_removal): Use it.
19460         (symtab_duplication): Likewise.
19461         (call_summary::hashable_uid): Remove.
19463 2018-06-08  Martin Liska  <mliska@suse.cz>
19465         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
19466         (initialize_growth_caches): Remove.
19467         (free_growth_caches): Likewise.
19468         (do_estimate_edge_time): Use edge_growth_cache.
19469         (do_estimate_edge_size): Likewise.
19470         (do_estimate_edge_hints): Likewise.
19471         * ipa-inline.c (reset_edge_caches): Likewise.
19472         (recursive_inlining): Likewise.
19473         (inline_small_functions): Likewise.
19474         * ipa-inline.h (initialize_growth_caches): Remove.
19475         (estimate_edge_size): Likewise.
19476         (estimate_edge_time): Likewise.
19477         (estimate_edge_hints): Likewise.
19478         (reset_edge_growth_cache): Likewise.
19479         * symbol-summary.h (call_summary::remove): New method.
19481 2018-06-08  Martin Liska  <mliska@suse.cz>
19483         * ipa-cp.c (class edge_clone_summary): New summary.
19484         (grow_edge_clone_vectors): Remove.
19485         (ipcp_edge_duplication_hook): Remove.
19486         (class edge_clone_summary_t): New call_summary class.
19487         (ipcp_edge_removal_hook): Remove.
19488         (edge_clone_summary_t::duplicate): New function.
19489         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
19490         (create_specialized_node): Likewise.
19491         (ipcp_driver): Initialize edge_clone_summaries and do not
19492         register hooks.
19494 2018-06-08  Martin Liska  <mliska@suse.cz>
19496         * symbol-summary.h (get): New function.
19497         (call_summary::m_initialize_when_cloning): New class member.
19499 2018-06-08  Martin Liska  <mliska@suse.cz>
19501         * cgraph.c (cgraph_node::remove): Do not recycle uid.
19502         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
19503         (symbol_table::allocate_cgraph_symbol): Do not set uid.
19504         * passes.c (uid_hash_t): Record removed_nodes by their uids.
19505         (remove_cgraph_node_from_order): Use the removed_nodes set.
19506         (do_per_function_toporder): Likwise.
19507         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
19508         instead of summary_uid.
19509         (symtab_removal): Likewise.
19510         (symtab_duplication): Likewise.
19512 2018-06-08  Martin Liska  <mliska@suse.cz>
19514         * ipa-cp.c (ipcp_store_bits_results): Use
19515         ipcp_transformation_sum.
19516         (ipcp_store_vr_results): Likewise.
19517         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
19518         to ...
19519         (ipcp_transformation_initialize): ... this.
19520         (ipa_set_node_agg_value_chain):
19521         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
19522         (write_ipcp_transformation_info): Likewise.
19523         (read_ipcp_transformation_info): Likewise.
19524         (ipcp_update_bits): Likewise.
19525         (ipcp_update_vr): Likewise.
19526         (ipcp_transform_function): Likewise.
19527         * ipa-prop.h: Rename ipcp_transformation_summary to
19528         ipcp_transformation.
19529         (class ipcp_transformation_t): New function summary.
19530         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
19531         (ipa_get_agg_replacements_for_node): Likewise.
19533 2018-06-08  Martin Liska  <mliska@suse.cz>
19535         * ipa-pure-const.c (struct funct_state_d): Do it class instead
19536         of struct.
19537         (class funct_state_summary_t): New function_summary class.
19538         (has_function_state): Remove.
19539         (get_function_state): Likewise.
19540         (set_function_state): Likewise.
19541         (add_new_function): Likewise.
19542         (funct_state_summary_t::insert): New function.
19543         (duplicate_node_data): Remove.
19544         (remove_node_data): Remove.
19545         (funct_state_summary_t::duplicate): New function.
19546         (register_hooks): Create new funct_state_summaries.
19547         (pure_const_generate_summary): Use it.
19548         (pure_const_write_summary): Likewise.
19549         (pure_const_read_summary): Likewise.
19550         (propagate_pure_const): Likewise.
19551         (propagate_nothrow): Likewise.
19552         (dump_malloc_lattice): Likewise.
19553         (propagate_malloc): Likewise.
19554         (execute): Do not register hooks, just remove summary
19555         instead.
19556         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
19557         constructor.
19559 2018-06-08  Martin Liska  <mliska@suse.cz>
19561         * ipa-reference.c (remove_node_data): Remove.
19562         (duplicate_node_data): Likewise.
19563         (class ipa_ref_var_info_summary_t): New class.
19564         (class ipa_ref_opt_summary_t): Likewise.
19565         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
19566         (get_reference_optimization_summary): Use
19567         ipa_ref_opt_sum_summaries.
19568         (set_reference_vars_info): Remove.
19569         (set_reference_optimization_summary): Likewise.
19570         (ipa_init): Create summaries.
19571         (init_function_info): Use function summary.
19572         (ipa_ref_opt_summary_t::duplicate): New function.
19573         (ipa_ref_opt_summary_t::remove): New function.
19574         (get_read_write_all_from_node): Fix GNU coding style.
19575         (propagate): Use function summary.
19576         (write_node_summary_p): Fix GNU coding style.
19577         (stream_out_bitmap): Likewise.
19578         (ipa_reference_read_optimization_summary): Use function summary.
19579         (ipa_reference_c_finalize): Do not release hooks.
19581 2018-06-08  Martin Liska  <mliska@suse.cz>
19583         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
19584         (analyze_function_body): Extract multiple calls of get_create.
19585         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
19586         * ipa-inline.c (recursive_inlining): Use ::get method.
19587         * ipa-inline.h (estimate_edge_growth): Likewise.
19589 2018-06-08  Martin Liska  <mliska@suse.cz>
19591         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
19592         HSA_INVALID.
19593         (hsa_function_summary::hsa_function_summary): Use the new enum
19594         value.
19595         (hsa_gpu_implementation_p): Use hsa_summaries::get.
19596         * hsa-gen.c (hsa_get_host_function): Likewise.
19597         (get_brig_function_name): Likewise.
19598         * ipa-hsa.c (process_hsa_functions): Likewise.
19599         (ipa_hsa_write_summary): Likewise.
19600         * symbol-summary.h (symtab_duplication): Use ::get function/
19601         (get): New function.
19603 2018-06-08  Martin Liska  <mliska@suse.cz>
19605         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
19606         of get.
19607         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
19608         (hsa_register_kernel): Likewise.
19609         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
19610         * hsa-gen.c (hsa_get_host_function): Likewise.
19611         (get_brig_function_name): Likewise.
19612         (generate_hsa): Likewise.
19613         (pass_gen_hsail::execute): Likewise.
19614         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
19615         (devirtualization_time_bonus): Likewise.
19616         (ipcp_propagate_stage): Likewise.
19617         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
19618         (edge_set_predicate): Likewise.
19619         (evaluate_conditions_for_known_args): Likewise.
19620         (evaluate_properties_for_edge): Likewise.
19621         (ipa_fn_summary::reset): Likewise.
19622         (ipa_fn_summary_t::duplicate): Likewise.
19623         (dump_ipa_call_summary): Likewise.
19624         (ipa_dump_fn_summary): Likewise.
19625         (analyze_function_body): Likewise.
19626         (compute_fn_summary): Likewise.
19627         (estimate_edge_devirt_benefit): Likewise.
19628         (estimate_edge_size_and_time): Likewise.
19629         (estimate_calls_size_and_time): Likewise.
19630         (estimate_node_size_and_time): Likewise.
19631         (inline_update_callee_summaries): Likewise.
19632         (remap_edge_change_prob): Likewise.
19633         (remap_edge_summaries): Likewise.
19634         (ipa_merge_fn_summary_after_inlining): Likewise.
19635         (ipa_update_overall_fn_summary): Likewise.
19636         (read_ipa_call_summary): Likewise.
19637         (inline_read_section): Likewise.
19638         (write_ipa_call_summary): Likewise.
19639         (ipa_fn_summary_write): Likewise.
19640         (ipa_free_fn_summary): Likewise.
19641         * ipa-hsa.c (process_hsa_functions): Likewise.
19642         (ipa_hsa_write_summary): Likewise.
19643         (ipa_hsa_read_section): Likewise.
19644         * ipa-icf.c (sem_function::merge): Likewise.
19645         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
19646         (do_estimate_edge_time): Likewise.
19647         (estimate_size_after_inlining): Likewise.
19648         (estimate_growth): Likewise.
19649         (growth_likely_positive): Likewise.
19650         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
19651         (inline_call): Likewise.
19652         * ipa-inline.c (caller_growth_limits): Likewise.
19653         (can_inline_edge_p): Likewise.
19654         (can_inline_edge_by_limits_p): Likewise.
19655         (compute_uninlined_call_time): Likewise.
19656         (compute_inlined_call_time): Likewise.
19657         (want_inline_small_function_p): Likewise.
19658         (edge_badness): Likewise.
19659         (update_caller_keys): Likewise.
19660         (update_callee_keys): Likewise.
19661         (recursive_inlining): Likewise.
19662         (inline_small_functions): Likewise.
19663         (inline_to_all_callers_1): Likewise.
19664         (dump_overall_stats): Likewise.
19665         (early_inline_small_functions): Likewise.
19666         (early_inliner): Likewise.
19667         * ipa-inline.h (estimate_edge_growth): Likewise.
19668         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
19669         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
19670         * ipa-prop.h (IPA_NODE_REF): Likewise.
19671         (IPA_EDGE_REF): Likewise.
19672         * ipa-pure-const.c (malloc_candidate_p): Likewise.
19673         (propagate_malloc): Likewise.
19674         * ipa-split.c (execute_split_functions): Likewise.
19675         * symbol-summary.h: Rename get to get_create.
19676         (get): Likewise.
19677         (get_create): Likewise.
19678         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
19680 2018-06-08  Martin Liska  <mliska@suse.cz>
19682         * symbol-summary.h (release): Move definition out of class
19683         declaration.
19684         (symtab_removal): Likewise.
19685         (symtab_duplication): Likewise.
19687 2018-06-08  Martin Liska  <mliska@suse.cz>
19689         * symbol-summary.h (function_summary): Move constructor
19690         implementation out of class declaration.
19691         (release): Likewise.
19692         (symtab_insertion): Likewise.
19693         (symtab_removal): Likewise.
19694         (symtab_duplication): Likewise.
19695         (get): Likewise.
19697 2018-06-08  Martin Liska  <mliska@suse.cz>
19699         * Makefile.in: Remove support for MPX (macros, related functions,
19700         fields in cgraph_node, ...).
19701         * builtin-types.def (BT_BND): Likewise.
19702         (BT_FN_BND_CONST_PTR): Likewise.
19703         (BT_FN_CONST_PTR_BND): Likewise.
19704         (BT_FN_VOID_PTR_BND): Likewise.
19705         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
19706         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
19707         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
19708         (expand_builtin_mempcpy_with_bounds): Likewise.
19709         (expand_builtin_memset_with_bounds): Likewise.
19710         (expand_builtin_memset_args): Likewise.
19711         (std_expand_builtin_va_start): Likewise.
19712         (expand_builtin): Likewise.
19713         (expand_builtin_with_bounds): Likewise.
19714         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
19715         (DEF_LIB_BUILTIN_CHKP): Likewise.
19716         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
19717         (DEF_CHKP_BUILTIN): Likewise.
19718         (BUILT_IN_MEMCPY): Likewise.
19719         (BUILT_IN_MEMMOVE): Likewise.
19720         (BUILT_IN_MEMPCPY): Likewise.
19721         (BUILT_IN_MEMSET): Likewise.
19722         (BUILT_IN_STPCPY): Likewise.
19723         (BUILT_IN_STRCAT): Likewise.
19724         (BUILT_IN_STRCHR): Likewise.
19725         (BUILT_IN_STRCPY): Likewise.
19726         (BUILT_IN_STRLEN): Likewise.
19727         (BUILT_IN_MEMCPY_CHK): Likewise.
19728         (BUILT_IN_MEMMOVE_CHK): Likewise.
19729         (BUILT_IN_MEMPCPY_CHK): Likewise.
19730         (BUILT_IN_MEMSET_CHK): Likewise.
19731         (BUILT_IN_STPCPY_CHK): Likewise.
19732         (BUILT_IN_STRCAT_CHK): Likewise.
19733         (BUILT_IN_STRCPY_CHK): Likewise.
19734         * calls.c (store_bounds): Likewise.
19735         (emit_call_1): Likewise.
19736         (special_function_p): Likewise.
19737         (maybe_warn_nonstring_arg): Likewise.
19738         (initialize_argument_information): Likewise.
19739         (finalize_must_preallocate): Likewise.
19740         (compute_argument_addresses): Likewise.
19741         (expand_call): Likewise.
19742         * cfgexpand.c (expand_call_stmt): Likewise.
19743         (expand_return): Likewise.
19744         (expand_gimple_stmt_1): Likewise.
19745         (pass_expand::execute): Likewise.
19746         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
19747         (cgraph_node::remove): Likewise.
19748         (cgraph_node::dump): Likewise.
19749         (cgraph_node::verify_node): Likewise.
19750         * cgraph.h (chkp_function_instrumented_p): Likewise.
19751         (symtab_node::get_alias_target): Likewise.
19752         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
19753         (cgraph_local_p): Likewise.
19754         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
19755         (cgraph_edge::rebuild_references): Likewise.
19756         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
19757         (walk_polymorphic_call_targets): Likewise.
19758         (cgraph_node::expand_thunk): Likewise.
19759         (symbol_table::output_weakrefs): Likewise.
19760         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
19761         (ix86_handle_option): Likewise.
19762         * config/i386/constraints.md: Likewise.
19763         * config/i386/i386-builtin-types.def (BND): Likewise.
19764         (VOID): Likewise.
19765         (PVOID): Likewise.
19766         (ULONG): Likewise.
19767         * config/i386/i386-builtin.def (BDESC_END): Likewise.
19768         (BDESC_FIRST): Likewise.
19769         (BDESC): Likewise.
19770         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
19771         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
19772         * config/i386/i386.c (enum reg_class): Likewise.
19773         (ix86_target_string): Likewise.
19774         (ix86_option_override_internal): Likewise.
19775         (ix86_conditional_register_usage): Likewise.
19776         (ix86_valid_target_attribute_inner_p): Likewise.
19777         (ix86_set_indirect_branch_type): Likewise.
19778         (ix86_set_current_function): Likewise.
19779         (ix86_function_arg_regno_p): Likewise.
19780         (init_cumulative_args): Likewise.
19781         (ix86_function_arg_advance): Likewise.
19782         (ix86_function_arg): Likewise.
19783         (ix86_pass_by_reference): Likewise.
19784         (ix86_function_value_regno_p): Likewise.
19785         (ix86_function_value_1): Likewise.
19786         (ix86_function_value_bounds): Likewise.
19787         (ix86_return_in_memory): Likewise.
19788         (ix86_setup_incoming_vararg_bounds): Likewise.
19789         (ix86_va_start): Likewise.
19790         (indirect_thunk_need_prefix): Likewise.
19791         (print_reg): Likewise.
19792         (ix86_print_operand): Likewise.
19793         (ix86_expand_call): Likewise.
19794         (ix86_output_function_return): Likewise.
19795         (reg_encoded_number): Likewise.
19796         (BDESC_VERIFYS): Likewise.
19797         (ix86_init_mpx_builtins): Likewise.
19798         (ix86_init_builtins): Likewise.
19799         (ix86_emit_cmove): Likewise.
19800         (ix86_emit_move_max): Likewise.
19801         (ix86_expand_builtin): Likewise.
19802         (ix86_builtin_mpx_function): Likewise.
19803         (ix86_get_arg_address_for_bt): Likewise.
19804         (ix86_load_bounds): Likewise.
19805         (ix86_store_bounds): Likewise.
19806         (ix86_load_returned_bounds): Likewise.
19807         (ix86_store_returned_bounds): Likewise.
19808         (ix86_class_likely_spilled_p): Likewise.
19809         (ix86_hard_regno_mode_ok): Likewise.
19810         (x86_order_regs_for_local_alloc): Likewise.
19811         (ix86_mitigate_rop): Likewise.
19812         (ix86_bnd_prefixed_insn_p): Likewise.
19813         (ix86_mpx_bound_mode): Likewise.
19814         (ix86_make_bounds_constant): Likewise.
19815         (ix86_initialize_bounds): Likewise.
19816         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
19817         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
19818         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
19819         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
19820         (TARGET_CHKP_BOUND_MODE): Likewise.
19821         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
19822         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
19823         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
19824         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
19825         * config/i386/i386.h (TARGET_MPX): Likewise.
19826         (TARGET_MPX_P): Likewise.
19827         (VALID_BND_REG_MODE): Likewise.
19828         (FIRST_BND_REG): Likewise.
19829         (LAST_BND_REG): Likewise.
19830         (enum reg_class): Likewise.
19831         (BND_REG_P): Likewise.
19832         (BND_REGNO_P): Likewise.
19833         (BNDmode): Likewise.
19834         (ADJUST_INSN_LENGTH): Likewise.
19835         * config/i386/i386.md: Likewise.
19836         * config/i386/i386.opt: Likewise.
19837         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
19838         (defined): Likewise.
19839         (LINK_MPX): Likewise.
19840         (MPX_SPEC): Likewise.
19841         (LIBMPX_SPEC): Likewise.
19842         (LIBMPXWRAPPERS_SPEC): Likewise.
19843         (CHKP_SPEC): Likewise.
19844         * config/i386/predicates.md: Likewise.
19845         * dbxout.c (dbxout_type): Likewise.
19846         * doc/extend.texi: Likewise.
19847         * doc/invoke.texi: Likewise.
19848         * doc/md.texi: Likewise.
19849         * doc/tm.texi: Likewise.
19850         * doc/tm.texi.in: Likewise.
19851         * dwarf2out.c (is_base_type): Likewise.
19852         (gen_formal_types_die): Likewise.
19853         (gen_subprogram_die): Likewise.
19854         (gen_type_die_with_usage): Likewise.
19855         (gen_decl_die): Likewise.
19856         (dwarf2out_late_global_decl): Likewise.
19857         * expr.c (expand_assignment): Likewise.
19858         (emit_storent_insn): Likewise.
19859         (store_expr_with_bounds): Likewise.
19860         (store_expr): Likewise.
19861         (expand_expr_real_1): Likewise.
19862         * expr.h (store_expr_with_bounds): Likewise.
19863         * function.c (use_register_for_decl): Likewise.
19864         (struct bounds_parm_data): Likewise.
19865         (assign_parms_augmented_arg_list): Likewise.
19866         (assign_parm_find_entry_rtl): Likewise.
19867         (assign_parm_is_stack_parm): Likewise.
19868         (assign_parm_load_bounds): Likewise.
19869         (assign_bounds): Likewise.
19870         (assign_parms): Likewise.
19871         (expand_function_start): Likewise.
19872         * gcc.c (CHKP_SPEC): Likewise.
19873         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
19874         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
19875         (wrestrict_dom_walker::check_call): Likewise.
19876         * gimple.c (gimple_build_call_from_tree): Likewise.
19877         * gimple.h (enum gf_mask): Likewise.
19878         (gimple_call_with_bounds_p): Likewise.
19879         (gimple_call_set_with_bounds): Likewise.
19880         * gimplify.c (gimplify_init_constructor): Likewise.
19881         * ipa-cp.c (initialize_node_lattices): Likewise.
19882         (propagate_constants_across_call): Likewise.
19883         (find_more_scalar_values_for_callers_subset): Likewise.
19884         * ipa-hsa.c (process_hsa_functions): Likewise.
19885         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
19886         * ipa-icf.c (sem_function::merge): Likewise.
19887         * ipa-inline.c (early_inliner): Likewise.
19888         * ipa-pure-const.c (warn_function_noreturn): Likewise.
19889         (warn_function_cold): Likewise.
19890         (propagate_pure_const): Likewise.
19891         * ipa-ref.h (enum GTY): Likewise.
19892         * ipa-split.c (find_retbnd): Likewise.
19893         (consider_split): Likewise.
19894         (split_function): Likewise.
19895         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
19896         * ipa.c (walk_polymorphic_call_targets): Likewise.
19897         (symbol_table::remove_unreachable_nodes): Likewise.
19898         (process_references): Likewise.
19899         (cgraph_build_static_cdtor_1): Likewise.
19900         * lto-cgraph.c (lto_output_node): Likewise.
19901         (output_refs): Likewise.
19902         (compute_ltrans_boundary): Likewise.
19903         (input_overwrite_node): Likewise.
19904         (input_node): Likewise.
19905         (input_cgraph_1): Likewise.
19906         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
19907         * passes.c (pass_manager::execute_early_local_passes): Likewise.
19908         (class pass_chkp_instrumentation_passes): Likewise.
19909         (make_pass_chkp_instrumentation_passes): Likewise.
19910         * passes.def: Likewise.
19911         * rtl.h (struct GTY): Likewise.
19912         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
19913         * stor-layout.c (layout_type): Likewise.
19914         * symtab.c: Likewise.
19915         * target.def: Likewise.
19916         * targhooks.c (default_chkp_bound_type): Likewise.
19917         (default_chkp_bound_mode): Likewise.
19918         (default_builtin_chkp_function): Likewise.
19919         (default_chkp_function_value_bounds): Likewise.
19920         (default_chkp_make_bounds_constant): Likewise.
19921         (default_chkp_initialize_bounds): Likewise.
19922         * targhooks.h (default_chkp_bound_type): Likewise.
19923         (default_chkp_bound_mode): Likewise.
19924         (default_builtin_chkp_function): Likewise.
19925         (default_chkp_function_value_bounds): Likewise.
19926         (default_chkp_make_bounds_constant): Likewise.
19927         (default_chkp_initialize_bounds): Likewise.
19928         * toplev.c (compile_file): Likewise.
19929         (process_options): Likewise.
19930         * tree-core.h (DEF_BUILTIN): Likewise.
19931         (DEF_BUILTIN_CHKP): Likewise.
19932         * tree-inline.c (declare_return_variable): Likewise.
19933         (remap_gimple_stmt): Likewise.
19934         (copy_bb): Likewise.
19935         (initialize_inlined_parameters): Likewise.
19936         (expand_call_inline): Likewise.
19937         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
19938         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
19939         (make_pass_ipa_chkp_produce_thunks): Likewise.
19940         (make_pass_chkp): Likewise.
19941         (make_pass_chkp_opt): Likewise.
19942         (make_pass_chkp_instrumentation_passes): Likewise.
19943         * tree-pretty-print.c (dump_generic_node): Likewise.
19944         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
19945         * tree-ssa-dce.c (propagate_necessity): Likewise.
19946         (eliminate_unnecessary_stmts): Likewise.
19947         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
19948         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
19949         * tree-ssa-sccvn.h: Likewise.
19950         * tree-ssa-strlen.c (get_string_length): Likewise.
19951         (valid_builtin_call): Likewise.
19952         (adjust_last_stmt): Likewise.
19953         (handle_builtin_strchr): Likewise.
19954         (handle_builtin_strcpy): Likewise.
19955         (handle_builtin_stxncpy): Likewise.
19956         (handle_builtin_memcpy): Likewise.
19957         (handle_builtin_strcat): Likewise.
19958         (strlen_check_and_optimize_stmt): Likewise.
19959         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
19960         * tree-streamer-in.c: Likewise.
19961         * tree-streamer.c (record_common_node): Likewise.
19962         * tree.c (tree_code_size): Likewise.
19963         (wide_int_to_tree_1): Likewise.
19964         (type_contains_placeholder_1): Likewise.
19965         (build_common_tree_nodes): Likewise.
19966         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
19967         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
19968         (POINTER_BOUNDS_P): Likewise.
19969         (BOUNDED_TYPE_P): Likewise.
19970         (BOUNDED_P): Likewise.
19971         (CALL_WITH_BOUNDS_P): Likewise.
19972         (pointer_bounds_type_node): Likewise.
19973         * value-prof.c (gimple_ic): Likewise.
19974         * var-tracking.c (vt_add_function_parameters): Likewise.
19975         * varasm.c (make_decl_rtl): Likewise.
19976         (assemble_start_function): Likewise.
19977         (output_constant): Likewise.
19978         (maybe_assemble_visibility): Likewise.
19979         * varpool.c (ctor_for_folding): Likewise.
19980         * chkp-builtins.def: Remove.
19981         * ipa-chkp.c: Remove.
19982         * ipa-chkp.h: Remove.
19983         * rtl-chkp.c: Remove.
19984         * rtl-chkp.h: Remove.
19985         * tree-chkp-opt.c: Remove.
19986         * tree-chkp.c: Remove.
19987         * tree-chkp.h: Remove.
19989 2018-06-07  Carl Love  <cel@us.ibm.com>
19991         * config/rs6000/vsx.md (vextract_fp_from_shorth,
19992         vextract_fp_from_shortl): Add BE support.
19994 2018-06-07  Paul Koning  <ni1d@arrl.net>
19996         * compare-elim.c (try_merge_compare): Don't merge compare if
19997         address contains a side effect.
19998         (try_eliminate_compare): Likewise.
20000 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
20002         * config.gcc: Support "tremont".
20003         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
20004         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20005         PROCESSOR_TREMONT.
20006         * config/i386/i386.c (m_TREMONT): Define.
20007         (processor_target_table): Add "tremont".
20008         (PTA_TREMONT): Define.
20009         (ix86_lea_outperforms): Add TARGET_TREMONT.
20010         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
20011         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
20012         and M_INTEL_GOLDMONT_PLUS.
20013         (fold_builtin_cpu): Add "tremont".
20014         (ix86_add_stmt_cost): Add TARGET_TREMONT.
20015         (ix86_option_override_internal): Add "tremont".
20016         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
20017         (processor_type): Add PROCESSOR_TREMONT.
20018         * config/i386/x86-tune.def: Add m_TREMONT.
20019         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
20021 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20023         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
20024         symbol defined for msp430i* devices to be lower case.
20026 2018-06-07  Richard Biener  <rguenther@suse.de>
20028         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
20029         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
20030         Properly wrap signed arithmetic if overflow wraps.
20032 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
20034         PR tree-optimization/69615
20035         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
20036         of a cast from a same precision integral SSA_NAME in a bb dominated
20037         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
20038         cast to utype if rhs2 has already a compatible type.
20040 2018-06-07  Richard Biener  <rguenther@suse.de>
20042         PR tree-optimization/85935
20043         * graphite-scop-detection.c (find_params_in_bb): Analyze
20044         condition operands with respect to the correct loop.  Assert
20045         the analysis doesn't fail.
20047 2018-06-04  Carl Love  <cel@us.ibm.com>
20049         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
20050         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
20051         as it is slightly cheaper.
20052         (first_match_or_eos_index_<mode>):
20053         Calculate index using natural element order.
20054         (first_match_index_<mode>):
20055         Calculate index using natural element order.
20056         (first_match_or_eos_index_<mode>):
20057         Calculate index using natural order.
20058         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
20059         for BE and LE modes.
20060         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
20061         P9V_BUILTIN_VCLZLSBB_V16QI.
20062         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
20063         specific.
20065 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20067         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
20068         indentation and line wrap for many prototypes.  Add missing
20069         @smallexample directives around block of prototypes for vec_xl and
20070         vec_xst.
20072 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
20074         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
20075         track if we pass or return IEEE 128-bit floating point.
20076         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
20077         C++ mangling that is compatible with GCC 8.1.
20078         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
20079         (init_cumulative_args): Note if we pass or return IEEE 128-bit
20080         floating point types.
20081         (rs6000_function_arg_advance_1): Likewise.
20082         (rs6000_mangle_type): Optionally generate mangled names that match
20083         what GCC 8.1 generated for IEEE 128-bit floating point types.
20084         (rs6000_globalize_decl_name): If we have an external function that
20085         passes or returns IEEE 128-bit types, generate a weak reference
20086         from the mangled name used in GCC 8.1 to the current mangled
20087         name.
20088         (rs6000_init_builtins): Make __ibm128 use the long double type if
20089         long double is IBM extended double.  Make __float128 use the long
20090         double type if long double is IEEE 128-bit.
20092         PR target/85657
20093         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
20094         macro for __ibm128 built-in functions.
20095         (PACK_IF): Add __ibm128 pack/unpack functions.
20096         (UNPACK_IF): Likewise.
20097         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
20098         enable long double built-in functions if long double is IEEE
20099         128-bit floating point.
20100         (rs6000_invalid_builtin): Update long double built-in function
20101         error message.
20102         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
20103         functions, adjust the built-in function to use the long double
20104         built-in function if __ibm128 and long double are the same type.
20105         * doc/extend.texi (PowerPC builtins): Update documention for
20106         __builtin_{,un}pack_longdouble.  Add documentation for
20107         __builtin_{,un}pack_ibm128.
20109 2018-06-06  Jim Wilson  <jimw@sifive.com>
20111         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
20112         (struct machine_function): New field interrupt_mode.
20113         (riscv_handle_type_attribute): New function.  Add forward declaration.
20114         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
20115         (riscv_expand_epilogue): Check interrupt_mode field.
20116         (riscv_set_current_function): Check interrupt attribute args and
20117         set interrupt_mode field.
20118         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
20119         (riscv_sret, riscv_uret): New.
20120         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
20121         new arguments to interrupt attribute.
20123 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
20125         PR target/63177
20126         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
20127         Don't handle -mcpu=power8 if -mpower9-vector is also used.
20129 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20131         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
20132         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
20133         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20134         several redundant entries.
20136 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
20138         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
20139         type from "rtx" to "rtx_insn *".
20140         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
20141         for local "call_insn", removing cast.
20142         (ix86_expand_call): Likewise, introducing a "call_insn" local.
20144 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
20146         PR tree-optimization/86066
20147         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
20148         for BIT_INSERT_EXPR stores.
20150 2018-06-06  Richard Biener  <rguenther@suse.de>
20152         PR tree-optimization/86062
20153         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
20154         component refs ontop
20155         of to be offsetted base.
20157 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20159         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
20160         to be static and remove check on interrupt attribute name.
20162 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20164         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20165         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
20167 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
20169         PR target/79924
20170         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
20171         second argument.
20172         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
20173         Remove second argument, change how error is called.
20174         (aarch64_layout_arg): Remove second argument from
20175         aarch64_err_no_fpadvsimd call.
20176         (aarch64_init_cumulative_args): Ditto.
20177         (aarch64_gimplify_va_arg_expr): Ditto.
20178         * config/aarch64/aarch64.md (mov<mode>): Ditto.
20180 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
20182         * config/i386/i386.md (simple_return_indirect_internal): New expander.
20183         (*simple_return_indirect_internal<mode>): Rename from
20184         simple_return_indirect_internal.  Use W mode iterator.
20185         (rstorssp): New expander.
20186         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
20187         (clrssbsy): New expander.
20188         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
20190 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20192         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
20193         __typeof__.
20194         (cmse_check_pointed_object): Likewise.
20196 2018-06-05  Martin Liska  <mliska@suse.cz>
20198         PR gcov-profile/47618
20199         * doc/invoke.texi: Document how -fprofile-dir format
20200         is extended.
20202 2018-06-05  Richard Biener  <rguenther@suse.de>
20204         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
20205         removal pretend DOM info isn't available so we do not update
20206         it and only remove edges, not dominated blocks.  Actually free
20207         DOM info in case we removed something.  Remove unreachable blocks.
20208         (mfb_keep_latches): Work with either DOM info or marked backedges.
20209         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
20210         first.  Mark backedges if DOM info isn't available.
20211         (Re-)compute DOM info after cleanup_control_flow_pre.
20213 2018-06-05  Richard Biener  <rguenther@suse.de>
20215         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
20216         (locus_discrim_hasher::hash): Adjust.
20217         (locus_discrim_hasher::equal): Likewise.
20218         (next_discriminator_for_locus): Work on line directly.
20219         (same_line_p): Pass in expanded locus1 as well.
20220         (assign_discriminators): Avoid redundant location expansions.
20222 2018-06-05  Richard Biener  <rguenther@suse.de>
20224         PR tree-optimization/86046
20225         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
20226         if required after clearing TREE_ADDRESSABLE.
20228 2018-06-05  Richard Biener  <rguenther@suse.de>
20230         PR tree-optimization/86047
20231         * tree-ssa-loop.c (for_each_index): Glob handling of all
20232         decls and constants and really handle all of them.
20234 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20236         PR target/81497
20237         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
20238         qualifier_void_pointer and qualifier_const_void_pointer.
20239         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
20240         (arm_init_builtins): Handle the above.
20241         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
20242         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
20243         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
20244         void intrinsics.
20246 2018-06-05  Martin Liska  <mliska@suse.cz>
20248         * auto-profile.c (read_autofdo_file): Do not use
20249         gcov_ctr_summary struct.
20250         (afdo_callsite_hot_enough_for_early_inline): Likewise.
20251         * coverage.c (struct counts_entry): Likewise.
20252         (read_counts_file): Read just single summary entry.
20253         (get_coverage_counts): Use gcov_summary struct.
20254         * coverage.h (get_coverage_counts): Likewise.
20255         * gcov-dump.c (dump_working_sets): Likewise.
20256         (tag_summary): Dump just single summary.
20257         * gcov-io.c (gcov_write_summary): Write just histogram
20258         summary.
20259         (gcov_read_summary): Read just single summary.
20260         (compute_working_sets): Use gcov_summary struct.
20261         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
20262         of GCOV_COUNTERS_SUMMABLE.
20263         (GCOV_COUNTERS_SUMMABLE): Remove.
20264         (GCOV_FIRST_VALUE_COUNTER): Replace with
20265         GCOV_COUNTER_V_INTERVAL.
20266         (struct gcov_ctr_summary): Remove.
20267         (struct gcov_summary): Directly use fields of former
20268         gcov_ctr_summary.
20269         (compute_working_sets): Use gcov_summary struct.
20270         * gcov.c (read_count_file): Do not use ctrs fields.
20271         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
20272         struct.
20273         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
20274         struct.
20275         * profile.c: Likewise.
20276         * profile.h: Likewise.
20278 2018-06-05  Martin Liska  <mliska@suse.cz>
20280         PR gcov-profile/84846
20281         * gcov.c (output_lines): Print working directory only
20282         in intermediate format.
20284 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
20286         * config/s390/s390-builtin-types.def: Add void function type.
20287         * config/s390/s390-builtins.def: Use the function type for the
20288         tbeginc builtin.
20290 2018-06-04  Jim Wilson  <jimw@sifive.com>
20292         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
20293         to int.
20294         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
20295         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
20296         handle EH_RETURN_DATA_REGNO registers properly.
20297         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
20298         (riscv_expand_epilogue): Update comment.  Change argument name and
20299         type.  Update code to use new name and type.  Pass new args to
20300         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
20301         EXCEPTION_RETURN.
20302         * config/riscv/riscv.md (NORMAL_RETURN): New.
20303         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
20304         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
20305         (eh_return): Call gen_eh_return_internal and emit barrier.
20306         (eh_return_internal): Call riscv_expand_epilogue.
20308 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
20310         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
20311         bit_insertion field and declare can_be_merged_into method.
20312         (merged_store_group::can_be_merged_into): New method.
20313         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
20314         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
20315         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
20317 2018-06-04  Richard Biener  <rguenther@suse.de>
20319         PR tree-optimization/85955
20320         * builtins.c (fold_builtin_sincos): Convert pointers to
20321         destination to appropriate type before dereferencing.
20323 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
20325         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
20327 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
20329         * expr.c (expand_expr_real_1): Force the operand into memory if
20330         its TYPE_MODE is BLKmode and if there is no integer mode for
20331         the number of bits being extracted.
20333 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
20335         PR target/85832
20336         PR target/86036
20337         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
20338         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
20340 2018-06-04  Richard Biener  <rguenther@suse.de>
20342         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
20343         (cleanup_tree_cfg_noloop): ... single caller.  Do
20344         start_recording_case_labels later.
20346 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
20348         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
20349         to _IMMINTRIN_H_INCLUDED.
20350         * config/i386/pconfigintrin.h: Ditto.
20351         * config/i386/waitpkgintrin.h: Ditto.
20352         * config/i386/immintrin.h: Add includes for sgxintrin.h,
20353         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
20354         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
20355         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
20356         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
20357         waitpkgintrin.h and cldemoteintrin.h.
20359 2018-06-04  Richard Biener  <rguenther@suse.de>
20361         PR tree-optimization/86038
20362         * tracer.c (find_best_successor): Check probability for
20363         being initialized, bail out if not.
20365 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
20367         PR target/86003
20368         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
20369         of bits to ignore when comparing architectures.
20371 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
20373         PR tree-optimization/69615
20374         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
20375         maximum or minimum of the type, try to merge it also as if
20376         range1 is + [-, x - 1] or + [x + 1, -].
20378         PR c++/86025
20379         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
20381 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
20383         PR tree-optimization/86034
20384         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
20385         the unsigned bitfield type in a bit insertion sequence if it does not
20386         have a larger precision than the bitfield size.
20387         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
20389 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
20391         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
20393 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
20395         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
20396         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
20397         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
20398         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
20400 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
20402         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
20403         Disable -fdelete-null-pointer-checks for ELF toolchain.
20405 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20406             Kito Cheng  <kito.cheng@gmail.com>
20408         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
20409         (nds32le-*-*, nds32be-*-*): Integrate checking process.
20410         (nds32*-*-*): Add glibc and uclibc conditions.
20411         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
20412         (TARGET_EXCEPT_UNWIND_INFO): Define.
20413         * config/nds32/elf.h: New file.
20414         * config/nds32/linux.h: New file.
20415         * config/nds32/nds32-elf.opt: New file.
20416         * config/nds32/nds32-linux.opt: New file.
20417         * config/nds32/nds32-fp-as-gp.c
20418         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
20419         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
20420         TARGET_LINUX_ABI.
20421         (nds32_asm_file_end): Ditto.
20422         (nds32_print_operand): Ditto.
20423         (nds32_insert_attributes): Ditto.
20424         (nds32_init_libfuncs): New function.
20425         (TARGET_HAVE_TLS): Define.
20426         (TARGET_INIT_LIBFUNCS): Define.
20427         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
20428         spec content.
20429         (TARGET_ELF): Apply different mcmodel setting.
20430         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
20431         been migrated into elf.h and linux.h files.
20432         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
20433         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
20434         (mcmodel): The content has been migrated into nds32-elf.opt and
20435         nds32-linux.opt files.
20436         * config/nds32/t-elf: New file.
20437         * config/nds32/t-linux: New file.
20439 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20440             Shiva Chen  <shiva0217@gmail.com>
20442         * config/nds32/constants.md (unspec_volatile_element): Add
20443         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
20444         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
20445         optimization.
20446         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
20447         (make_pass_nds32_fp_as_gp): Declare.
20448         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
20449         optmization pass.
20450         (nds32_asm_function_end_prologue): Remove unused asm output.
20451         (nds32_asm_function_begin_epilogue): Remove unused asm output.
20452         (nds32_asm_file_start): Output necessary fp_as_gp information.
20453         (nds32_option_override): Adjust register usage.
20454         (nds32_expand_prologue): Consider fp_as_gp situation.
20455         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
20456         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
20457         (epilogue): Ditto.
20458         (return): Ditto.
20459         (simple_return): Ditto.
20460         (omit_fp_begin): Output special directive for fp_as_gp.
20461         (omit_fp_end): Output special directive for fp_as_gp.
20462         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
20463         mforbid-fp-as-gp): New options.
20465 2018-06-01  Mark Wielaard  <mark@klomp.org>
20467         * dwarf2out.c (dwarf2out_finish): Remove generation of
20468         DW_AT_loclists_base.
20470 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
20472         * gimple-ssa-store-merging.c: Include gimple-fold.h.
20473         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
20474         (struct merged_store_group): Add bit_insertion field.
20475         (dump_char_array): Use standard hexadecimal format.
20476         (merged_store_group::merged_store_group): Set bit_insertion to false.
20477         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
20478         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
20479         also print the mask in the dump file.
20480         (pass_store_merging::gate): Minor tweak.
20481         (imm_store_chain_info::coalesce_immediate): Fix wrong association
20482         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
20483         stores with INTEGER_CST stores.
20484         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
20485         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
20486         and use it throughout.  Generate bit insertion sequences if need be.
20487         (pass_store_merging::process_store): Remove redundant condition.
20488         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
20490 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
20492         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
20493         the 128-bit floating point types.  Fix function comment.
20495 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20497         * config/aarch64/aarch64-simd.md
20498         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
20499         mnemonics.
20500         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
20501         mnemonics.
20503 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
20505         PR tree-optimization/85989
20506         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
20507         variable.
20508         (backprop::intersect_uses): Check it when deciding whether this
20509         is a backedge reference.
20510         (backprop::process_block): Add each phi to m_visited_phis
20511         after visiting it, then clear it at the end.
20513 2018-06-01  Richard Biener  <rguenther@suse.de>
20515         * tree-vectorizer.h (vect_dr_stmt): New function.
20516         (vect_get_load_cost): Adjust.
20517         (vect_get_store_cost): Likewise.
20518         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
20519         Use vect_dr_stmt instead of DR_SMTT.
20520         (vect_record_base_alignments): Likewise.
20521         (vect_calculate_target_alignment): Likewise.
20522         (vect_compute_data_ref_alignment): Likewise and make static.
20523         (vect_update_misalignment_for_peel): Likewise.
20524         (vect_verify_datarefs_alignment): Likewise.
20525         (vector_alignment_reachable_p): Likewise.
20526         (vect_get_data_access_cost): Likewise.  Pass down
20527         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
20528         (vect_get_peeling_costs_all_drs): Likewise.
20529         (vect_peeling_hash_get_lowest_cost): Likewise.
20530         (vect_enhance_data_refs_alignment): Likewise.
20531         (vect_find_same_alignment_drs): Likewise.
20532         (vect_analyze_data_refs_alignment): Likewise.
20533         (vect_analyze_group_access_1): Likewise.
20534         (vect_analyze_group_access): Likewise.
20535         (vect_analyze_data_ref_access): Likewise.
20536         (vect_analyze_data_ref_accesses): Likewise.
20537         (vect_vfa_segment_size): Likewise.
20538         (vect_small_gap_p): Likewise.
20539         (vectorizable_with_step_bound_p): Likewise.
20540         (vect_prune_runtime_alias_test_list): Likewise.
20541         (vect_analyze_data_refs): Likewise.
20542         (vect_supportable_dr_alignment): Likewise.
20543         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
20544         (vect_gen_prolog_loop_niters): Likewise.
20545         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
20546         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
20547         modify DR_STMT.
20548         (vect_recog_mask_conversion_pattern): Likewise.
20549         (vect_try_gather_scatter_pattern): Likewise.
20550         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
20551         to vect_get_store_cost.
20552         (vect_get_store_cost): Get stmt_info instead of DR.
20553         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
20554         (vect_get_load_cost): Get stmt_info instead of DR.
20556 2018-06-01  Richard Biener  <rguenther@suse.de>
20558         PR middle-end/86017
20559         * gimple-fold.c (var_decl_component_p): Also allow offsetted
20560         vars wrapped in MEM_REFs.
20562 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
20564         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
20565         Fix subreg tests so that we only return a choice between
20566         GENERAL_REGS and FP_REGS if the original classes included both.
20568 2018-06-01  Richard Biener  <rguenther@suse.de>
20570         PR ipa/85960
20571         * tree-ssa-structalias.c (get_function_part_constraint):
20572         Handle NULL fi->decl.
20573         (find_func_aliases_for_call): Properly handle indirect
20574         fi from direct call.
20575         (find_func_clobbers): Likewise.
20576         (ipa_pta_execute): Likewise.
20577         (create_variable_info_for): For functions that are ifunc_resolver
20578         resolve to a varinfo that contains the result of the resolver call.
20579         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
20580         aliases.
20582 2018-05-31  Michael Collison  <michael.collison@arm.com>
20584         * config/aarch64/aarch64.md:
20585         (*fix_to_zero_extenddfdi2): New pattern.
20586         * gcc.target/aarch64/fix_extend1.c: New testcase.
20588 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
20590         PR middle-end/78809
20591         PR middle-end/83026
20592         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
20593         and BUILT_IN_STRNCMP_EQ.
20594         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
20595         BUILT_IN_STRNCMP_EQ.
20596         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
20597         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
20598         (gimple_fold_builtin): Likewise.
20599         * tree-ssa-strlen.c (compute_string_length): New function.
20600         (determine_min_obsize): New function.
20601         (handle_builtin_string_cmp): New function to handle calls to
20602         string compare functions.
20603         (strlen_optimize_stmt): Add handling to builtin string compare
20604         calls.
20605         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
20606         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
20607         * tree.c (build_common_builtin_nodes): Add new defines of
20608         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
20610 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
20612         PR target/85984
20613         * bb-reorder.c (pass_partition_blocks::gate): Return false for
20614         functions with naked attribute.
20616 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
20618         * config/i386/sse.md (avx_vec_concat<mode>):
20619         Substitute concat_tg_mode mode attribute with xtg_mode.
20620         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
20621         (concat_tg_mode): Remove mode attribute.
20623 2018-05-31  Martin Sebor  <msebor@redhat.com>
20625         PR c/82063
20626         * calls.c (alloc_max_size): Correct a logic error/typo.
20627         Treat excessive arguments as infinite.  Warn for invalid arguments.
20628         * doc/invoke.texi (-Walloc-size-larger-than): Update.
20630 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
20632         PR target/85829
20633         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
20634         and movx for Haswell.
20636 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
20637             Cesar Philippidis  <cesar@codesourcery.com>
20639         PR middle-end/85879
20640         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
20641         when emitting error on private/firstprivate reductions.
20642         * omp-low.c (lower_omp_target): Avoid reference-type processing
20643         on pointers for firstprivate clause.
20645 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
20647         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
20648         (st1x2): Likewise.
20649         (st1x3): Likewise.
20650         * config/aarch64/aarch64-simd.md
20651         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
20652         (aarch64_ld1_x3_<mode>): Likewise
20653         (aarch64_st1x2<VALLDIF:mode>): Likewise
20654         (aarch64_st1_x2_<mode>): Likewise
20655         (aarch64_st1x3<VALLDIF:mode>): Likewise
20656         (aarch64_st1_x3_<mode>): Likewise
20657         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
20658         (vld1_s8_x3): Likewise.
20659         (vld1_u16_x3): Likewise.
20660         (vld1_s16_x3): Likewise.
20661         (vld1_u32_x3): Likewise.
20662         (vld1_s32_x3): Likewise.
20663         (vld1_u64_x3): Likewise.
20664         (vld1_s64_x3): Likewise.
20665         (vld1_f16_x3): Likewise.
20666         (vld1_f32_x3): Likewise.
20667         (vld1_f64_x3): Likewise.
20668         (vld1_p8_x3): Likewise.
20669         (vld1_p16_x3): Likewise.
20670         (vld1_p64_x3): Likewise.
20671         (vld1q_u8_x3): Likewise.
20672         (vld1q_s8_x3): Likewise.
20673         (vld1q_u16_x3): Likewise.
20674         (vld1q_s16_x3): Likewise.
20675         (vld1q_u32_x3): Likewise.
20676         (vld1q_s32_x3): Likewise.
20677         (vld1q_u64_x3): Likewise.
20678         (vld1q_s64_x3): Likewise.
20679         (vld1q_f16_x3): Likewise.
20680         (vld1q_f32_x3): Likewise.
20681         (vld1q_f64_x3): Likewise.
20682         (vld1q_p8_x3): Likewise.
20683         (vld1q_p16_x3): Likewise.
20684         (vld1q_p64_x3): Likewise.
20685         (vst1_s64_x2): Likewise.
20686         (vst1_u64_x2): Likewise.
20687         (vst1_f64_x2): Likewise.
20688         (vst1_s8_x2): Likewise.
20689         (vst1_p8_x2): Likewise.
20690         (vst1_s16_x2): Likewise.
20691         (vst1_p16_x2): Likewise.
20692         (vst1_s32_x2): Likewise.
20693         (vst1_u8_x2): Likewise.
20694         (vst1_u16_x2): Likewise.
20695         (vst1_u32_x2): Likewise.
20696         (vst1_f16_x2): Likewise.
20697         (vst1_f32_x2): Likewise.
20698         (vst1_p64_x2): Likewise.
20699         (vst1q_s8_x2): Likewise.
20700         (vst1q_p8_x2): Likewise.
20701         (vst1q_s16_x2): Likewise.
20702         (vst1q_p16_x2): Likewise.
20703         (vst1q_s32_x2): Likewise.
20704         (vst1q_s64_x2): Likewise.
20705         (vst1q_u8_x2): Likewise.
20706         (vst1q_u16_x2): Likewise.
20707         (vst1q_u32_x2): Likewise.
20708         (vst1q_u64_x2): Likewise.
20709         (vst1q_f16_x2): Likewise.
20710         (vst1q_f32_x2): Likewise.
20711         (vst1q_f64_x2): Likewise.
20712         (vst1q_p64_x2): Likewise.
20713         (vst1_s64_x3): Likewise.
20714         (vst1_u64_x3): Likewise.
20715         (vst1_f64_x3): Likewise.
20716         (vst1_s8_x3): Likewise.
20717         (vst1_p8_x3): Likewise.
20718         (vst1_s16_x3): Likewise.
20719         (vst1_p16_x3): Likewise.
20720         (vst1_s32_x3): Likewise.
20721         (vst1_u8_x3): Likewise.
20722         (vst1_u16_x3): Likewise.
20723         (vst1_u32_x3): Likewise.
20724         (vst1_f16_x3): Likewise.
20725         (vst1_f32_x3): Likewise.
20726         (vst1_p64_x3): Likewise.
20727         (vst1q_s8_x3): Likewise.
20728         (vst1q_p8_x3): Likewise.
20729         (vst1q_s16_x3): Likewise.
20730         (vst1q_p16_x3): Likewise.
20731         (vst1q_s32_x3): Likewise.
20732         (vst1q_s64_x3): Likewise.
20733         (vst1q_u8_x3): Likewise.
20734         (vst1q_u16_x3): Likewise.
20735         (vst1q_u32_x3): Likewise.
20736         (vst1q_u64_x3): Likewise.
20737         (vst1q_f16_x3): Likewise.
20738         (vst1q_f32_x3): Likewise.
20739         (vst1q_f64_x3): Likewise.
20740         (vst1q_p64_x3): Likewise.
20742 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20744         * config/msp430/msp430.c (msp430_output_labelref): Prepend
20745         user_label_prefix to name.
20747         * tree-core.h: Update comment about the format of NAME string
20748         passed to handler in attribute_spec.
20750         * config/msp430/msp430.md: Remove erroneous subreg expression from
20751         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
20752         zero_extend{q,h}isi2.
20754 2018-05-30  Borislav Petkov  <bp@suse.de>
20756         * doc/extend.texi: Document some architecture specific
20757         constraints and sort entries.
20759 2018-05-30  Martin Sebor  <msebor@redhat.com>
20761         PR middle-end/85369
20762         * builtins.c (expand_builtin_stpcpy_1): New function.
20763         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
20764         only if the former succeeds.
20766 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
20768         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
20769         in saphira.
20771 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20773         * doc/invoke.texi (-flinker-output): Document
20775 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20777         * passes.c (ipa_write_summaries): Only modify statements if body
20778         is in memory.
20779         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
20780         incrementally linking.
20781         (ipa_passes): Likewise.
20782         * lto-cgraph.c (lto_output_node): When incrementally linking do not
20783         pass down resolution info.
20784         * common.opt (flag_incremental_link): Update info.
20785         * gcc.c (plugin specs): Turn flinker-output=* to
20786         -plugin-opt=-linker-output-known
20787         * toplev.c (compile_file): Also cut compilation when doing incremental
20788         link.
20789         * flag-types. (enum lto_partition_model): Add
20790         LTO_LINKER_OUTPUT_NOLTOREL.
20791         (invoke.texi): Add -flinker-output docs.
20792         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
20793         link same way as WPA; do not stream in dead initializers.
20795         * dwarf2out.c (dwarf2out_die_ref_for_decl,
20796         darf2out_register_external_decl): Support incremental link.
20798 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20800         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
20802 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20804         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
20805         it down to simple_object_copy_lto_debug_sections.
20806         (run_gcc): Determine incremental LTO link time and configure
20807         lto1 into non-wpa mode, disable renaming of debug sections.
20809 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20811         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20812         descriptions of various incorrectly documented functions.
20814 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20816         Revert:
20817         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
20818         address check not strict.
20820 2018-05-30  Richard Biener  <rguenther@suse.de>
20822         PR tree-optimization/85964
20823         * tracer.c (better_p): Drop initialized count check, we only
20824         call the function with initialized counts now.
20825         (find_best_successor): Do find a best edge if one
20826         has uninitialized count.
20827         (find_best_predecessor): Likewise.  Do BB frequency check only
20828         if count is initialized.
20830 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
20832         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
20833         (aarch64_ldrstr_offset_compare): New.
20834         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
20835         load/store orderings.
20836         (aarch64_gen_adjusted_ldpstp): Likewise.
20838 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
20840         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
20841         Check for subset of GENERAL_REGS and FP_REGS.
20842         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
20843         r=w alternative.
20845 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
20847         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
20848         and wi::to_poly_offset.  Add the current offset and then check
20849         whether the sum fits, rather than using an unchecked addition of
20850         a checked term.  Check for a shwi rather than a uhwi.
20851         * expr.c (get_bit_range): Use tree_to_poly_uint64.
20852         (store_constructor): Use poly_int_tree_p.
20853         (expand_expr_real_1): Likewise.
20854         * function.c (assign_temp): Likewise.
20855         * fold-const.c (const_binop): Use poly_int_tree_p and
20856         wi::to_poly_offset.
20857         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
20858         division.
20859         * ipa-icf-gimple.c (func_checker::compare_operand): Use
20860         to_poly_offset for MEM offsets.
20861         * ipa-icf.c (sem_variable::equals): Likewise.
20862         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
20863         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20864         wi::to_poly_offset for BIT_FIELD_REF offsets.
20865         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
20866         wi::to_poly_offset.
20867         * var-tracking.c (emit_note_insn_var_location): Use
20868         tree_to_poly_uint64.
20870 2018-05-29  Jim Wilson  <jimw@sifive.com>
20872         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
20874 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
20876         PR target/85950
20877         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
20878         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
20879         sequence.
20880         (sse4_1_round<mode>2): Use nonimmediate_operand
20881         for operand 1 predicate.
20883 2018-05-29  Martin Sebor  <msebor@redhat.com>
20884             Richard Biener  <rguenther@suse.de>
20886         PR testsuite/85888
20887         * calls.c (get_size_range): Call determine_value_range instead
20888         of get_value_range..
20889         * tree-vrp.h (determine_value_range): Declared new function.
20890         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
20892 2018-05-29  Richard Biener  <rguenther@suse.de>
20894         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
20895         sure to use non-pattern stmts for get_earlier_stmt arguments.
20896         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
20897         called on pattern stmts.
20898         (get_later_stmt): Likewise.
20900 2018-05-29  Martin Liska  <mliska@suse.cz>
20902         PR gcov-profile/85759
20903         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
20904         env variables.
20906 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
20908         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
20909         VEC_UNPACK_*_EXPR.
20910         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
20911         VEC_PACK_*_EXPR.
20913         PR target/85918
20914         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20915         VEC_PACK_FLOAT_EXPR): New tree codes.
20916         * tree-pretty-print.c (op_code_prio): Handle
20917         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20918         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20919         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20920         * tree-inline.c (estimate_operator_cost): Likewise.
20921         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
20922         * fold-const.c (const_binop): Likewise.
20923         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
20924         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20925         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
20926         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
20927         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20928         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20929         * expr.c (expand_expr_real_2): Likewise.
20930         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
20931         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
20932         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
20933         optabs.
20934         * optabs.c (expand_widen_pattern_expr): For
20935         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
20936         sign from result type rather than operand's type.
20937         (expand_binop_directly): For vec_packu_float_optab and
20938         vec_packs_float_optab allow result type to be different from operand's
20939         type.
20940         * optabs-tree.c (optab_for_tree_code): Handle
20941         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20942         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
20943         * tree-vect-generic.c (expand_vector_operations_1):  Handle
20944         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20945         VEC_PACK_FLOAT_EXPR.
20946         * tree-vect-stmts.c (supportable_widening_operation): Handle
20947         FIX_TRUNC_EXPR.
20948         (supportable_narrowing_operation): Handle FLOAT_EXPR.
20949         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
20950         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
20951         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
20952         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
20953         mode attributes.
20954         (vec_pack<floatprefix>_float_<mode>): New expander.
20955         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
20956         attributes.
20957         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
20958         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
20959         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
20960         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
20961         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
20962         Document.
20963         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
20964         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
20965         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20966         VEC_PACK_FLOAT_EXPR): Document.
20968 2018-05-29  Richard Biener  <rguenther@suse.de>
20970         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
20971         member.
20972         (stmt_vec_info_vec): Make pointer.
20973         (init_stmt_vec_info_vec): Remove.
20974         (free_stmt_vec_info_vec): Likewise.
20975         (set_stmt_vec_info_vec): New function.
20976         (free_stmt_vec_infos): Likewise.
20977         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
20978         (set_vinfo_for_stmt): Likewise.
20979         (get_earlier_stmt): Likewise.
20980         (get_later_stmt): Likewise.
20981         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
20982         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
20983         (vec_info::~vec_info): Free stmt_vec_infos.
20984         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
20985         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
20986         (pass_slp_vectorize::execute): Likewise.
20987         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
20988         (free_stmt_vec_info_vec): Likewise.
20989         (set_stmt_vec_info_vec): New function.
20990         (free_stmt_vec_infos): Likewise.
20991         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
20992         the global stmt_vec_info_vec.
20993         * tree-parloops.c (gather_scalar_reductions): Use
20994         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
20995         vector.
20997 2018-05-29  Richard Biener  <rguenther@suse.de>
20999         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
21001 2018-05-29  Martin Liska  <mliska@suse.cz>
21002             David Malcolm  <dmalcolm@redhat.com>
21004         * vec.c (test_reverse): New.
21005         (vec_c_tests): Add new test.
21006         * vec.h (vl_ptr>::reverse): New function.
21008 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
21010         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
21012         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
21013         and later.
21015 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
21017         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
21019 2018-05-28  Richard Biener  <rguenther@suse.de>
21021         PR tree-optimization/85933
21022         * tree-vect-data-refs.c (vect_record_base_alignments): Only
21023         look at stmts marked as vectorizable.
21025 2018-05-28  Richard Biener  <rguenther@suse.de>
21027         PR tree-optimization/85934
21028         * tree-vect-generic.c (expand_vector_operations_1): Hoist
21029         vector boolean check before scalar optimization.
21031 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
21033         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
21034         for armv5te.
21036 2018-05-28  Mark Wielaard  <mark@klomp.org>
21038         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
21039         if it is an expression containing a minus sign.
21041 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
21043         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
21045 2018-05-27  Paul Koning  <ni1d@arrl.net>
21047         * config/pdp11/pdp11.md (truncsihi2): Remove.
21049 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
21050             Chung-Ju Wu  <jasonwucj@gmail.com>
21052         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
21053         implementation.
21054         (unaligned_store_dw): Ditto.
21055         * config/nds32/nds32-memory-manipulation.c
21056         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
21057         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
21058         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
21059         (emit_setmem_word_loop): Rename to ...
21060         (emit_setmem_doubleword_loop): ... this.
21061         (nds32_gen_dup_4_byte_to_word_value): New function.
21062         (nds32_gen_dup_8_byte_to_double_word_value): New function.
21063         (nds32_expand_setmem_loop): Refine implementation.
21064         (nds32_expand_setmem_loop_v3m): Ditto.
21065         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
21066         pattern.
21068 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21070         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
21072 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21074         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
21075         (nds32_init_machine_status): Initialize machine->attr_naked_p and
21076         machine->attr_no_prologue_p.
21077         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
21078         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
21079         (nds32_expand_epilogue): Consider attr_naked_p.
21080         (nds32_expand_epilogue_v3pop): Likewise.
21081         (nds32_can_use_return_insn): Likewise.
21082         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
21083         attr_no_prologue_p fields.
21084         * config/nds32/nds32.opt (mret-in-naked-func): New option.
21086 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
21088         PR target/85918
21089         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
21090         attributes.
21091         * config/i386/sse.md
21092         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
21093         Rename to ...
21094         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
21095         ... this.
21096         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
21097         Rename to ...
21098         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
21099         ... this.
21100         (*<floatsuffix>floatv2div2sf2): Rename to ...
21101         (*float<floatunssuffix>v2div2sf2): ... this.
21102         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
21103         (float<floatunssuffix>v2div2sf2_mask): ... this.
21104         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
21105         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
21106         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
21107         to ...
21108         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
21109         ... this.
21110         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21111         Rename to ...
21112         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21113         ... this.
21114         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21115         Rename to ...
21116         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21117         ... this.
21118         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
21119         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
21120         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
21121         gen_ufix_truncv8dfv8si2.
21122         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
21123         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
21124         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
21125         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
21126         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
21127         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
21128         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
21129         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
21131 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
21133         PR target/85900
21134         PR target/85345
21135         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
21137 2018-05-25  Jim Wilson  <jimw@sifive.com>
21139         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
21140         * config/riscv/riscv.c (struct machine_function): Add
21141         interrupt_handler_p and attribute_checked_p fields.
21142         (riscv_attribute_table): Add interrupt.
21143         (riscv_interrupt_type_p): New.
21144         (riscv_save_reg_p): Save extra regs for interrupt handler.
21145         (riscv_use_save_libcall): Return false  for interrupt handler.
21146         (riscv_first_stack_step): Add forward declaration.
21147         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
21148         for interrupt handler with large frame.  Use it for saved reg list.
21149         (riscv_expand_prologue): Move flag_stack_usage_info support to
21150         eliminate duplication.
21151         (riscv_expand_epilogue): Generate mret for interrupt handler.
21152         (riscv_epilogue_uses): New.
21153         (riscv_can_use_return_insn): Return false for interrupt handler.
21154         (riscv_function_ok_for_sibcall): Likewise.
21155         (riscv_set_current_function): Add interrupt handler support.
21156         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
21157         * config/riscv/riscv.md (UNSPECV_MRET): New.
21158         (GP_REGNUM): New.
21159         (riscv_frflags, riscv_fsflags): Use tab after opcode.
21160         (riscv_mret): New.
21161         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
21163 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
21165         PR tree-optimization/85712
21166         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
21167         this candidate has already been replaced in-situ by a copy.
21169 2018-05-25  Jason Merrill  <jason@redhat.com>
21171         PR c++/80485 - inline function non-zero address.
21172         * symtab.c (nonzero_address): Check DECL_COMDAT.
21174 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
21176         PR target/83628
21177         * config/alpha/alpha.md (ashlsi3): New insn pattern.
21178         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
21179         extension of SImode operation.  Use const123_operand predicate.
21180         (*saddsi_1): Remove.
21181         (*saddl_se_1): Ditto.
21182         (*ssubsi_1): Ditto.
21183         (*ssubl_se_1): Ditto.
21184         * config/alpha/predicates.md (const123_operand): New predicate.
21185         * config/alpha/constraints.md (P): Use IN_RANGE.
21187 2018-05-25  Richard Biener  <rguenther@suse.de>
21189         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
21190         defaulted to true.
21191         (ref_maybe_used_by_stmt_p): Likewise.
21192         (stmt_may_clobber_ref_p): Likewise.
21193         (stmt_may_clobber_ref_p_1): Likewise.
21194         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
21195         and pass it along.
21196         (ref_maybe_used_by_stmt_p): Likewise.
21197         (stmt_may_clobber_ref_p): Likewise.
21198         (stmt_may_clobber_ref_p_1): Likewise.
21199         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
21200         the alias oracle to disambiguate DRs with stmts DR analysis
21201         couldn't handle.
21202         (vect_analyze_data_refs): Do not give up on not analyzable
21203         DRs for BB vectorization.  Remove code truncating the dataref
21204         vector.
21206 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
21208         PR target/85832
21209         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
21210         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
21211         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
21213 2018-05-25  Richard Biener  <rguenther@suse.de>
21215         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
21216         function, combining stmt data ref gathering and fatal analysis
21217         parts.
21218         (vect_analyze_data_refs): Remove now redudnant code and simplify.
21219         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
21220         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
21221         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
21222         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
21224 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
21226         PR tree-optimization/85720
21227         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
21228         SCC if all partitions are builtins.
21229         (version_loop_by_alias_check): New parameter.  Generate cancelable
21230         runtime alias check if all partitions are builtins.
21231         (distribute_loop): Update call to above function.
21233 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
21235         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
21236         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
21237         (parm_default_def_partition_arg): Ditto.
21238         (set_parm_default_def_partition): Ditto.
21239         (get_parm_default_def_partitions): Ditto and make it static.
21240         (get_undefined_value_partitions): Ditto and make it static.
21241         (remove_ssa_form): Refactor call to init_var_map here.
21242         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
21243         computation for loop region.
21244         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
21245         (register_default_def): Delete.
21246         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
21247         (parm_default_def_partition_arg): Ditto.
21248         (set_parm_default_def_partition): Ditto.
21249         (get_parm_default_def_partitions): Ditto and make it static.
21250         (get_undefined_value_partitions): Ditto and make it static.
21251         (coalesce_with_default, coalesce_with_default): Update comment.
21252         (create_coalesce_list_for_region): New func factored out from
21253         create_outofssa_var_map.
21254         (populate_coalesce_list_for_outofssa): New func factored out from
21255         create_outofssa_var_map and coalesce_ssa_name.
21256         (create_outofssa_var_map): Delete.
21257         (coalesce_ssa_name): Refactor to support live range computation.
21258         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
21259         (get_parm_default_def_partitions): Delete.
21260         (get_undefined_value_partitions): Ditto.
21261         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
21262         computation for loop region.
21263         (new_tree_live_info, loe_visit_block): Ditto.
21264         (live_worklist, set_var_live_on_entry): Ditto.
21265         (calculate_live_on_exit, verify_live_on_entry): Ditto.
21266         * tree-ssa-live.h (struct _var_map): New fields.
21267         (init_var_map): Change decl.
21268         (region_contains_p): New.
21270 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
21272         * tree-ssa-live.h (live_merge_and_clear): Delete.
21274 2018-05-25  Richard Biener  <rguenther@suse.de>
21276         PR c++/85912
21277         * tree-dump.c (dequeue_and_dump): Remove access to removed
21278         operand 2 of a SWITCH_EXPR.
21280 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21282         * doc/sourcebuild.texi (vect_double_cond_arith): Include
21283         multiplication and division.
21284         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
21285         (cond_udiv@var{m}, cond_umod@var{m}): Document.
21286         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
21287         (cond_udiv_optab, cond_umod_optab): New optabs.
21288         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
21289         (IFN_COND_RDIV): New internal functions.
21290         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
21291         TRUNC_MOD_EXPR and RDIV_EXPR.
21292         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
21293         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
21294         New unspecs.
21295         (SVE_INT_BINARY): Include mult.
21296         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
21297         (optab, sve_int_op): Handle mult.
21298         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
21299         UNSPEC_COND_DIV.
21300         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
21301         for SVE_INT_BINARY_SD.
21303 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21305         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
21306         (optab, sve_int_op): Handle div and udiv.
21307         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
21308         for SVE_INT_BINARY_SD.
21309         (*<optab><mode>3): New insn for the same.
21311 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21313         * tree-vect-patterns.c: Include predict.h.
21314         (vect_recog_divmod_pattern): Restrict check for division support
21315         to when optimizing for size.
21317 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21319         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
21320         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
21321         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
21322         (gimple_match_op::set_op): Likewise.
21323         (gimple_resimplify4): Declare.
21324         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
21325         (expr::gen_transform): Likewise.
21326         (decision_tree::gen): Generate a simplification routine for 4 operands.
21327         * gimple-match-head.c (gimple_simplify): Add an overload for
21328         4 operands.  In the top-level function, handle up to 4 call
21329         arguments and call gimple_resimplify4.
21330         (gimple_resimplify4): New function.
21331         (build_call_internal): Pass a fourth operand.
21332         (maybe_push_to_seq): Likewise.
21333         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
21334         Fold VEC_COND_EXPRs of an operation and a default value into
21335         an IFN_COND_* function if possible.
21336         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
21337         New unspecs.
21338         (SVE_COND_FP_BINARY): Include them.
21339         (optab, sve_fp_op): Handle them.
21340         (SVE_INT_BINARY_REV): New code iterator.
21341         (SVE_COND_FP_BINARY_REV): New int iterator.
21342         (commutative): New int attribute.
21343         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
21344         Declare.
21345         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
21346         function.
21347         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
21348         (*cond_<optab><mode>): New patterns for reversed operands.
21350 2018-05-25  Richard Biener  <rguenther@suse.de>
21352         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
21353         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
21354         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
21355         (STMT_VINFO_GROUPED_ACCESS): Adjust.
21356         * tree-vect-data-refs.c (everywhere): Adjust users.
21357         * tree-vect-loop.c (everywhere): Likewise.
21358         * tree-vect-slp.c (everywhere): Likewise.
21359         * tree-vect-stmts.c (everywhere): Likewise.
21360         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
21362 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21364         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
21365         Rename to...
21366         (gcc_cv_as_section_exclude): ... this.
21367         Try Solaris as #exclude syntax.
21368         * configure: Regenerate.
21369         * config.in: Regenerate.
21370         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
21371         SECTION_EXCLUDE.
21372         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
21373         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
21375         * varasm.c (default_elf_asm_named_section): Don't check if
21376         HAVE_GAS_SECTION_EXCLUDE is defined.
21378 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21380         * doc/md.texi: Update the documentation of the cond_* optabs
21381         to mention the new final operand.  Fix GET_MODE_NUNITS call.
21382         Describe the scalar case too.
21383         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
21384         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
21385         instead of 2.
21386         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
21387         (get_conditional_internal_fn): Update comment.
21388         * tree-vect-loop.c (vectorizable_reduction): Pass the original
21389         accumulator value as a final argument to conditional functions.
21390         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
21391         a define_expand and add an "else" operand.  Assert for now that
21392         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
21393         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
21394         (*cond_<optab><mode>): New patterns.
21395         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
21396         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21397         (UNSPEC_COND_EOR): Delete.
21398         (optab): Remove associated mappings.
21399         (SVE_INT_BINARY): New code iterator.
21400         (sve_int_op): Remove int attribute and add "minus" to the code
21401         attribute.
21402         (SVE_COND_INT_OP): Delete.
21403         (SVE_COND_FP_OP): Rename to...
21404         (SVE_COND_FP_BINARY): ...this.
21406 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
21408         * optabs.c (can_reuse_operands_p): New function.
21409         (maybe_legitimize_operands): Try to reuse the results for
21410         earlier operands.
21412 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
21414         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
21415         Add {q} suffix to insn mnemonic.
21417 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
21419         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
21420         (msp430_warn_func_return): New.
21422 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
21424         * fold-const.c (tree_nonzero_bits): New function.
21425         * fold-const.h (tree_nonzero_bits): Likewise.
21426         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
21427         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
21429 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
21431         PR target/85900
21432         PR target/85345
21433         * varasm.c (assemble_alias): Check ifunc_resolver only on
21434         FUNCTION_DECL.
21436 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
21438         PR target/85903
21439         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
21440         when memory input operand is handled.
21442 2018-05-24  Luis Machado  <luis.machado@linaro.org>
21444         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
21445         global.
21446         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
21448 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
21450         * match.pd: Delay FMA folds until after vectorization.
21452 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
21454         PR target/83009
21455         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21456         address check not strict.
21458 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
21460         * gimple-match.h (gimple_match_op): New class.
21461         (mprts_hook): Replace parameters with a gimple_match_op *.
21462         (maybe_build_generic_op): Likewise.
21463         (gimple_simplified_result_is_gimple_val): Replace parameters with
21464         a const gimple_match_op *.
21465         (gimple_simplify): Replace code_helper * and tree * parameters with
21466         a gimple_match_op * parameter.
21467         (gimple_resimplify1): Replace code_helper *, tree and tree *
21468         parameters with a gimple_match_op * parameter.
21469         (gimple_resimplify2): Likewise.
21470         (gimple_resimplify3): Likewise.
21471         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
21472         parameters with a gimple_match_op * parameter.
21473         * gimple-match-head.c (gimple_simplify): Change prototypes of
21474         auto-generated functions to take a gimple_match_op * instead of
21475         separate code_helper * and tree * parameters.  Make the same
21476         change in the top-level overload and update calls to the
21477         gimple_resimplify routines.  Update calls to the auto-generated
21478         functions and to maybe_push_res_to_seq in the publicly-facing
21479         operation-specific gimple_simplify overloads.
21480         (gimple_match_op::MAX_NUM_OPS): Define.
21481         (gimple_resimplify1): Replace rcode and ops with a single res_op
21482         parameter.  Update call to gimple_simplify.
21483         (gimple_resimplify2): Likewise.
21484         (gimple_resimplify3): Likewise.
21485         (mprts_hook): Replace parameters with a gimple_match_op *.
21486         (maybe_build_generic_op): Likewise.
21487         (build_call_internal): Replace type, nargs and ops with
21488         a gimple_match_op *.
21489         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
21490         with a single gimple_match_op *.  Update calls to mprts_hook,
21491         build_call_internal and gimple_simplified_result_is_gimple_val.
21492         Factor out code that is common to the tree_code and combined_fn cases.
21493         * genmatch.c (expr::gen_transform): Replace tem_code and
21494         tem_ops with a gimple_match_op called tem_op.  Update calls
21495         to the gimple_resimplify functions and maybe_push_res_to_seq.
21496         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
21497         res_ops.  Update call to the gimple_resimplify functions.
21498         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
21499         (decision_tree::gen): Make the functions take a gimple_match_op *
21500         called res_op instead of separate res_code and res_ops parameters.
21501         Update call accordingly.
21502         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
21503         and ops with a single res_op parameter.  Update calls to
21504         maybe_build_generic_op and maybe_push_res_to_seq.
21505         (fold_stmt_1): Update calls to gimple_simplify and
21506         replace_stmt_with_simplification.
21507         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
21508         and gimple_simplified_result_is_gimple_val.
21509         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
21510         gimple_simplify.
21511         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
21512         with a gimple_match_op *.
21513         (vn_nary_build_or_lookup): Likewise.  Update call to
21514         vn_nary_build_or_lookup_1.
21515         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
21516         gimple_match_op *.  Update calls to the gimple_resimplify routines
21517         and to gimple_simplified_result_is_gimple_val.
21518         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
21519         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
21520         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
21521         (visit_nary_op): Likewise.
21522         (visit_reference_op_load): Likewise.
21524 2018-05-23  Luis Machado  <luis.machado@linaro.org>
21526         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
21527         modifier for printing the step amount.
21529 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
21531         PR target/78849
21532         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
21533         types.
21535 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
21537         * doc/sourcebuild.texi (Endianness): New subsubsection.
21539 2018-05-23  Luis Machado  <luis.machado@linaro.org>
21541         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21542         <prefetch_dynamic_strides>: New const bool field.
21543         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21544         prefetch_dynamic_strides.
21545         (exynosm1_prefetch_tune): Likewise.
21546         (thunderxt88_prefetch_tune): Likewise.
21547         (thunderx_prefetch_tune): Likewise.
21548         (thunderx2t99_prefetch_tune): Likewise.
21549         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
21550         false.
21551         (aarch64_override_options_internal): Update to set
21552         PARAM_PREFETCH_DYNAMIC_STRIDES.
21553         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21554         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21555         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21556         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21557         prefetch-dynamic-strides setting.
21559 2018-05-23  Luis Machado  <luis.machado@linaro.org>
21561         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21562         <minimum_stride>: New const int field.
21563         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21564         minimum_stride field defaulting to -1.
21565         (exynosm1_prefetch_tune): Likewise.
21566         (thunderxt88_prefetch_tune): Likewise.
21567         (thunderx_prefetch_tune): Likewise.
21568         (thunderx2t99_prefetch_tune): Likewise.
21569         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
21570         <default_opt_level>: Set to 3.
21571         (aarch64_override_options_internal): Update to set
21572         PARAM_PREFETCH_MINIMUM_STRIDE.
21573         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21574         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21575         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21576         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21577         stride is constant and is below the minimum stride threshold.
21579 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21581         * config/arm/arm-cpus.in (mode26): Delete.
21582         (armv4): Delete mode26 reference.
21583         * config/arm/arm.c (arm_configure_build_target): Delete use of
21584         isa_bit_mode26.
21586 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
21588         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
21589         New insn pattern.
21590         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
21591         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
21592         for non-SSE modes.
21593         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
21594         (floatunsdidf2): Ditto.
21596 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
21598         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
21599         (fixuns_trunc<mode>si2_avx512f): Ditto.
21600         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
21601         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
21602         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
21604 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
21606         PR rtl-optimization/79985
21607         * df-scan.c (df_insn_refs_collect): Remove special case for
21608         global registers and asm statements.
21610 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
21612         * extend.texi (Global Register Variables): Rewrite the bullet list.
21613         Note that the register is available for allocation. Note that access
21614         via inline asm must use constraints. Add note about async-signal
21615         handlers. Remove paragraph about automagic register selection.
21617 2018-05-23  Richard Biener  <rguenther@suse.de>
21619         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
21620         of fixed offset from memset VN.
21622 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
21624         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
21625         first_interp field.
21626         (alloc_cand_and_find_basis): Initialize first_interp field.
21627         (slsr_process_mul): Modify first_interp field.
21628         (slsr_process_add): Likewise.
21629         (slsr_process_cast): Modify first_interp field for each new
21630         interpretation.
21631         (slsr_process_copy): Likewise.
21632         (dump_candidate): Dump first_interp field.
21633         (replace_mult_candidate): Process all interpretations, not just
21634         subsequent ones.
21635         (replace_rhs_if_not_dup): Likewise.
21636         (replace_one_candidate): Likewise.
21638 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
21640         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
21641         Add new boolean.
21642         (aarch64_needs_frame_chain): New function.
21643         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
21645 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
21647         PR target/84882
21648         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
21649         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
21650         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
21651         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
21652         as true for strict-align.
21653         (aarch64_can_inline_p): Perform checks even when callee has no
21654         attributes to check for strict alignment.
21655         * doc/extend.texi (AArch64 Function Attributes): Document
21656         no-strict-align.
21657         * doc/invoke.texi: (AArch64 Options): Likewise.
21659 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
21661         PR tree-optimization/85853
21662         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
21663         the handling of the root of the node to...
21664         (vect_slp_analyze_node_operations_1): ...this new function,
21665         and run the whole thing with the child nodes' def types
21666         set according to their SLP node's def type.
21668 2018-05-23  Richard Biener  <rguenther@suse.de>
21670         PR middle-end/85874
21671         * tree-data-ref.c (create_runtime_alias_checks): Defer
21672         and ignore overflow warnings.
21674 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
21676         PR tree-optimization/85822
21677         * tree-vrp.c (is_masked_range_test): Fix handling of negative
21678         constants.
21680 2018-05-23  Richard Biener  <rguenther@suse.de>
21682         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
21683         memset constants via native_interpret_expr.
21685 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
21687         PR target/85345
21688         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
21689         attribute.
21690         (cgraph_node::create_alias): Likewise.
21691         (cgraph_node::get_availability): Check ifunc_resolver instead
21692         of looking up ifunc attribute.
21693         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
21694         * varasm.c (do_assemble_alias): Likewise.
21695         (assemble_alias): Likewise.
21696         (default_binds_local_p_3): Likewise.
21697         * cgraph.h (cgraph_node): Add ifunc_resolver.
21698         (cgraph_node::only_called_directly_or_aliased_p): Return false
21699         for IFUNC resolver.
21700         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
21701         attribute.
21702         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
21703         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
21704         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
21705         instead of looking up ifunc attribute.
21707 2018-05-22  Luis Machado  <luis.machado@linaro.org>
21709         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
21711 2018-05-22  Martin Sebor  <msebor@redhat.com>
21713         PR middle-end/85359
21714         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
21715         only when expasion succeeds.
21716         (expand_builtin_strcmp): Same.
21717         (expand_builtin_strncmp): Same.
21719 2018-05-22  Martin Sebor  <msebor@redhat.com>
21721         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
21723 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21724             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21726         * config/aarch64/aarch64-ldpstp.md: Replace uses of
21727         aarch64_mem_pair_operand with memory_operand and delete operand swapping
21728         code.
21729         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21730         Add check for legitimate_address.
21731         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
21732         (aarch64_swap_ldrstr_operands): New.
21733         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
21734         Define prototype.
21736 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21737             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21739         * config/aarch64/aarch64.md: New patterns to generate stp
21740         and ldp.
21741         (store_pair_sw, store_pair_dw): New patterns to generate stp for
21742         single words and double words.
21743         (load_pair_sw, load_pair_dw): Likewise.
21744         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
21745         Delete.
21746         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
21747         Delete.
21748         * config/aarch64/aarch64-ldpstp.md: Modify peephole
21749         for different mode ldpstp and add peephole for merged zero stores.
21750         Likewise for loads.
21751         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21752         Add size check.
21753         (aarch64_gen_store_pair): Rename calls to match new patterns.
21754         (aarch64_gen_load_pair): Rename calls to match new patterns.
21755         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
21756         (load_pair<DREG:mode><DREG2:mode>): ... This.
21757         (store_pair<mode>): Rename to...
21758         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
21759         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
21760         New mode iterators.
21761         (V_INT_EQUIV): Handle SImode.
21762         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
21763         New predicate.
21765 2018-05-22  Martin Sebor  <msebor@redhat.com>
21767         PR c/85623
21768         * calls.c (maybe_warn_nonstring_arg): Use string length to set
21769         or ajust the presumed bound on an operation to avoid unnecessary
21770         warnings.
21772 2018-05-22  Martin Sebor  <msebor@redhat.com>
21774         PR tree-optimization/85826
21775         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
21776         assuming that a DECL necesarily has a constant size.
21778 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
21780         PR middle-end/85862
21781         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
21783 2018-05-22  Richard Biener  <rguenther@suse.de>
21785         PR tree-optimization/85834
21786         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
21787         non-constant and non-zero memset arguments.
21789 2018-05-22  Martin Liska  <mliska@suse.cz>
21791         PR ipa/85607
21792         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
21794 2018-05-22  Richard Biener  <rguenther@suse.de>
21796         PR tree-optimization/85863
21797         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
21798         comparisons when vectype is specified.
21799         (vectorizable_condition): Do not specify vectype for
21800         vect_is_simple_cond when SLP vectorizing.
21802 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
21804         PR target/85657
21805         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
21806         define __ibm128 as long double.
21807         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
21808         as a distinct type when IEEE 128-bit support is enabled.
21809         (init_float128_ieee): Fix up conversions between IFmode and IEEE
21810         128-bit types to use the correct functions.
21811         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
21812         convert between 128-bit floating point types that have different
21813         modes but the same representation, instead of using gen_lowpart to
21814         makean alias.
21815         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
21816         KFmode.
21817         (IFKF_reg): New attributes to give the register constraints for
21818         IFmode and KFmode.
21819         (extend<mode>tf2_internal): New insns to mark an explicit
21820         conversion between 128-bit floating point types that have a
21821         different mode but share the same representation.
21823 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
21825         PR tree-optimization/85814
21826         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
21827         a null return from get_strinfo when unsharing the next
21828         strinfo in the chain.
21830 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21832         PR gcc/84923
21833         * varasm.c (weak_finish): Clean up weak_decls.
21835 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21837         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
21838         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
21839         UNSPEC_UADALP values.
21840         * config/aarch64/iterators.md (ABAL): New int iterator.
21841         (ABDL2): Likewise.
21842         (ADALP): Likewise.
21843         (sur): Add mappings for the above.
21844         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
21845         New define_insn.
21846         (aarch64_<sur>abal<mode>_4): Likewise.
21847         (aarch64_<sur>adalp<mode>_3): Likewise.
21848         (<sur>sadv16qi): New define_expand.
21850 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
21852         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
21853         (*movdf_internal): Ditto.
21854         (*rcpsf2_sse): Ditto.
21855         (*rsqrtsf2_sse): Ditto.
21856         (*sqrt<mode>2_sse): Ditto.
21858 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
21860         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
21861         eor3q<mode>4.
21862         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
21863         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
21864         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
21865         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
21866         vbcaxq_s64): New.
21867         * config/aarch64/arm_neon.h: Likewise.
21868         * config/aarch64/iterators.md (VQ_I): New.
21870 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
21872         * config.gcc: Add arc/t-multilib-linux to tmake_file for
21873         arc*-*-linux*.
21874         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
21875         MULTILIB_DIRNAMES
21877 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21879         * config/nds32/constraints.md (S): New constraint.
21880         * config/nds32/nds32.md (call_internal): Use constraint S.
21881         (call_value_internal): Likewise.
21882         (sibcall_internal): Likewise.
21883         (sibcall_value_internal): Likewise.
21885 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21886             Chung-Ju Wu  <jasonwucj@gmail.com>
21888         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
21889         into consideration.
21891 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21892             Chung-Ju Wu  <jasonwucj@gmail.com>
21894         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
21895         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
21896         (nds32_rtx_costs_impl): Simplify.
21897         (nds32_address_cost_impl): Simplify.
21898         (nds32_init_rtx_costs): New function.
21899         (nds32_rtx_costs_speed_prefer): Likewise.
21900         (nds32_rtx_costs_size_prefer): Likewise.
21901         (nds32_address_cost_speed_prefer): Likewise.
21902         (nds32_address_cost_speed_fwprop): Likewise.
21903         (nds32_address_cost_size_prefer): Likewise.
21904         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
21905         * config/nds32/nds32.c (nds32_option_override): Use
21906         nds32_init_rtx_costs function.
21908 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21910         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
21911         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
21912         (TARGET_PIPELINE_N8): Likewise.
21913         (TARGET_PIPELINE_N10): Likewise.
21914         (TARGET_PIPELINE_N13): Likewise.
21915         (TARGET_PIPELINE_GRAYWOLF): Likewise.
21917 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
21919         * config/nds32/nds32-fpu.md: Update copyright year.
21921 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21923         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
21925 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21927         * config/nds32/nds32.c
21928         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
21929         * config/nds32/nds32.opt (minline-asm-r15): New option.
21931 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21933         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
21934         MASK_HW_ABS.
21935         * config/nds32/nds32.md (abssi2): New pattern.
21937 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
21939         * config/i386/i386.md (rex64namesuffix): New mode attribute.
21940         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
21941         Merge insn pattern from sse_cvtsi2ss<round_name> and
21942         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
21943         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
21944         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
21945         using SWI48 mode iterator.
21946         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
21947         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
21948         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
21949         pattern from sse_cvttss2si<round_saeonly_name>
21950         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
21951         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
21952         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
21953         using SWI48 mode iterator.
21954         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
21955         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
21956         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
21957         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
21958         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
21959         using SWI48 mode iterator.
21960         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
21961         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
21962         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
21963         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
21964         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
21965         SWI48 mode iterator.
21966         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
21967         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
21968         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
21969         pattern from sse_cvttsd2si<round_saeonly_name>
21970         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
21972 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21974         * config/nds32/nds32-md-auxiliary.c
21975         (nds32_valid_smw_lwm_base_p): Refine.
21976         (nds32_output_smw_single_word): Refine.
21977         (nds32_output_smw_double_word): New.
21978         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
21980 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21982         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
21983         (nds32_output_stack_pop): Refine.
21984         (nds32_expand_unaligned_load): Refine.
21985         (nds32_expand_unaligned_store): Refine.
21987 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
21988             Chung-Ju Wu  <jasonwucj@gmail.com>
21990         * config/nds32/constants.md: Add TP_REGNUM constant.
21991         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
21992         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
21993         UNSPEC_ADD32.
21994         * config/nds32/nds32-doubleword.md: Consider flag_pic.
21995         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
21996         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
21997         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
21998         and PIC code generation.
21999         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
22000         code generation.
22001         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
22002         optimization.
22003         * config/nds32/nds32.md: Support TLS and PIC.
22004         * config/nds32/nds32.c: Support TLS and PIC.
22005         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
22006         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
22007         predicate.
22009 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
22011         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
22012         mode with E_ prefix.
22014 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
22015             Chung-Ju Wu  <jasonwucj@gmail.com>
22017         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
22018         * config/nds32/nds32-md-auxiliary.c
22019         (symbolic_reference_mentioned_p): New.
22020         (nds32_legitimize_ict_address): New.
22021         (nds32_expand_ict_move): New.
22022         (nds32_indirect_call_referenced_p): New.
22023         (nds32_symbol_binds_local_p): Delete.
22024         (nds32_long_call_p): Modify.
22025         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
22026         * config/nds32/nds32-protos.h
22027         (symbolic_reference_mentioned_p): Declare.
22028         (nds32_legitimize_ict_address): Declare.
22029         (nds32_expand_ict_move): Declare.
22030         (nds32_indirect_call_referenced_p): Declare.
22031         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
22032         (nds32_relax_group): Use nds32_ict_const_p as condition.
22033         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
22034         (nds32_asm_file_start): Output ict_model directive in asm code.
22035         (nds32_legitimate_address_p): Consider indirect call.
22036         (nds32_print_operand): Consider indirect call.
22037         (nds32_print_operand_address): Consider indirect call.
22038         (nds32_insert_attributes): Handle "indirect_call" attribute.
22039         (TARGET_LEGITIMATE_ADDRESS_P): Define.
22040         (TARGET_LEGITIMATE_CONSTANT_P): Define.
22041         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
22042         (TARGET_DELEGITIMIZE_ADDRESS): Define.
22043         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
22044         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
22045         (TARGET_ICT_MODEL_SMALL): Define.
22046         (TARGET_ICT_MODEL_LARGE): Define.
22047         * config/nds32/nds32.md (movsi): Consider ict model.
22048         (call, call_value): Consider ict model.
22049         (sibcall, sibcall_value): Consider ict model.
22050         * config/nds32/nds32.opt (mict-model): New option.
22051         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
22052         model.
22054 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
22055             Monk Chiang  <sh.chiang04@gmail.com>
22056             Jim Wilson <jimw@sifive.com>
22058         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
22059         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
22060         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
22061         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
22062         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
22063         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
22064         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
22065         compute save_libcall_adjustment properly.
22066         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
22067         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
22068         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
22069         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
22070         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
22071         (ABI_SPEC): Handle mabi=ilp32e.
22072         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
22073         (RVE): Add RVE mask.
22074         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
22075         <-march>: Add rv32e as an example.
22077 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
22079         PR c++/82899
22080         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
22081         (intra_create_variable_infos): Handle C++ constructors.
22083 2018-05-18  Martin Liska  <mliska@suse.cz>
22085         * passes.def: Remove a redundant pass.
22087 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
22089         PR bootstrap/85838
22090         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
22092 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22094         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
22095         (ARMv4): Update.
22096         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
22097         (ARMv6m): Update.
22098         (armv2, armv2a, armv3, armv3m): Delete architectures.
22099         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
22100         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
22101         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
22102         Delete cpus.
22103         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
22104         (*mulsidi3adddi): Likewise.
22105         (mulsidi3): Likewise.
22106         (*mulsidi3_nov6): Likewise.
22107         (umulsidi3): Likewise.
22108         (umulsidi3_nov6): Likewise.
22109         (umaddsidi4): Likewise.
22110         (*umulsidi3adddi): Likewise.
22111         (smulsi3_highpart): Likewise.
22112         (*smulsi3_highpart_nov6): Likewise.
22113         (umulsi3_highpart): Likewise.
22114         (*umulsi3_highpart_nov6): Likewise.
22115         * config/arm/arm.h (arm_arch3m): Delete.
22116         * config/arm/arm.c (arm_arch3m): Delete.
22117         (arm_option_override_internal): Update armv3-related comment.
22118         (arm_configure_build_target): Delete use of isa_bit_mode32.
22119         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
22120         (arm_rtx_costs_internal): Delete check of arm_arch3m.
22121         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
22122         (mulsa3): Likewise.
22123         (mulusa3): Likewise.
22124         * config/arm/arm-protos.h (arm_arch3m): Delete.
22125         * config/arm/arm-tables.opt: Regenerate.
22126         * config/arm/arm-tune.md: Likewise.
22127         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
22128         deleted architectures.
22130 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22132         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
22133         (armv5t, armv5te): New features.
22134         (ARMv5, ARMv5e): Delete fgroups.
22135         (ARMv5t, ARMv5te): Adjust for above changes.
22136         (ARMv6m): Likewise.
22137         (armv5, armv5e): Delete arches.
22138         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
22139         arm_arch5.
22140         (*call_reg_arm): Likewise.
22141         (*call_value_reg_armv5): Likewise.
22142         (*call_value_reg_arm): Likewise.
22143         (*call_symbol): Likewise.
22144         (*call_value_symbol): Likewise.
22145         (*sibcall_insn): Likewise.
22146         (*sibcall_value_insn): Likewise.
22147         (clzsi2): Likewise.
22148         (prefetch): Likewise.
22149         (define_split and define_peephole2 dependent on arm_arch5):
22150         Likewise.
22151         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
22152         arm_arch5e.
22153         (TARGET_ARM_QBIT): Likewise.
22154         (TARGET_DSP_MULTIPLY): Likewise.
22155         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
22156         (arm_arch5, arm_arch5e): Delete.
22157         (arm_arch5t, arm_arch5te): Declare.
22158         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
22159         (arm_arch5t): Declare.
22160         (arm_option_reconfigure_globals): Update for the above.
22161         (arm_options_perform_arch_sanity_checks): Update comment, replace
22162         use of arm_arch5 with arm_arch5t.
22163         (use_return_insn): Likewise.
22164         (arm_emit_call_insn): Likewise.
22165         (output_return_instruction): Likewise.
22166         (arm_final_prescan_insn): Likewise.
22167         (arm_coproc_builtin_available): Likewise.
22168         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
22169         arm_arch5e with arm_arch5t and arm_arch5te.
22170         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
22171         (arm_arch5t, arm_arch5te): Declare.
22172         * config/arm/arm-tables.opt: Regenerate.
22173         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
22174         * config/arm/t-multilib: Likewise.
22175         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
22176         instead of arm_arch5.
22177         (*call_reg_thumb1): Likewise.
22178         (*call_value_reg_thumb1_v5): Likewise.
22179         (*call_value_reg_thumb1): Likewise.
22180         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
22181         unreachable path.
22182         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
22184 2018-05-18  Martin Liska  <mliska@suse.cz>
22186         PR gcov-profile/84846
22187         * doc/gcov.texi: Document -t option of gcov tool.
22189 2018-05-18  Martin Liska  <mliska@suse.cz>
22191         PR gcov-profile/84846
22192         * gcov.c (print_usage): Add new -t option.
22193         (process_args): Handle the option.
22194         (generate_results): Use stdout as output when requested by
22195         the option.
22197 2018-05-18  Martin Liska  <mliska@suse.cz>
22199         PR gcov-profile/84846
22200         * coverage.c (coverage_init): Write PWD to .gcno file.
22201         * doc/gcov.texi: Document how working directory is printed.
22202         * gcov-dump.c (dump_gcov_file): Print PWD.
22203         * gcov.c (output_intermediate_file): Likewise.
22204         (read_graph_file): Read PWD string.
22205         (output_lines): Print PWD.
22207 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22209         PR middle-end/85817
22210         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
22211         for retval and return false if all args to phi are zero.
22213 2018-05-18  Richard Biener  <rguenther@suse.de>
22215         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
22216         method.
22217         (evrp_dom_walker::before_dom_children): Call it.
22219 2018-05-18  Richard Biener  <rguenther@suse.de>
22221         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
22222         results when processing array refs with variable index.
22224 2018-05-18  Toon Moene  <toon@moene.org>
22226         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
22227         directly after that of -floop-interchange. Indicate that both
22228         options are enabled by default when specifying -O3.
22230 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22232         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
22233         iterator.  Delete separate integer-mode vec_set<mode> expander.
22234         (aarch64_simd_vec_setv2di): Delete.
22235         (vec_setv2di): Delete.
22236         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
22237         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
22238         the "w, r" alternative.
22240 2018-05-18  Martin Liska  <mliska@suse.cz>
22242         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
22243         * tree-pass.h (make_pass_lower_switch_O0): New function.
22244         * tree-switch-conversion.c (node_has_low_bound): Remove.
22245         (node_has_high_bound): Likewise.
22246         (node_is_bounded): Likewise.
22247         (class pass_lower_switch): Make it a template type and create
22248         two instances.
22249         (pass_lower_switch::execute): Add template argument.
22250         (make_pass_lower_switch): New function.
22251         (make_pass_lower_switch_O0): New function.
22252         (do_jump_if_equal): Remove.
22253         (emit_case_nodes): Simplify to just handle all 3 cases and leave
22254         all the hard work to tree optimization passes.
22256 2018-05-18  Martin Liska  <mliska@suse.cz>
22258         * dbgcnt.c (limit_low): Renamed from limit.
22259         (limit_high): New variable.
22260         (dbg_cnt_is_enabled): Check for upper limit.
22261         (dbg_cnt): Adjust dumping.
22262         (dbg_cnt_set_limit_by_index): Add new argument for high
22263         value.
22264         (dbg_cnt_set_limit_by_name): Likewise.
22265         (dbg_cnt_process_single_pair): Parse new format.
22266         (dbg_cnt_process_opt): Use strtok.
22267         (dbg_cnt_list_all_counters): Remove 'value' and add
22268         'limit_high'.
22269         * doc/invoke.texi: Document changes.
22271 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
22273         * doc/sourcebuild.texi (scalar_all_fma): Document.
22274         * tree.def (FMA_EXPR): Delete.
22275         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
22276         * internal-fn.c (ternary_direct): New macro.
22277         (expand_ternary_optab_fn): Likewise.
22278         (direct_ternary_optab_supported_p): Likewise.
22279         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
22280         * builtins.c (fold_builtin_fma): Delete.
22281         (fold_builtin_3): Don't call it.
22282         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
22283         * expr.c (expand_expr_real_2): Likewise.
22284         * fold-const.c (operand_equal_p): Likewise.
22285         (fold_ternary_loc): Likewise.
22286         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
22287         * gimple.c (DEFTREECODE): Likewise.
22288         * gimplify.c (gimplify_expr): Likewise.
22289         * optabs-tree.c (optab_for_tree_code): Likewise.
22290         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
22291         * tree-eh.c (operation_could_trap_p): Likewise.
22292         (stmt_could_throw_1_p): Likewise.
22293         * tree-inline.c (estimate_operator_cost): Likewise.
22294         * tree-pretty-print.c (dump_generic_node): Likewise.
22295         (op_code_prio): Likewise.
22296         * tree-ssa-loop-im.c (stmt_cost): Likewise.
22297         * tree-ssa-operands.c (get_expr_operands): Likewise.
22298         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
22299         * fold-const-call.h (fold_fma): Delete.
22300         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
22301         CFN_FNMA and CFN_FNMS.
22302         (fold_fma): Delete.
22303         * genmatch.c (combined_fn): New enum.
22304         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
22305         (commutative_op): New function.
22306         (commutate): Use it.  Handle more than 2 operands.
22307         (dt_operand::gen_gimple_expr): Use commutative_op.
22308         (parser::parse_expr): Allow :c to be used with non-binary
22309         operators if the commutative operand is known.
22310         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
22311         CFN_FMS, CFN_FNMA and CFN_FNMS.
22312         (backprop::process_assign_use): Remove FMA_EXPR handling.
22313         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
22314         (gen_hsa_fma): New function.
22315         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
22316         IFN_FNMA and IFN_FNMS.
22317         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
22318         * gimple-fold.h (follow_all_ssa_edges): Declare.
22319         * gimple-fold.c (follow_all_ssa_edges): New function.
22320         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
22321         gimple_build interface and use follow_all_ssa_edges to fold the result.
22322         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
22323         instead of checking for optabs directly.
22324         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
22325         rather than FMA_EXPRs.
22326         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
22327         call to IFN_FMA instead of an FMA_EXPR.
22329 2018-05-17  Jim Wilson  <jimw@sifive.com>
22331         * expr.c (do_tablejump): When converting index to Pmode, if we have a
22332         sign extended promoted subreg, and the range does not have the sign bit
22333         set, then do a sign extend.
22335         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
22336         test, check for sign extended subreg and/or constant operands, and
22337         do a sign extend in that case.
22339 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
22341         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
22342         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
22343         Add untyped.
22344         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
22345         Change logics_shift_reg to logics_shift_imm.
22346         (thunderx2t99_fp_loadpair_basic): Delete.
22347         (thunderx2t99_fp_storepair_basic): Delete.
22348         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
22349         (thunderx2t99_asimd_polynomial): Delete.
22350         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
22351         and neon_fp_mul_d_scalar_q.
22352         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
22353         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
22354         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
22355         (thunderx2t99_asimd_lut): Add missing tbl types.
22356         (thunderx2t99_asimd_ext): Delete.
22357         (thunderx2t99_asimd_load1_1_mult): Delete.
22358         (thunderx2t99_asimd_load1_2_mult): Delete.
22359         (thunderx2t99_asimd_load1_ldp): New.
22360         (thunderx2t99_asimd_load1): New.
22361         (thunderx2t99_asimd_load2): Add missing *load2* types.
22362         (thunderx2t99_asimd_load3): New.
22363         (thunderx2t99_asimd_load4): New.
22364         (thunderx2t99_asimd_store1_1_mult): Delete.
22365         (thunderx2t99_asimd_store1_2_mult): Delete.
22366         (thunderx2t99_asimd_store2_mult): Delete.
22367         (thunderx2t99_asimd_store2_onelane): Delete.
22368         (thunderx2t99_asimd_store_stp): New.
22369         (thunderx2t99_asimd_store1): New.
22370         (thunderx2t99_asimd_store2): New.
22371         (thunderx2t99_asimd_store3): New.
22372         (thunderx2t99_asimd_store4): New.
22374 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
22376         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
22377         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
22379 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
22380             Segher Boessenkool  <segher@kernel.crashing.org>
22382         PR target/85698
22383         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
22384         operand.
22386 2018-05-17  Richard Biener  <rguenther@suse.de>
22388         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
22389         for pruning loop and prune defs feeding only already visited PHIs.
22391 2018-05-17  Richard Biener  <rguenther@suse.de>
22393         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
22395 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
22396             Richard Biener  <rguenther@suse.de>
22398         PR tree-optimization/85793
22399         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
22400         for VMAT_ELEMENTWISE.
22402 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
22404         * internal-fn.h (lookup_internal_fn): Declare
22405         * internal-fn.c (lookup_internal_fn): New function.
22406         * gimple.c (gimple_build_call_from_tree): Handle calls to
22407         internal functions.
22408         * gimple-pretty-print.c (dump_gimple_call): Print "." before
22409         internal function names.
22410         * tree-pretty-print.c (dump_generic_node): Likewise.
22411         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
22413 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
22415         * gimple-fold.h (gimple_build): Make the function forms take
22416         combined_fn rather than built_in_function.
22417         (gimple_simplify): Likewise.
22418         * gimple-match-head.c (gimple_simplify): Likewise.
22419         * gimple-fold.c (gimple_build): Likewise.
22420         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
22421         rather than gimple_build_call_internal.
22422         (get_initial_defs_for_reduction): Likewise.
22423         (vect_create_epilog_for_reduction): Likewise.
22424         (vectorizable_live_operation): Likewise.
22426 2018-05-17  Martin Liska  <mliska@suse.cz>
22428         * gimple-ssa-sprintf.c (format_directive): Do not use
22429         space in between 'G_' and '('.
22431 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
22433         PR target/85323
22434         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
22435         even if the mask is not all ones.
22437         PR target/85323
22438         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
22439         vector.
22440         (ix86_gimple_fold_builtin): Likewise.
22442         PR target/85323
22443         * config/i386/i386.c: Include tree-vector-builder.h.
22444         (ix86_vector_shift_count): New function.
22445         (ix86_fold_builtin): Fold shift builtins by scalar count.
22446         (ix86_gimple_fold_builtin): Likewise.
22448         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
22449         _mm512_setzero): New intrinsics.
22451 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
22452             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22454         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
22455         code generation for cases where splatting a value is not useful.
22456         * simplify-rtx.c (simplify_ternary_operation): Simplify
22457         vec_merge across a vec_duplicate and a paradoxical subreg forming
22458         a vector mode to a vec_concat.
22460 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
22462         * config.gcc: Support "goldmont-plus".
22463         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
22464         "goldmont-plus".
22465         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22466         PROCESSOR_GOLDMONT_PLUS.
22467         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
22468         (processor_target_table): Add "goldmont-plus".
22469         (PTA_GOLDMONT_PLUS): Define.
22470         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
22471         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
22472         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
22473         (fold_builtin_cpu): Add "goldmont-plus".
22474         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
22475         (ix86_option_override_internal): Add "goldmont-plus".
22476         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
22477         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
22478         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
22479         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
22481 2018-05-17  Richard Biener  <rguenther@suse.de>
22483         PR tree-optimization/85757
22484         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
22485         remove defs that only feed that PHI from further processing.
22487 2018-05-16  Jim Wilson  <jimw@sifive.com>
22489         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
22490         asterisk to name.
22491         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
22493 2018-05-16  Mark Wielaard  <mark@klomp.org>
22495         * dwarf2out.c (count_index_strings): New function.
22496         (output_indirect_strings): Call count_index_strings and generate
22497         header for dwarf_version >= 5.
22499 2018-05-16  Mark Wielaard  <mark@klomp.org>
22501         * dwarf2out.c (dwarf_FORM): New function.
22502         (set_indirect_string): Use dwarf_FORM.
22503         (reset_indirect_string): Likewise.
22504         (size_of_die): Likewise.
22505         (value_format): Likewise.
22506         (output_die): Likewise.
22507         (add_skeleton_AT_string): Likewise.
22508         (output_macinfo_op): Likewise.
22509         (index_string): Likewise.
22510         (output_index_string_offset): Likewise.
22511         (output_index_string): Likewise.
22512         (count_index_strings): Likewise.
22514 2018-05-16  Carl Love  <cel@us.ibm.com>
22516         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
22517         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
22519 2018-05-16  Martin Jambor  <mjambor@suse.cz>
22521         * ipa-prop.c (ipa_free_all_edge_args): Remove.
22522         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
22524 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
22526         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
22527         (fnma<mode>4): Likewise.
22528         (fms<mode>4): Likewise.
22529         (fnms<mode>4): Likewise.
22530         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
22531         (aarch64_fnma<mode>4): Likewise.
22532         (aarch64_fms<mode>4): Likewise.
22533         (aarch64_fnms<mode>4): Likewise.
22534         (aarch64_fnmadd<mode>4): Likewise.
22536 2018-05-16  Jason Merrill  <jason@redhat.com>
22538         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
22540 2018-05-16  Richard Biener  <rguenther@suse.de>
22542         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
22543         (dump_stmt_cost): Declare.
22544         (add_stmt_cost): Dump cost we add.
22545         (add_stmt_costs): New function.
22546         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
22547         No longer exported.
22548         (vect_analyze_stmt): Adjust prototype.
22549         (vectorizable_condition): Likewise.
22550         (vectorizable_live_operation): Likewise.
22551         (vectorizable_reduction): Likewise.
22552         (vectorizable_induction): Likewise.
22553         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
22554         cost vector to pass to vectorizable_ and record afterwards.
22555         (vect_model_reduction_cost): Take cost vector argument and adjust.
22556         (vect_model_induction_cost): Likewise.
22557         (vectorizable_reduction): Likewise.
22558         (vectorizable_induction): Likewise.
22559         (vectorizable_live_operation): Likewise.
22560         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
22561         SLP_TREE_NUMBER_OF_VEC_STMTS.
22562         (vect_analyze_slp_cost_1): Remove.
22563         (vect_analyze_slp_cost): Likewise.
22564         (vect_slp_analyze_node_operations): Take visited args and
22565         a target cost vector.  Avoid processing already visited stmt sets.
22566         (vect_slp_analyze_operations): Use a local cost vector to gather
22567         costs and register those of non-discarded instances.
22568         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
22569         (vect_schedule_slp_instance): Remove copying of
22570         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
22571         zero.
22572         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
22573         adding cost.  Record cost entry location.
22574         (vect_prologue_cost_for_slp_op): Function to compute cost of
22575         a constant or invariant generated for SLP vect in the prologue,
22576         split out from vect_analyze_slp_cost_1.
22577         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
22578         (vect_model_promotion_demotion_cost): Likewise.
22579         (vect_model_store_cost): Likewise, make static.
22580         (vect_model_load_cost): Likewise.
22581         (vectorizable_bswap): Add cost vector arg and adjust.
22582         (vectorizable_call): Likewise.
22583         (vectorizable_simd_clone_call): Likewise.
22584         (vectorizable_conversion): Likewise.
22585         (vectorizable_assignment): Likewise.
22586         (vectorizable_shift): Likewise.
22587         (vectorizable_operation): Likewise.
22588         (vectorizable_store): Likewise.
22589         (vectorizable_load): Likewise.
22590         (vectorizable_condition): Likewise.
22591         (vectorizable_comparison): Likewise.
22592         (can_vectorize_live_stmts): Likewise.
22593         (vect_analyze_stmt): Likewise.
22594         (vect_transform_stmt): Adjust calls to vectorizable_*.
22595         * tree-vectorizer.c: Include gimple-pretty-print.h.
22596         (dump_stmt_cost): New function.
22598 2018-05-16  Richard Biener  <rguenther@suse.de>
22600         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
22601         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
22602         * tree-ssa-dse.c: Include tree-ssa-loop.h.
22603         (check_name): New callback.
22604         (dse_classify_store): Track cycles via a visited bitmap of PHI
22605         defs and simplify handling of in-loop and across loop dead stores
22606         and properly fail for loop-variant refs.  Handle byte-tracking with
22607         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
22608         limiting the walk.
22610 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
22612         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
22613         (vect_get_mask_type_for_stmt): Likewise.
22614         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
22615         split out from...
22616         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
22617         to determine the statement's vector type and the vector type that
22618         should be used for calculating nunits.  Deal with cases in which
22619         the type has to be deferred.
22620         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
22621         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
22622         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
22623         (vect_determine_vf_for_stmt): New functions, split out from...
22624         (vect_determine_vectorization_factor): ...here.
22625         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
22626         (vect_get_mask_type_for_stmt): New functions, split out from
22627         vect_determine_vectorization_factor.
22629 2018-05-16  Richard Biener  <rguenther@suse.de>
22631         * tree-cfg.c (verify_gimple_assign_ternary): Properly
22632         verify the [VEC_]COND_EXPR embedded comparison.
22634 2018-05-15  Martin Sebor  <msebor@redhat.com>
22636         PR tree-optimization/85753
22637         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
22638         RECORD_TYPE in addition to ARRAY_TYPE.
22640 2018-05-15  Martin Sebor  <msebor@redhat.com>
22642         PR middle-end/85643
22643         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
22645 2018-05-15  Richard Biener  <rguenther@suse.de>
22647         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
22648         add by_clobber_p one.  Change algorithm to collect all defs
22649         representing uses we need to walk and try reducing them to
22650         a single one before failing.
22651         (dse_dom_walker::dse_optimize_stmt): Adjust.
22653 2018-05-13  Mark Wielaard  <mark@klomp.org>
22655         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
22656         (size_of_loc_descr): Likewise.
22657         (output_loc_operands): Likewise.
22658         (output_loc_operands_raw): Likewise.
22659         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
22660         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
22661         (hash_loc_operands): Likewise.
22662         (compare_loc_operands): Likewise.
22664 2018-05-14  Mark Wielaard  <mark@klomp.org>
22666         * dwarf2out.c (count_index_addrs): New function.
22667         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
22669 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22671         PR tree-optimization/83648
22672         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
22673         return value as malloc candidate.
22675 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22677         PR ipa/85734
22678         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
22679         param as true in call to suggest_attribute.
22681 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
22683         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
22684         -mreadonly-in-sdata.
22686 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22688         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
22689         New pattern.
22690         (aarch64_crypto_aesd_fused): Likewise.
22692 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
22694         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
22695         (movsi_aarch64): Likewise.
22696         (load_pairsi): Likewise.
22697         (load_pairdi): Likewise.
22698         (store_pairsi): Likewise.
22699         (store_pairdi): Likewise.
22700         (load_pairsf): Likewise.
22701         (load_pairdf): Likewise.
22702         (store_pairsf): Likewise.
22703         (store_pairdf): Likewise.
22704         (zero_extend): Likewise.
22705         (trunc): Swap alternatives.
22706         (fcvt_target): Add '?' to prefer w over r.
22708 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
22710         PR target/85756
22711         * config/i386/i386.md: Disallow non-commutative arithmetics in
22712         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
22713         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
22714         in the peephole2 before it.
22716 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
22718         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
22719         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
22720         (ix86_handle_option): Handle -mcldemote.
22721         * config.gcc: New header.
22722         * config/i386/cldemoteintrin.h: New file.
22723         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
22724         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
22725         -mcldemote.
22726         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22727         OPTION_MASK_ISA_CLDEMOTE.
22728         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
22729         (ix86_valid_target_attribute_inner_p): Ditto.
22730         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
22731         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
22732         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
22733         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
22734         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
22735         (cldemote): New.
22736         * config/i386/i386.opt: Add -mcldemote.
22737         * config/i386/x86intrin.h: New header.
22738         * doc/invoke.texi: Add -mcldemote.
22740 2018-05-14  Richard Biener  <rguenther@suse.de>
22742         * doc/match-and-simplify.texi: Adjust :s documentation.
22744 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
22746         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
22747         intended memcpy size.
22748         (REORDER_45): Likewise.
22750 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
22752         * sort.cc: New file.
22753         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
22754         * vec.c (qsort_chk): Use gcc_qsort.
22755         * Makefile.in (OBJS-libcommon): Add sort.o.
22756         (build/sort.o): New target.  Use it...
22757         (BUILD_RTL): ... here, and...
22758         (build/gencfn-macros): ... here, and...
22759         (build/genmatch): ... here.
22761 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22762             Chung-Ju Wu  <jasonwucj@gmail.com>
22764         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
22765         * config/nds32/nds32-graywolf.md: New file.
22766         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
22767         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
22768         pipeline.
22769         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
22770         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
22771         * config/nds32/nds32.md (pipeline_model): Add graywolf.
22772         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
22773         * config/nds32/pipelines.md: Include n15 settings.
22775 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22776             Chung-Ju Wu  <jasonwucj@gmail.com>
22778         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
22779         * config/nds32/nds32-n13.md: New file.
22780         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
22781         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
22782         pipeline.
22783         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
22784         * config/nds32/nds32.md (pipeline_model): Add n13.
22785         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
22786         * config/nds32/pipelines.md: Include n13 settings.
22788 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22789             Chung-Ju Wu  <jasonwucj@gmail.com>
22791         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
22792         * config/nds32/nds32-n10.md: New file.
22793         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
22794         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
22795         pipeline.
22796         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
22797         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
22798         * config/nds32/nds32.md (pipeline_model): Add n10.
22799         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
22800         * config/nds32/pipelines.md: Include n10 settings.
22802 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
22803             Kito Cheng  <kito.cheng@gmail.com>
22804             Chung-Ju Wu  <jasonwucj@gmail.com>
22806         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
22807         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22808         Add enum values for DSP extension instructions.
22809         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
22810         New constraints.
22811         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
22812         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
22813         New code iterators.
22814         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
22815         * config/nds32/nds32-dspext.md: New file for DSP implementation.
22816         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
22817         * config/nds32/nds32-intrinsic.md: Likewise.
22818         * config/nds32/nds32_intrinsic.h: Likewise.
22819         * config/nds32/nds32-md-auxiliary.c: Likewise.
22820         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
22821         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
22822         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
22823         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
22824         * config/nds32/nds32-protos.h: New declarations for DSP extension.
22825         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
22826         TYPE_DMAC in switch statement.
22827         * config/nds32/nds32.c: New checking and implementation for DSP
22828         extension instructions.
22829         * config/nds32/nds32.h: Likewise.
22830         * config/nds32/nds32.md: Likewise.
22831         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
22832         * config/nds32/predicates.md: Implement new predicates for DSP
22833         extension.
22835 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
22837         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
22838         Reformat alternatives and attributes so it is easier to identify
22839         which constraints/attributes go with which instruction.
22840         (mov<mode>_hardfloat32, FMOVE64): Likewise.
22841         (mov<mode>_softfloat32, FMOVE64): Likewise.
22842         (mov<mode>_hardfloat64, FMOVE64): Likewise.
22843         (mov<mode>_softfloat64, FMOVE64): Likewise.
22845 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22847         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22848         subsection.
22849         (Basic PowerPC Built-in Functions): The new name of the
22850         subsection previously known as "PowerPC Built-in Functions".
22851         (Basic PowerPC Built-in Functions Available on all Configurations):
22852         New subsubsection.
22853         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
22854         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
22855         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
22856         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
22858 2018-05-11  Martin Jambor  <mjambor@suse.cz>
22860         PR ipa/85655
22861         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
22862         single const.
22864 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22866         PR target/85733
22867         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
22869 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
22871         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
22872         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
22873         (ix86_handle_option): Handle -mwaitpkg.
22874         * config.gcc: New header.
22875         * config/i386/cpuid.h (bit_WAITPKG): New bit.
22876         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
22877         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
22878         function type.
22879         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22880         OPTION_MASK_ISA_WAITPKG.
22881         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
22882         (ix86_option_override_internal): Add PTA_WAITPKG.
22883         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
22884         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
22885         IX86_BUILTIN_TPAUSE.
22886         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
22887         __builtin_ia32_umwait and __builtin_ia32_tpause.
22888         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
22889         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
22890         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
22891         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
22892         UNSPECV_TPAUSE): New.
22893         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
22894         * config/i386/i386.opt: Add -mwaitpkg.
22895         * config/i386/waitpkgintrin.h: New file.
22896         * config/i386/x86intrin.h: New header.
22897         * doc/invoke.texi: Add -mwaitpkg.
22899 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22901         PR target/85606
22902         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
22903         equivalent.
22904         (cortex-m0): Use armv6s-m isa.
22905         (cortex-m0plus): Likewise.
22906         (cortex-m1): Likewise.
22907         (cortex-m0.small-multiply): Likewise.
22908         (cortex-m0plus.small-multiply): Likewise.
22909         (cortex-m1.small-multiply): Likewise.
22911 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
22912             Jakub Jelinek  <jakub@redhat.com>
22914         PR tree-optimization/85692
22915         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
22916         source permute as well.
22918 2018-05-11  Martin Liska  <mliska@suse.cz>
22920         PR sanitizer/85556
22921         * doc/extend.texi: Document LLVM style format for no_sanitize
22922         attribute.
22924 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
22926         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
22927         mode_supports_vsx_dform_quad to mode_supports_dq_form.
22928         (mode_supports_vsx_dform_quad): Likewise.
22929         (mode_supports_vmx_dform): Move these functions to be next to the
22930         other mode_supports functions.
22931         (mode_supports_dq_form): Likewise.
22932         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
22933         mode_supports_dq_form.
22934         (reg_offset_addressing_ok_p): Likewise.
22935         (offsettable_ok_by_alignment): Likewise.
22936         (rs6000_legitimate_offset_address_p): Likewise.
22937         (legitimate_lo_sum_address_p): Likewise.
22938         (rs6000_legitimize_address): Likewise.
22939         (rs6000_legitimize_reload_address): Likewise.
22940         (rs6000_secondary_reload_inner): Likewise.
22941         (rs6000_preferred_reload_class): Likewise.
22942         (rs6000_output_move_128bit): Likewise.
22944 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22946         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
22947         Generate SImode target register for null target.
22948         <case IX86_BUILTIN_XGETBV>: Ditto.
22949         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
22950         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
22952 2018-05-10  Carl Love  <cel@us.ibm.com>
22954         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
22955         dcbtt and dcbtstt if operands[2] is 0.
22957 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22959         PR target/85693
22960         * config/i386/sse.md (usadv64qi): New expander.
22962 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
22964         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
22965         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
22966         -maltivec=be support.
22967         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
22968         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
22969         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
22970         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
22971         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
22972         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
22973         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
22974         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
22975         altivec_vsumsws): Adjust.
22976         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
22977         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
22978         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
22979         support.
22980         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
22981         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
22982         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
22983         (altivec_lve<VI_char>x): Delete expand.
22984         (*altivec_lve<VI_char>x_internal): Rename to...
22985         (altivec_lve<VI_char>x): ... this.
22986         (altivec_lvxl_<mode>): Delete expand.
22987         (*altivec_lvxl_<mode>_internal): Rename to ...
22988         (altivec_lvxl_<mode>): ... this.
22989         (altivec_stvxl_<mode>): Delete expand.
22990         (*altivec_stvxl_<mode>_internal): Rename to ...
22991         (altivec_stvxl_<mode>): ... this.
22992         (altivec_stve<VI_char>x): Delete expand.
22993         (*altivec_stve<VI_char>x_internal): Rename to ...
22994         (altivec_stve<VI_char>x): ... this.
22995         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
22996         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
22997         reduc_plus_scal_<mode>): Adjust.
22998         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
22999         comment.
23000         (rs6000_cpu_cpp_builtins): Adjust.
23001         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
23002         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
23003         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
23004         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
23005         -maltivec=be support.
23006         (rs6000_split_vec_extract_var): Adjust.
23007         (rs6000_split_v4si_init): Adjust.
23008         (swap_selector_for_mode): Delete.
23009         (altivec_expand_lvx_be, altivec_expand_stvx_be,
23010         altivec_expand_stvex_be): Delete.
23011         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
23012         -maltivec=be support.
23013         (rs6000_gimple_fold_builtin): Ditto.
23014         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
23015         Adjust.
23016         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
23017         (TARGET_DIRECT_MOVE_64BIT): Adjust.
23018         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
23019         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
23020         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
23021         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
23022         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
23023         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
23024         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
23025         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
23026         anonymous split): Adjust.
23027         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
23028         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
23030 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
23032         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
23033         when --with-gxx-include-dir is also specified.
23034         * configure: Regenerate.
23036 2018-05-09  Jim Wilson  <jimw@sifive.com>
23038         PR target/84797
23039         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
23040         * config/riscv/t-withmultilib: New.
23041         * config/riscv/withmultilib.h: New.
23042         * doc/install.texi: Document RISC-V --with-multilib-list support.
23044 2018-05-09  Richard Biener  <rguenther@suse.de>
23046         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
23047         vector.
23048         (vect_bb_vectorization_profitable_p): Adjust.  Compute
23049         actual scalar cost using the cost vector and the add_stmt_cost
23050         machinery.
23052 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23054         PR rtl-optimization/85645
23055         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
23056         in the REG_CFA_REGISTER note for LR, don't leave it empty.
23058 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23060         PR rtl-optimization/85645
23061         * shrink-wrap.c (spread_components): Return a boolean saying if
23062         anything was changed.
23063         (try_shrink_wrapping_separate): Iterate spread_components until
23064         nothing changes anymore.
23066 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23068         PR rtl-optimization/85645
23069         * regrename.c (build_def_use): Also kill the chains that include the
23070         destination of a REG_CFA_REGISTER note.
23072 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
23074         PR rtl-optimization/85645
23075         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
23076         insn that has a REG_CFA_REGISTER note.
23078 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
23080         * cfgexpand.c (expand_clobber): New function.
23081         (expand_gimple_stmt_1): Use it.
23082         * tree-vect-stmts.c (vect_clobber_variable): New function,
23083         split out from...
23084         (vectorizable_simd_clone_call): ...here.
23085         (vectorizable_store): Emit a clobber either side of an
23086         IFN_STORE_LANES sequence.
23087         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
23089 2018-05-09  Tom de Vries  <tom@codesourcery.com>
23091         PR target/85626
23092         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
23093         (define_insn "trap_if_false"): Add exit after trap.
23095 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
23097         PR rtl-optimization/85638
23098         * bb-reorder.c: Include common/common-target.h.
23099         (create_forwarder_block): New function extracted from...
23100         (fix_up_crossing_landing_pad): ...here.  Rename into...
23101         (dw2_fix_up_crossing_landing_pad): ...this.
23102         (sjlj_fix_up_crossing_landing_pad): New function.
23103         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
23104         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
23105         from both partitions and exit the loop after one iteration.
23107 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23109         Revert:
23110         * doc/extend.texi (PowerPC Built-in Functions): Rename this
23111         subsection.
23112         (Basic PowerPC Built-in Functions): The new name of the
23113         subsection previously known as "PowerPC Built-in Functions".
23114         (Basic PowerPC Built-in Functions Available on all Configurations):
23115         New subsubsection.
23116         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23117         subsubsection.
23118         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23119         subsubsection.
23120         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23121         subsubsection.
23122         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23123         subsubsection.
23125 2018-05-08  Jim Wilson  <jimw@sifive.com>
23127         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
23128         (LD_EMUL_SUFFIX): New.
23129         (LINK_SPEC): Use it.
23131 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23133         * doc/extend.texi (PowerPC Built-in Functions): Rename this
23134         subsection.
23135         (Basic PowerPC Built-in Functions): The new name of the
23136         subsection previously known as "PowerPC Built-in Functions".
23137         (Basic PowerPC Built-in Functions Available on all Configurations):
23138         New subsubsection.
23139         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23140         subsubsection.
23141         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23142         subsubsection.
23143         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23144         subsubsection.
23145         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23146         subsubsection.
23148 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
23150         PR target/85683
23151         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
23152         after cmpelim optimization.
23154 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
23156         * config.gcc: Support "goldmont".
23157         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
23158         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23159         PROCESSOR_GOLDMONT.
23160         * config/i386/i386.c (m_GOLDMONT): Define.
23161         (processor_target_table): Add "goldmont".
23162         (PTA_GOLDMONT): Define.
23163         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
23164         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
23165         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
23166         (fold_builtin_cpu): Add "goldmont".
23167         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
23168         (ix86_option_override_internal): Add "goldmont".
23169         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
23170         (processor_type): Add PROCESSOR_GOLDMONT.
23171         * config/i386/i386.md: Add CPU "glm".
23172         * config/i386/glm.md: New file.
23173         * config/i386/x86-tune.def: Add m_GOLDMONT.
23174         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
23176 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
23178         PR target/85572
23179         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
23180         E_V4DImode.
23181         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
23182         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
23183         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
23185         PR target/85317
23186         * config/i386/i386.c (ix86_fold_builtin): Handle
23187         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
23189         PR target/85480
23190         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
23191         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
23193 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
23195         PR target/85658
23196         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
23197         (check_arch): Likewise.
23198         (check_fpu): Return the result rather than printing it.
23199         (end arch): Fix operator precedence.
23200         (end cpu): Likewise.
23201         (END): Print the result from check_fpu.
23203 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
23204             Alan Hayward  <alan.hayward@arm.com>
23205             David Sherwood  <david.sherwood@arm.com>
23207         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
23208         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
23209         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
23210         (*fcmuo<mode>_and): New patterns.
23212 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
23214         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
23215         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
23216         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
23217         (cmp_op, sve_imm_con): New code attributes.
23218         (SVE_COND_INT_CMP, imm_con): Delete.
23219         (cmp_op): Remove above unspecs from int attribute.
23220         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
23221         to...
23222         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
23223         comparison-specific unspecs.
23224         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
23225         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
23226         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
23227         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
23228         (*vec_fcm<cmp_op><mode>): Rename to...
23229         (*fcm<cmp_op><mode>): ...this and adjust likewise.
23230         (*vec_fcmuo<mode>): Rename to...
23231         (*fcmuo<mode>): ...this and adjust likewise.
23232         (*pred_fcm<cmp_op><mode>): New pattern.
23233         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
23234         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
23235         functions.
23236         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
23237         and UNORDERED.
23238         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
23239         (aarch64_emit_sve_predicated_cond): New function.
23240         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
23241         (aarch64_emit_unspec_cond_or): Replace with...
23242         (aarch64_emit_sve_or_conds): ...this new function.  Use
23243         aarch64_emit_sve_ptrue_op for the individual comparisons and
23244         aarch64_emit_binop to OR them together.
23245         (aarch64_emit_inverted_unspec_cond): Replace with...
23246         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
23247         aarch64_emit_sve_ptrue_op for the comparison and
23248         aarch64_emit_unop to invert the result.
23249         (aarch64_expand_sve_vec_cmp_float): Update after the above
23250         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
23252 2018-05-07  Nathan Sidwell  <nathan@acm.org>
23254         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
23255         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
23256         (Backwards Compatibility): Likewise.
23258 2018-05-07  Luis Machado  <luis.machado@linaro.org>
23260         PR bootstrap/85681
23261         Revert:
23262         2018-05-07  Luis Machado  <luis.machado@linaro.org>
23264         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23265         <prefetch_dynamic_strides>: New const bool field.
23266         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23267         prefetch_dynamic_strides.
23268         (exynosm1_prefetch_tune): Likewise.
23269         (thunderxt88_prefetch_tune): Likewise.
23270         (thunderx_prefetch_tune): Likewise.
23271         (thunderx2t99_prefetch_tune): Likewise.
23272         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23273         to false.
23274         (aarch64_override_options_internal): Update to set
23275         PARAM_PREFETCH_DYNAMIC_STRIDES.
23276         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23277         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23278         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23279         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23280         prefetch-dynamic-strides setting.
23282         2018-05-07  Luis Machado  <luis.machado@linaro.org>
23284         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23285         <minimum_stride>: New const int field.
23286         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23287         minimum_stride field.
23288         (exynosm1_prefetch_tune): Likewise.
23289         (thunderxt88_prefetch_tune): Likewise.
23290         (thunderx_prefetch_tune): Likewise.
23291         (thunderx2t99_prefetch_tune): Likewise.
23292         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23293         (aarch64_override_options_internal): Update to set
23294         PARAM_PREFETCH_MINIMUM_STRIDE.
23295         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23296         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23297         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23298         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23299         stride is constant and is below the minimum stride threshold.
23301 2018-05-07  Luis Machado  <luis.machado@linaro.org>
23303         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
23304         to 512.
23306 2018-05-07  Luis Machado  <luis.machado@linaro.org>
23308         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23309         <prefetch_dynamic_strides>: New const bool field.
23310         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23311         prefetch_dynamic_strides.
23312         (exynosm1_prefetch_tune): Likewise.
23313         (thunderxt88_prefetch_tune): Likewise.
23314         (thunderx_prefetch_tune): Likewise.
23315         (thunderx2t99_prefetch_tune): Likewise.
23316         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23317         to false.
23318         (aarch64_override_options_internal): Update to set
23319         PARAM_PREFETCH_DYNAMIC_STRIDES.
23320         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23321         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23322         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23323         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23324         prefetch-dynamic-strides setting.
23326 2018-05-07  Luis Machado  <luis.machado@linaro.org>
23328         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23329         <minimum_stride>: New const int field.
23330         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23331         minimum_stride field.
23332         (exynosm1_prefetch_tune): Likewise.
23333         (thunderxt88_prefetch_tune): Likewise.
23334         (thunderx_prefetch_tune): Likewise.
23335         (thunderx2t99_prefetch_tune): Likewise.
23336         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23337         (aarch64_override_options_internal): Update to set
23338         PARAM_PREFETCH_MINIMUM_STRIDE.
23339         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23340         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23341         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23342         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23343         stride is constant and is below the minimum stride threshold.
23345 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
23347         PR c++/85659
23348         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
23349         the type is addressable.  Don't force op into register if it has
23350         BLKmode.
23352 2018-05-05  Roland McGrath  <mcgrathr@google.com>
23354         PR other/77609
23355         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
23356         any section for which we don't know a specific type it should have,
23357         regardless of name.  Previously this was done only for the exact
23358         names ".init_array", ".fini_array", and ".preinit_array".
23359         (default_elf_asm_named_section): Add comment about
23360         relationship with default_section_type_flags and SECTION_NOTYPE.
23361         (get_section): Don't consider it a type conflict if one side has
23362         SECTION_NOTYPE and the other doesn't, as long as neither has the
23363         SECTION_BSS et al used in the default_section_type_flags logic.
23365 2018-05-05  Tom de Vries  <tom@codesourcery.com>
23367         PR target/85653
23368         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
23369         (workaround_barsyncs): New function.
23370         (nvptx_reorg): Use workaround_barsyncs.
23371         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
23372         (define_expand "nvptx_membar_cta"): New define_expand.
23373         (define_insn "*nvptx_membar_cta"): New insn.
23375 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
23377         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
23378         To improve optimization opportunities.
23379         * builtin-types.def: The new needed builtin types for the above.
23381 2018-05-04  Richard Biener  <rguenther@suse.de>
23383         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
23384         * gimple-ssa-store-merging.c
23385         (imm_store_chain_info::output_merged_store): Remove redundant create,
23386         release split_store vector contents on failure.
23387         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
23388         scalar stmt vector on cache hit.
23390 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
23392         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
23393         Xilinx FP support.
23394         * config.gcc (powerpc-xilinx-eabi*): Remove.
23395         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
23396         support.
23397         (fusion_addis_mem_combo_load): Ditto.
23398         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
23399         FP support.
23400         (rs6000_cpu_cpp_builtins): Ditto.
23401         * config/rs6000/rs6000-linux.c
23402         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
23403         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
23404         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
23405         support.
23406         (rs6000_setup_reg_addr_masks): Ditto.
23407         (rs6000_init_hard_regno_mode_ok): Ditto.
23408         (rs6000_option_override_internal): Ditto.
23409         (legitimate_lo_sum_address_p): Ditto.
23410         (rs6000_legitimize_address): Ditto.
23411         (rs6000_legitimize_reload_address): Ditto.
23412         (rs6000_legitimate_address_p): Ditto.
23413         (abi_v4_pass_in_fpr): Ditto.
23414         (setup_incoming_varargs): Ditto.
23415         (rs6000_gimplify_va_arg): Ditto.
23416         (rs6000_split_multireg_move): Ditto.
23417         (rs6000_savres_strategy): Ditto.
23418         (rs6000_emit_prologue_components): Ditto.
23419         (rs6000_emit_epilogue_components): Ditto.
23420         (rs6000_emit_prologue): Ditto.
23421         (rs6000_emit_epilogue): Ditto.
23422         (rs6000_elf_file_end): Ditto.
23423         (rs6000_function_value): Ditto.
23424         (rs6000_libcall_value): Ditto.
23425         * config/rs6000/rs6000.h: Ditto.
23426         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
23427         (TARGET_MINMAX): ... this.  New.
23428         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
23429         * config/rs6000/rs6000.md: Remove Xilinx FP support.
23430         (*movsi_internal1_single): Delete.
23431         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
23432         mfpu=, mxilinx-fpu): Delete.
23433         * config/rs6000/singlefp.h: Delete.
23434         * config/rs6000/sysv4.h: Remove Xilinx FP support.
23435         * config/rs6000/t-rs6000: Ditto.
23436         * config/rs6000/t-xilinx: Delete.
23437         * config/rs6000/titan.md: Adjust for fp_type removal.
23438         * config/rs6000/vsx.md: Remove Xilinx FP support.
23439         (VStype_simple): Delete.
23440         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
23441         * config/rs6000/xfpu.h: Delete.
23442         * config/rs6000/xfpu.md: Delete.
23443         * config/rs6000/xilinx.h: Delete.
23444         * config/rs6000/xilinx.opt: Delete.
23445         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
23446         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
23448 2018-05-04  Tom de Vries  <tom@codesourcery.com>
23450         PR libgomp/85639
23451         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
23452         if ignore == 0.
23454 2018-05-04  Richard Biener  <rguenther@suse.de>
23456         PR middle-end/85627
23457         * tree-complex.c (update_complex_assignment): We are always in SSA form.
23458         (expand_complex_div_wide): Likewise.
23459         (expand_complex_operations_1): Likewise.
23460         (expand_complex_libcall): Preserve EH info of the original stmt.
23461         (tree_lower_complex): Handle removed blocks.
23462         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
23463         on complex multiplication and division libcall builtins.
23465 2018-05-04  Richard Biener  <rguenther@suse.de>
23467         PR middle-end/85574
23468         * fold-const.c (negate_expr_p): Restrict negation of operand
23469         zero of a division to when we know that can happen without
23470         overflow.
23471         (fold_negate_expr_1): Likewise.
23473 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
23475         PR libstdc++/85466
23476         * real.h (real_nextafter): Declare.
23477         * real.c (real_nextafter): New function.
23478         * fold-const-call.c (fold_const_nextafter): New function.
23479         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
23480         CASE_CFN_NEXTTOWARD.
23481         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
23482         even when arg1_mode is different from arg0_mode.
23484 2018-05-03  Nathan Sidwell  <nathan@acm.org>
23486         * doc/extend.texi (Deprecated Features): Remove
23487         -ffriend-injection.
23488         (Backwards Compatibility): Likewise.
23489         * doc/invoke.texi (C++ Language Options): Likewise.
23490         (C++ Dialect Options): Likewise.
23492 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
23494         PR target/85530
23495         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
23496         _mm512_mask_mullox_epi64): New intrinsics.
23498 2018-05-03  Tom de Vries  <tom@codesourcery.com>
23500         PR testsuite/85106
23501         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
23502         dump files): Add offload-tree.
23504 2018-05-03  Richard Biener  <rguenther@suse.de>
23506         PR tree-optimization/85615
23507         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
23508         to loops not nested in BBs loop father to avoid creating multi-entry
23509         loops.
23511 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23513         PR tree-optimization/70291
23514         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
23515         arguments.  Change return type to tree.  Emit libcall as a new
23516         statement rather than replacing existing one when inplace_p is true.
23517         (expand_complex_multiplication_components): New function.
23518         (expand_complex_multiplication): Expand floating-point complex
23519         multiplication using the above.
23520         (expand_complex_division): Rename inner_type parameter to type.
23521         Update expand_complex_libcall call-site.
23522         (expand_complex_operations_1): Update expand_complex_multiplication
23523         and expand_complex_division call-sites.
23525 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
23527         PR target/85582
23528         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
23529         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
23530         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
23531         the highest significant bit of the shift count mask is clear.  In
23532         check whether and[sq]i3 is needed verify that all significant bits
23533         of the shift count other than the highest are set.
23535 2018-05-02  Tom de Vries  <tom@codesourcery.com>
23537         PR libgomp/82428
23538         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
23539         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
23540         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
23541         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
23542         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
23543         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
23544         __builtin_goacc_parlevel_size.
23546 2018-05-02  Richard Biener  <rguenther@suse.de>
23548         PR tree-optimization/85597
23549         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
23550         do not use split vect_get_vec_defs call but call vect_get_slp_defs
23551         directly.
23553 2018-05-02  Tom de Vries  <tom@codesourcery.com>
23555         PR testsuite/85106
23556         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
23557         dump files): Add ltrans-tree.
23559 2018-05-02  Tom de Vries  <tom@codesourcery.com>
23561         PR testsuite/85106
23562         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
23563         dump files): Add wpa-ipa.
23565 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
23567         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
23568         powerpc*-*-linux*paired* target.
23569         * config/rs6000/750cl.h: Delete.
23570         * config/rs6000/paired.h: Delete.
23571         * config/rs6000/paired.md: Delete.
23572         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
23573         float support.
23574         * config/rs6000/rs6000-builtin.def: Remove paired float support.
23575         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
23576         comment.  Remove paired float support.
23577         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
23578         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
23579         VECTOR_PAIRED.
23580         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
23581         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
23582         declarations.
23583         * config/rs6000/rs6000.c: Remove paired float support.
23584         (paired_expand_vector_init, paired_expand_vector_move,
23585         paired_emit_vector_compare, paired_emit_vector_cond_expr,
23586         (paired_expand_lv_builtin, paired_expand_stv_builtin,
23587         paired_expand_builtin, paired_expand_predicate_builtin,
23588         paired_init_builtins): Delete.
23589         * config/rs6000/rs6000.h: Remove paired float support.
23590         * config/rs6000/rs6000.md: Remove paired float support.
23591         (move_from_CR_ov_bit): Delete.
23592         * config/rs6000/rs6000.opt (mpaired): Delete.
23593         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
23594         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
23596 2018-05-02  Richard Biener  <rguenther@suse.de>
23598         PR middle-end/85567
23599         * gimplify.c (gimplify_save_expr): When in SSA form allow
23600         SAVE_EXPRs to compute to SSA vars.
23602 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
23604         PR target/85582
23605         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
23606         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
23607         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
23608         clobber operands[2], instead use a new pseudo.  Formatting fixes.
23610 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
23612         PR tree-optimization/85586
23613         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
23614         exit early for statements in the same group if the accesses are
23615         not strided.
23617 2018-05-02  Tom de Vries  <tom@codesourcery.com>
23619         PR lto/85451
23620         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
23621         error message.
23623 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
23625         PR tree-optimization/85143
23626         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
23628 2018-05-01  Tom de Vries  <tom@codesourcery.com>
23630         PR lto/85451
23631         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
23632         not found" error message.
23634 2018-05-01  Tom de Vries  <tom@codesourcery.com>
23636         PR other/83786
23637         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
23638         * vec.c (test_ordered_remove_if): New function.
23639         (vec_c_tests): Call test_ordered_remove_if.
23640         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
23641         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
23642         * tree-vect-patterns.c (vect_pattern_recog_1): Use
23643         VEC_ORDERED_REMOVE_IF.
23645 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23647         PR tree-optimization/82665
23648         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
23649         pointer subtraction where arguments come from a memchr call.
23651 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
23653         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
23654         --push-state --as-needed and --pop-state instead of --as-needed and
23655         --no-as-needed if ld supports it.
23656         * configure: Regenerated.
23658         PR web/85578
23659         * doc/install.texi2html: Replace _002d with - and _002a with * in
23660         generated html files using sed.
23662 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23664         PR c++/85523
23665         * gcc-rich-location.c (blank_line_before_p): New function.
23666         (use_new_line): New function.
23667         (gcc_rich_location::add_fixit_insert_formatted): New function.
23668         * gcc-rich-location.h
23669         (gcc_rich_location::add_fixit_insert_formatted): New function.
23671 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23673         * selftest.c (assert_streq): Rename "expected" and "actual" to
23674         "val1" and "val2".  Extend NULL-handling to cover both inputs
23675         symmetrically, while still requiring both to be non-NULL for a pass.
23676         * selftest.h (assert_streq): Rename "expected" and "actual" to
23677         "val1" and "val2".
23678         (ASSERT_EQ): Likewise.
23679         (ASSERT_EQ_AT): Likewise.
23680         (ASSERT_KNOWN_EQ): Likewise.
23681         (ASSERT_KNOWN_EQ_AT): Likewise.
23682         (ASSERT_NE): Likewise.
23683         (ASSERT_MAYBE_NE): Likewise.
23684         (ASSERT_MAYBE_NE_AT): Likewise.
23685         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
23686         the assertion to pass.
23687         (ASSERT_STREQ_AT): Likewise.
23689 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23691         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
23692         interaction with -pie.
23694 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23696         * selftest.h: Fix alphabetization of per-source-file selftest
23697         declarations.
23699 2018-04-30  Jason Merrill  <jason@redhat.com>
23701         PR c++/61982 - dead stores to destroyed objects.
23702         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
23703         of clobber.
23705 2018-04-30  Jason Merrill  <jason@redhat.com>
23707         * tree.c (build_clobber): New.
23708         * tree.h: Declare it.
23709         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
23711 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23713         * diagnostic-show-locus.c (layout::layout): Update for
23714         location_get_source_line returning a char_span.
23715         (struct char_span): Move to input.h.
23716         (struct correction): Update for fields in char_span becoming
23717         private.
23718         (struct source_line): Update for location_get_source_line
23719         returning a char_span.
23720         (layout::print_line): Likewise.
23721         * edit-context.c (edited_file::print_content): Likewise.
23722         (edited_file::print_diff_hunk): Likewise.
23723         (edited_file::print_run_of_changed_lines): Likewise.
23724         (edited_file::get_num_lines): Likewise.
23725         (edited_line::edited_line): Likewise.
23726         * final.c (asm_show_source): Likewise.
23727         * input.c (location_get_source_line): Convert return type
23728         from const char * to char_span, losing the final "line_len"
23729         param.
23730         (dump_location_info): Update for the above.
23731         (get_substring_ranges_for_loc): Likewise.  Use a char_span
23732         when handling the literal within the line.
23733         (test_reading_source_line): Update for location_get_source_line
23734         returning a char_span.
23735         * input.h (class char_span): Move here from
23736         diagnostic-show-locus.c, converting from a struct to a class.
23737         Make data members private.
23738         (char_span::operator bool): New.
23739         (char_span::length): New.
23740         (char_span::get_buffer): New.
23741         (char_span::operator[]): New.
23742         (char_span::subspan): Make const.
23743         (char_span::xstrdup): New.
23744         (location_get_source_line): Convert return type from const char *
23745         to char_span, losing the final "line_size" param.
23747 2018-04-30  Jan Hubicka  <jh@suse.cz>
23749         * lto-wrapper.c (ltrans_priorities): New static var.
23750         (cmp_priority): New.
23751         (run_gcc): Read priorities and if doing parallel build order
23752         the Makefile by them.
23754 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23756         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
23758 2018-04-30  Richard Biener  <rguenther@suse.de>
23760         * tree-cfg.c (verify_address): Remove base argument, add
23761         flag whether to check TREE_ADDRESSABLE and do that.
23762         (verify_expr): Remove.
23763         (verify_types_in_gimple_reference): Add pieces from verify_expr.
23764         (verify_gimple_assign_single): Likewise.
23765         (verify_gimple_switch): Likewise.
23766         (verify_expr_location_1): Dereference tp once.  Add (disabled)
23767         piece from verify_expr.
23768         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
23770 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23772         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
23774 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23776         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
23777         (small_data_pattern): Likewise.
23778         (arc_rewrite_small_data): Likewise.
23779         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
23780         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
23781         (get_symbol_alignment): New function.
23782         (legitimate_small_data_address_p): Likewise.
23783         (legitimate_scaled_address): Update, call
23784         legitimate_small_data_address_p.
23785         (output_sdata): New static variable.
23786         (arc_print_operand): Update how we handle small data operands.
23787         (arc_print_operand_address): Likewise.
23788         (arc_legitimate_address_p): Update, use
23789         legitimate_small_data_address_p.
23790         (arc_rewrite_small_data_p): Remove.
23791         (arc_rewrite_small_data_1): Likewise.
23792         (arc_rewrite_small_data): Likewise.
23793         (small_data_pattern): Likewise.
23794         (compact_sda_memory_operand): Update to use
23795         legitimate_small_data_address_p and get_symbol_alignment.
23796         (prepare_move_operands): Don't rewite sdata pattern.
23797         (prepare_extend_operands): Remove.
23798         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
23799         pattern.
23800         (zero_extendqisi2): Likewise.
23801         (zero_extendhisi2): Likewise.
23802         (extendqihi2): Likewise.
23803         (extendqisi2): Likewise.
23804         (extendhisi2): Likewise.
23805         (addsi3): Likewise.
23806         (subsi3): Likewise.
23807         (andsi3): Likewise.
23808         * config/arc/constraints.md (Usd): Change it to memory constraint.
23810 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23812         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
23813         as source of std instructions.
23814         * config/arc/arc.md (movsi_insn): Update pattern predicate to
23815         allow 6-bit constants as source for store instructions.
23816         (movdi_insn): Update instruction pattern to allow 6-bit constants
23817         as source for store instructions.
23819 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23821         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
23823 2018-04-30  Nathan Sidwell  <nathan@acm.org>
23824             Sandra Loosemore <sandra@codesourcery.com>
23826         * dumpfile.c (dump_open): Allow '-' for stdout.
23827         * doc/invoke.texi (Developer Options): Document dump filename
23828         determination early.  Document stdin/stdout selection.
23830 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
23832         Microblaze Target: PIC data text relative
23834         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
23835         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
23836         Add declaration.
23837         * config/microblaze/microblaze.h (microblaze_constant_address_p):
23838         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
23839         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
23840         New addressing mode for data-text relative position indepenedent code.
23841         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
23842         'ADDRESS_SYMBOLIC_TXT_REL'.
23843         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
23844         (microblaze_legitimate_pic_operand): Exclude function calls from
23845         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
23846         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
23847         addresses cases.
23848         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23849         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23850         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
23851         for 'address + offset'.
23852         (microblaze_expand_prologue): Add new function prologue call for
23853         'r20' assignation.
23854         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
23855         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
23856         table in case of TARGET_PIC_DATA_TEXT_REL.
23857         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
23858         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
23859         Add new macros 'UNSPEC_TEXT',
23860         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
23861         + exclude function calls from 'UNSPEC_PLT' in case of data text
23862         relative mode.
23863         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23864         new target hook for generating address diff vector tables in case of
23865         flag_pic.
23866         * doc/tm.texi : Regenerate.
23867         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
23868         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
23869         of addr diff vector generation.
23870         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23871         target hook definition.
23872         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
23873         Add default function for generate_pic_addr_diff_vec -> flag_pic.
23874         * doc/invoke.texi (Add new pic option): Add new microblaze pic
23875         option for data text relative.
23877 2018-04-30  Richard Biener  <rguenther@suse.de>
23879         * tree-chrec.h (evolution_function_is_constant_p): Remove
23880         redundant check.
23881         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
23883 2018-04-30  Richard Biener  <rguenther@suse.de>
23885         PR bootstrap/85571
23886         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
23888 2018-04-30  Richard Biener  <rguenther@suse.de>
23890         PR tree-optimization/28364
23891         PR tree-optimization/85275
23892         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
23893         copying first exit test.
23895 2018-04-28  Mark Wielaard  <mark@klomp.org>
23897         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
23898         dwarf_version >= 5.
23899         (dwarf_AT): Handle DW_AT_addr_base.
23900         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
23902 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
23904         PR target/84431
23905         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
23906         (*ashl<dwi>3_doubleword_mask_1): Ditto.
23907         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
23908         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
23910 2018-04-28  Richard Biener  <rguenther@suse.de>
23912         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
23913         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
23914         to reflect use.  Only add interesting stmts.
23916 2018-04-27  Martin Jambor  <mjambor@suse.cz>
23918         PR ipa/85549
23919         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
23920         the jump function allows for passing through aggregate values.
23922 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
23924         * input.h (in_system_header_at): Convert from macro to inline
23925         function.
23926         (from_macro_expansion_at): Likewise.
23927         (from_macro_definition_at): Likewise.
23929 2018-04-27  Jeff Law  <law@redhat.com>
23931         * config.gcc: Mark tile* targets as deprecated/obsolete.
23933 2018-04-27  Richard Biener  <rguenther@suse.de>
23935         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
23936         fix for ILP32.
23938 2018-04-27  Richard Biener  <rguenther@suse.de>
23940         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
23942 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
23944         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
23945         with Yd constraint. Set "preferred_for_speed" attribute from
23946         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
23947         with Yd constraint.
23948         (*movdi_internal): Ditto.
23949         (movti_interunit splitters): Remove
23950         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
23951         (movdi_interunit splitters): Ditto.
23952         * config/i386/constraints.md (Ye): Remove.
23953         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
23955 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23957         PR target/85512
23958         * config/aarch64/constraints.md (Usg): Limit to 31.
23959         (Usj): Limit to 63.
23961 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
23963         PR tree-optimization/85529
23964         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
23965         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
23966         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
23967         zero extension or masking of the MSB bit.
23968         (optimize_range_tests): Add FIRST_BB argument, pass it through
23969         to optimize_range_tests_var_bound.
23970         (maybe_optimize_range_tests, reassociate_bb): Adjust
23971         optimize_range_tests callers.
23973 2018-04-26  Richard Biener  <rguenther@suse.de>
23974             Jakub Jelinek  <jakub@redhat.com>
23976         * cgraph.h (symbol_table): Just declare debug method here.
23977         * symtab.c (symbol_table::debug): Define.
23979 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
23981         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
23983 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
23985         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
23986         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
23987         (*movdi_internal): Substitute Yi and Yj constraint with x
23988         and Ym and Yn constraint with y constraint.  Update "isa"
23989         attribute and set "preferred_for_speed" attribute from
23990         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
23991         (*movsi_internal): Ditto.
23992         (*movdf_internal): Ditto.
23993         (*movsf_internal): Ditto.
23994         (*zero_extendsidi2): Ditto.
23995         * config/i386/sse.md (vec_set<mode>_0): Ditto.
23996         (sse2_loadld): Ditto.
23997         (*vec_extract<ssevecmodelower>_0): Ditto.
23998         (*vec_extractv4si_0_zext_sse4): Ditto.
23999         (vec_concatv2di): Ditto.
24000         (*vec_dup<mode>): Ditto.
24001         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
24002         * config/i386/constraints.md (Yi): Remove.
24003         (Yj): Remove.
24004         (Ym): Remove.
24005         (Yn): Remove.
24007 2018-04-26  Nathan Sidwell  <nathan@acm.org>
24009         * dumpfile.c (dump_open): New.
24010         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
24011         (dump_finish): Detect stdio/stderr by value not name.
24013 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
24015         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
24017 2018-04-26  Tom de Vries  <tom@codesourcery.com>
24019         PR target/84952
24020         * config/nvptx/nvptx.c (verify_neutering_jumps)
24021         (verify_neutering_labels): New function
24022         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
24024 2018-04-26  Tom de Vries  <tom@codesourcery.com>
24026         PR target/84025
24027         * config/nvptx/nvptx.c (needs_neutering_p): New function.
24028         (nvptx_single): Use needs_neutering_p to skip over insns that do not
24029         need neutering.
24031 2018-04-26  Richard Biener <rguenther@suse.de>
24032             Tom de Vries  <tom@codesourcery.com>
24034         PR lto/85422
24035         * lto-streamer-out.c (output_function): Fixup loops if required to match
24036         discovery done in the reader.
24038 2018-04-26  Richard Biener  <rguenther@suse.de>
24040         PR tree-optimization/85116
24041         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
24042         have a loop exit from the single latch predecessor.  Remove
24043         case of header with just condition.
24044         (ch_base::copy_headers): Exclude infinite loops from any
24045         processing.
24046         (pass_ch::execute): Record exits.
24048 2018-04-26  Richard Biener  <rguenther@suse.de>
24050         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
24051         prologue cost vector and pass it to vect_get_load_cost.
24052         (vect_get_peeling_costs_all_drs): Likewise.
24053         (vect_peeling_hash_get_lowest_cost): Likewise.
24054         (vect_enhance_data_refs_alignment): Likewise.
24056 2018-04-26  Richard Biener  <rguenther@suse.de>
24058         PR middle-end/85450
24059         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
24060         checking of integer<->pointer conversions.
24061         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
24062         sign-/zero-extending pointer types.
24063         (expand_omp_for_static_chunk): Likewise.
24065 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
24066             Jean Lee  <xiaoyur347@gmail.com>
24068         * config/mips/mips.c (mips_asan_shadow_offset): New function.
24069         (TARGET_ASAN_SHADOW_OFFSET): Define.
24070         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
24071         true for -fsanitize=address.
24073 2018-04-25  Mark Wielaard  <mark@klomp.org>
24075         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
24076         shorter ones.
24078 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
24080         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
24081         than "alu", remove explicit "memory" and "imm_disp" attributes.
24082         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
24084         PR middle-end/85414
24085         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
24086         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
24087         gen_lowpart_no_emit.
24089 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
24091         PR target/85473
24092         * config/i386/i386.c (ix86_expand_builtin): Change memory
24093         operand to XI, extend p0 to Pmode.
24094         * config/i386/i386.md: Change unspec volatile and operand
24095         1 mode to XI, change operand 0 mode to P.
24097 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24099         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
24100         GET_MODE_MASK before any checking.
24101         (nds32_can_use_bset_p): Likewise.
24102         (nds32_can_use_btgl_p): Likewise.
24104 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24106         * config/nds32/nds32-doubleword.md: New define_split pattern for
24107         illegal register number.
24109 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24111         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
24113 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
24115         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
24117 2018-04-25  Richard Biener  <rguenther@suse.de>
24119         * lto-streamer.h (LTO_major_version): Bump to 8.
24121 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
24123         * BASE-VER: Set to 9.0.0.
24125 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
24127         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
24128         in __abskf2 and __powikf2.
24130 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24132         PR target/85512
24133         * config/aarch64/constraints.md (Usg, Usj): New constraints.
24134         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
24135         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
24136         Use the above on operand 2.  Reindent.
24137         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
24139 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
24141         PR target/85485
24142         * common/config/i386/i386-common.c (ix86_handle_option): Don't
24143         handle OPT_mcet.
24144         * config/i386/i386.opt (mcet): Removed.
24145         * doc/install.texi: Remove -mcet documentation.
24146         * doc/invoke.texi: Likewise.
24148 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
24150         PR target/85485
24151         * doc/install.texi: Remove -mcet from bootstrap-cet.
24153 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24155         PR target/85511
24156         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
24157         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
24158         if TARGET_64BIT.
24160         PR target/85503
24161         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
24162         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
24163         containing a CONST_VECTOR.
24165 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
24167         * doc/install.texi: Update newlib dependency for nvptx.
24169 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24171         PR target/85508
24172         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
24173         instead of INTVAL when shifting x left.
24175 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
24177         PR tree-optimization/85478
24178         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
24179         vect_grouped_store_supported for single element vectors.
24181 2018-04-24  Richard Biener  <rguenther@suse.de>
24183         PR target/85491
24184         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
24185         load cost increase to the case of non-constant step.
24187 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
24189         PR target/84828
24190         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
24191         destination if any_malformed_asm.
24193 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
24195         PR middle-end/85496
24196         * expr.c (store_field): In the bitfield case, if the value comes from
24197         a function call and is returned in registers by means of a PARALLEL,
24198         do not change the mode of the temporary unless BLKmode and VOIDmode.
24200 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
24202         PR rtl-optimization/85423
24203         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
24204         dependencies to debug insns when the previous insn is non-debug.
24206 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
24208         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
24209         enums into a single definition.
24210         (fls): Fix predicates and printing.
24211         (seti): Likewise.
24213 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
24215         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
24216         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
24217         and short u6 immediate.
24218         (check_if_valid_sleep_operand): Remove.
24219         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
24221 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
24223         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
24224         flag_always_save_lp condition.
24225         * config/nds32/nds32.opt (malways-save-lp): New option.
24227 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
24229         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
24230         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
24231         * config/nds32/nds32.h
24232         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
24233         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
24235 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
24237         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
24238         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
24240 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
24241             Chung-Ju Wu  <jasonwucj@gmail.com>
24243         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
24244         Declare.
24245         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
24246         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
24248 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
24250         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
24252 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
24254         * config/nds32/nds32-protos.h (nds32_data_alignment,
24255         nds32_local_alignment): Declare.
24256         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
24257         nds32_local_alignment): New functions.
24258         (TARGET_CONSTANT_ALIGNMENT): Define.
24259         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
24261 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
24263         * config/nds32/nds32.c
24264         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
24265         (TARGET_MODES_TIEABLE_P): Likewise.
24267 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
24269         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
24270         level Ofast and Og.
24272 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
24273             Chung-Ju Wu  <jasonwucj@gmail.com>
24275         * config/nds32/constants.md (unspec_volatile_element): Add enum values
24276         for unaligned access.
24277         * config/nds32/nds32-intrinsic.c: Implementation of expanding
24278         unaligned access.
24279         * config/nds32/nds32-intrinsic.md: Likewise.
24280         * config/nds32/nds32_intrinsic.h: Likewise.
24281         * config/nds32/nds32.h (nds32_builtins): Likewise.
24282         * config/nds32/nds32.opt (munaligned-access): New option.
24283         * config/nds32/nds32.c (nds32_asm_file_start): Display
24284         flag_unaligned_access status.
24286 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
24288         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
24289         -mno-relax is present.
24290         * config/riscv/linux.h (LINK_SPEC): Ditto.
24292 2018-04-20  Martin Sebor  <msebor@redhat.com>
24294         PR c/85365
24295         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
24296         for null pointers.
24297         (gimple_fold_builtin_stxcpy_chk): Same.
24298         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
24300 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
24302         PR target/85456
24303         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
24304         __powikf2 when long double is IEEE 128-bit.
24306 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
24308         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
24309         step to make sure stack always aligned.
24311 2018-04-20  Carl Love  <cel@us.ibm.com>
24313         PR target/83402
24314         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
24315         size check for arg0.
24317 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
24318             Tom de Vries  <tom@codesourcery.com>
24320         PR target/85445
24321         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
24322         Emit insns for calls too.
24323         (nvptx_find_par): Always look for worker-level predecessor insn.
24324         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
24325         calls.
24326         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
24327         (nvptx_process_pars): Propagate frames for calls.
24329 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
24331         PR target/85469
24332         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
24333         Removed.
24334         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
24335         (ix86_handle_option): Don't handle OPT_mibt.
24336         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
24337         __SHSTK__.
24338         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
24339         has_ibt and ibt.
24340         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
24341         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
24342         (ix86_target_macros): Define __CET__ with flag_cf_protection
24343         for -fcf-protection.
24344         * config/i386/i386.c (isa2_opts): Remove -mibt.
24345         * config/i386/i386.h (TARGET_IBT): Removed.
24346         (TARGET_IBT_P): Likewise.
24347         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
24348         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
24349         * config/i386/i386.opt (mcet): Update help message.
24350         (mshstk): Likewise.
24351         (mibt): Removed.
24352         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
24353         -mcet as an alias for -mshstk.
24355 2018-04-20  Richard Biener <rguenther@suse.de>
24357         PR middle-end/85475
24358         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
24359         complexity by forcing a single use of the multiply operand.
24361 2018-04-20  Martin Jambor  <mjambor@suse.cz>
24363         ipa/85449
24364         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
24365         recursion dependency to only apply to non-clones.
24367 2018-04-20  Martin Jambor  <mjambor@suse.cz>
24369         ipa/85447
24370         * ipa-cp.c (create_specialized_node): Check that clones of
24371         self-recursive edges exist during IPA-CP.
24373 2018-04-19  Toon Moene  <toon@moene.org>
24375         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
24376         by -O3.
24378 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
24380         PR tree-optimization/85467
24381         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
24382         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
24383         VECTOR_CST element to type.
24385 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
24387         PR target/85397
24388         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
24389         * config/i386/i386.md (builtin_setjmp_setup): Removed.
24390         (builtin_longjmp): Likewise.
24391         (save_stack_nonlocal): New pattern.
24392         (restore_stack_nonlocal): Likewise.
24394 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
24396         PR target/85404
24397         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24398         Replace ASM_OUTPUT_LABEL with fprintf.
24400 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
24402         PR target/85417
24403         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24404         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
24405         * config/i386/i386-c.c (ix86_target_macros_internal): Also
24406         define __IBT__ and __SHSTK__ for -fcf-protection.
24407         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
24408         TARGET_IBT.
24409         (ix86_trampoline_init): Likewise.
24410         (x86_output_mi_thunk): Likewise.
24411         (ix86_notrack_prefixed_insn_p): Likewise.
24412         (ix86_option_override_internal): Don't disallow -fcf-protection.
24413         * config/i386/i386.md (rdssp<mode>): Also enable for
24414         -fcf-protection.
24415         (incssp<mode>): Likewise.
24416         (nop_endbr): Likewise.
24417         * config/i386/i386.opt (mcet): Change help message to built-in
24418         functions only.
24419         (mibt): Likewise.
24420         (mshstk): Likewise.
24421         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
24422         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
24423         enable CET built-in functions.
24425 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
24427         * common/config/i386/i386-common.c
24428         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
24429         OPTION_MASK_ISA_MOVDIRI_UNSET,
24430         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
24431         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
24432         * config.gcc (movdirintrin.h): New header.
24433         * config/i386/cpuid.h (bit_MOVDIRI,
24434         bit_MOVDIR64B): New bits.
24435         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
24436         and -mmvodir64b.
24437         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
24438         (VOID, PVOID, PCVOID)): New function types.
24439         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
24440         __builtin_ia32_directstoreu_u64,
24441         __builtin_ia32_movdir64b): New builtins.
24442         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
24443         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
24444         and -mmovdiri.
24445         (ix86_valid_target_attribute_inner_p): Ditto.
24446         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
24447         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
24448         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
24449         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
24450         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
24451         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
24452         (movdiri<mode>, movdir64b_<mode>): New.
24453         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
24454         * config/i386/immintrin.h: Include movdirintrin.h.
24455         * config/i386/movdirintrin.h: New file.
24456         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
24458 2018-04-19  Richard Biener  <rguenther@suse.de>
24460         PR middle-end/85455
24461         * cfg.c (clear_bb_flags): When loop state says we have
24462         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
24464 2018-04-19  Richard Biener  <rguenther@suse.de>
24466         PR tree-optimization/84737
24467         * tree-vect-data-refs.c (vect_copy_ref_info): New function
24468         copying restrict info.
24469         (vect_setup_realignment): Use it.
24470         * tree-vectorizer.h (vect_copy_ref_info): Declare.
24471         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
24472         the first DR to all generated stores.
24473         (vectorizable_load): Likewise for loads.
24475 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
24477         PR tree-optimization/85446
24478         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
24479         the integral and pointer types to have the same precision.
24481         * doc/install.texi: Document --disable-cet being the default and
24482         --enable-cet=auto.
24484 2018-04-18  Martin Liska  <mliska@suse.cz>
24486         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
24487         style.
24489 2018-04-18  Martin Liska  <mliska@suse.cz>
24491         Revert
24492         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24494         PR ipa/83983
24495         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24496         arguments if they are comparable.
24498 2018-04-18  Martin Liska  <mliska@suse.cz>
24500         Revert
24501         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
24503         PR lto/84805
24504         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
24505         incomplete types.
24507 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
24509         PR target/85388
24510         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
24511         ENDBR after calling __morestack.
24513 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
24515         PR jit/85384
24516         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
24517         by using gcc_base_ver to generate a gcc_driver_version, and use
24518         it when generating GCC_DRIVER_NAME.
24519         * configure: Regenerate.
24521 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
24523         PR target/81084
24524         * config.gcc: Obsolete powerpc*-*-*spe*.
24526 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
24528         PR debug/84637
24529         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
24530         (stabstr_D): Change type of unum from unsigned int to
24531         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
24532         type.
24534 2018-04-17  Jim Wilson  <jimw@sifive.com>
24536         PR 84856
24537         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
24538         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
24539         Set arg_pointer_offset after using pretend_args_size.
24541 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
24543         PR rtl-optimization/85431
24544         * dse.c (record_store): Ignore zero width stores.
24546         PR sanitizer/85230
24547         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
24548         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
24549         __builtin_stack_restore rather than after it.
24550         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
24551         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
24552         argument instead of virtual_dynamic_stack_rtx.
24554 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24556         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
24557         New prototype.
24558         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24559         Add note to error message to explain internal mapping of overloaded
24560         built-in function name to non-overloaded built-in function name.
24561         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
24562         function.
24564 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
24566         PR target/85424
24567         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
24568         where the inputs overlap with the output.
24570 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
24572         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
24573         (=v, v) alternative and explicit "memory" attribute.
24574         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
24575         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
24576         attributes.
24577         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
24578         "sselog1" type instead of "sselog".
24579         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
24580         "sselog".  Remove explicit "memory" attribute.
24581         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
24582         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
24583         attributes.
24584         (vec_extract_hi_v32hi): Merge all alternatives into one, use
24585         "sselog1" type instead of "sselog".  Remove explicit "memory"
24586         attribute.
24587         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
24588         use "sselog1" type instead of "sselog".  Remove explicit "memory"
24589         attribute.
24590         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
24591         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
24592         attributes.
24593         (vec_extract_hi_v64qi): Merge all alternatives into one, use
24594         "sselog1" type instead of "sselog".  Remove explicit "memory"
24595         attribute.
24596         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
24597         use "sselog1" type instead of "sselog".  Remove explicit "memory"
24598         attribute.
24600         PR target/85430
24601         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
24603         PR middle-end/85414
24604         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
24605         on a SUBREG.
24607 2018-04-17  Martin Jambor  <mjambor@suse.cz>
24609         PR ipa/85421
24610         * ipa-cp.c (create_specialized_node): Call
24611         expand_all_artificial_thunks if necessary.
24613 2018-04-17  Martin Liska  <mliska@suse.cz>
24615         PR lto/85405
24616         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
24617         in message, remote space in between '_G' and '('.
24619 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
24621         PR target/85281
24622         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
24623         avx512f_vmcmp<mode>3<round_saeonly_name>,
24624         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
24625         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
24626         avx512f_rndscale<mode><round_saeonly_name>,
24627         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
24628         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
24629         Use %<iptr>2 instead of %2 for -masm=intel.
24630         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
24631         avx512f_vcvttss2usi<round_saeonly_name>,
24632         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
24633         -masm=intel.
24634         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
24635         avx512f_vcvttsd2usi<round_saeonly_name>,
24636         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
24637         Use %q1 instead of %1 for -masm=intel.
24638         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
24639         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
24640         of %3 for -masm=intel.
24641         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
24642         -masm=intel.
24643         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
24644         -masm=intel.
24645         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
24646         -masm=intel.
24647         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
24648         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
24649         %g1.
24650         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
24651         -masm=intel.
24652         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
24653         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
24654         %g1 and one with %0 and %1.
24655         (avx512er_vmrcp28<mode><round_saeonly_name>,
24656         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
24657         %1 for -masm=intel.
24658         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
24659         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
24660         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
24661         of %0 and %{%4%} for -masm=intel.
24662         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
24663         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
24664         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
24665         order of %0 and %{%5%}%{z%} for -masm=intel.
24667 2018-04-17  Jan Hubicka  <jh@suse.cz>
24669         PR lto/85405
24670         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
24672 2018-04-17  Martin Liska  <mliska@suse.cz>
24674         PR ipa/85329
24675         * multiple_target.c (create_dispatcher_calls): Set apostrophes
24676         for target_clone error message.  Make default implementation
24677         clone to be a local declaration.
24678         (separate_attrs): Add new argument and check for an empty
24679         string.
24680         (expand_target_clones): Handle it.
24681         (ipa_target_clone): Make redirection just for target_clones
24682         functions.
24684 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
24685             Tom de Vries  <tom@codesourcery.com>
24687         PR middle-end/84955
24688         * omp-expand.c (expand_oacc_for): Add dummy false branch for
24689         tiled basic blocks without omp continue statements.
24691 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
24693         PR target/83660
24694         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
24695         vec_extract expression as having side effects to make sure it gets
24696         a cleanup point.
24698 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
24700         PR target/85403
24701         * config/i386/i386.c (get_builtin_code_for_version): Check
24702         error_mark_node.
24704 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
24706         PR target/84331
24707         * config.gcc: Support "skylake".
24708         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
24709         PROCESSOR_SKYLAKE.
24710         * config/i386/i386.c (m_SKYLAKE): Define.
24711         (processor_target_table): Add "skylake".
24712         (ix86_option_override_internal): Add "skylake".
24713         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
24714         PROCESSOR_CANNONLAKE.
24715         (get_builtin_code_for_version): Fix priority for
24716         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
24717         PROCESSOR_SKYLAKE-AVX512.
24718         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
24719         (processor_type): Add PROCESSOR_SKYLAKE.
24721 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
24722             Jason Merrill  <jason@redhat.com>
24724         PR c++/85112
24725         * convert.c (convert_to_integer_1): Use direct recursion for
24726         enumeral types and types with a precision less than the number
24727         of bits in their mode.
24729 2018-04-16  Julia Koval  <julia.koval@intel.com>
24731         PR target/84413
24732         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
24733         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
24735 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
24737         PR target/85293
24738         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
24739         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
24740         and -mno-direct-move.
24742 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
24744         PR target/83402
24745         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
24746         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
24747         Ensure negative shifts result in {0}.
24749 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
24751         PR rtl-optimization/79916
24752         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
24753         regs (if any) to define how to gnerate SD moves when LRA is in
24754         progress.
24756 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
24758         PR rtl-optimization/85393
24759         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
24760         * except.c (expand_dw2_landing_pad_for_region): Make static.
24761         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
24762         a label and unconditional jump to old_bb, rather than
24763         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
24764         basic block.
24766         PR rtl-optimization/85376
24767         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
24768         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
24769         instead of a specific value.
24771 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24772             Bin Cheng  <bin.cheng@arm.com>
24774         PR tree-optimization/82965
24775         PR tree-optimization/83991
24776         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
24777         by_profile_only parameter.
24778         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
24779         information if the loop was predicted to iterate too many times.
24780         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
24782 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24784         PR lto/71991
24785         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
24786         always inline.
24788 2018-04-13  Martin Liska  <mliska@suse.cz>
24789             Jakub Jelinek  <jakub@redhat.com>
24791         PR middle-end/81657
24792         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
24793         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
24794         * builtins.c (expand_builtin_memory_copy_args): Use
24795         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
24796         handle dest_addr == pc_rtx.
24798 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
24800         PR target/85291
24801         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
24802         asked to not generate direct moves.
24803         (fix_trunc<mode>si2_stfiwx): Similar.
24804         (fix_trunc<mode>si2_internal): Similar.
24806 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24808         PR debug/83157
24809         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
24810         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
24811         lookup if dest in some wider mode is known to be const0_rtx and
24812         if so, record permanent equivalence for it to be ZERO_EXTEND of
24813         the narrower mode destination.
24815 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24817         * lto-streamer-out.c (output_function): Revert 259346.
24818         * omp-expand.c (expand_oacc_for): Likewise.
24820 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
24822         PR rtl-optimization/85354
24823         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
24824         * sel-sched.c (sel_global_init): ... here.
24826 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
24828         PR target/85238
24829         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
24830         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
24831         mode for PE-COFF targets.
24832         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
24833         (i386_pe_asm_lto_end): Likewise.
24834         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
24835         (TARGET_ASM_LTO_END): Likewise.
24836         * config/i386/winnt.c (saved_debug_info_level): New static variable.
24837         (i386_pe_asm_lto_start): New function.
24838         (i386_pe_asm_lto_end): Likewise.
24840 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24841             Richard Biener  <rguenther@suse.de>
24843         PR middle-end/84955
24844         * lto-streamer-out.c (output_function): Fix CFG loop state before
24845         streaming out.
24846         * omp-expand.c (expand_oacc_for): Handle calls to internal
24847         functions like regular functions.
24849 2018-04-12  Richard Biener  <rguenther@suse.de>
24851         PR lto/85371
24852         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
24853         for the early LTO debug to properly generate references to it
24854         during DIE emission.  Do not re-use that for the skeleton for
24855         split-dwarf.
24856         (dwarf2out_early_finish): Likewise.
24858 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24860         PR target/85328
24861         * config/i386/sse.md
24862         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
24863         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
24864         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
24865         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
24866         and output is a reg, avoid creating invalid lowpart subreg, but
24867         instead split into a 512-bit move.  Don't split if not AVX512VL,
24868         input is xmm16+ reg and output is a mem.
24869         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
24870         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
24871         xmm16+ reg and output is a mem.
24873 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24875         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
24876         also for flag_dwarf2_cfi_asm.
24878 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24880         PR rtl-optimization/85342
24881         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
24882         a bool scalar var inside of the loop instead.  Don't try to update
24883         recog_data.operand after failed apply_change_group.
24885 2018-04-12  Tom de Vries  <tom@codesourcery.com>
24887         PR target/85296
24888         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
24889         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
24890         array with flexible array member as array without given dimension.
24891         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
24892         argument for undefined param to true.
24894 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
24896         PR target/85321
24897         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24898         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
24899         from PowerPC section.
24900         * config/rs6000/sysv4.opt (mcall-): Improve help text.
24901         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
24902         help text that is too long.
24903         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
24904         help text that is too long.
24905         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
24906         help text that is too long.
24908 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
24910         * config/alpha/alpha.md (stack_probe_internal): Rename
24911         from "probe_stack".  Update all callers.
24913 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24915         PR rtl-optimization/84566
24916         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
24917         sched_macro_fuse_insns.
24919 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24921         PR target/84301
24922         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
24923         (compute_block_dependences): ... from here.
24925 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24927         PR tree-optimization/85331
24928         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
24929         from int to HOST_WIDE_INT.
24931 2018-04-11  Martin Jambor  <mjambor@suse.cz>
24933         PR ipa/84149
24934         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
24935         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
24936         not the same as the source val.
24937         (cgraph_edge_brings_value_p): New parameter.
24938         (gather_edges_for_value): Pass destination value to
24939         cgraph_edge_brings_value_p.
24940         (perhaps_add_new_callers): Likewise.
24941         (get_info_about_necessary_edges): Likewise and exclude values brought
24942         only by self-recursive edges.
24943         (create_specialized_node): Redirect only clones of self-calling edges.
24944         (+self_recursive_pass_through_p): New function.
24945         (find_more_scalar_values_for_callers_subset): Use it.
24946         (find_aggregate_values_for_callers_subset): Likewise.
24947         (known_aggs_to_agg_replacement_list): Removed.
24948         (decide_whether_version_node): Re-calculate known constants for all
24949         remaining context clones.
24951 2018-04-11  Richard Biener  <rguenther@suse.de>
24953         PR lto/85339
24954         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
24955         from early DWARF output.
24956         (dwarf2out_early_finish): Output line info unconditionally into
24957         early DWARF and add reference to it.
24959 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24961         PR target/85281
24962         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
24963         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
24964         other than V2DFmode using iptr mode attribute.
24965         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
24967 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24969         PR rtl-optimization/84659
24970         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
24972 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24974         PR debug/85302
24975         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
24976         SIZEP is NULL.
24977         (output_loc_list): Pass address of a dummy size variable even in the
24978         locview handling loop.
24979         (index_location_lists): Add comment on why skip_loc_list_entry can't
24980         call size_of_locs.
24982 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24984         PR target/85261
24985         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
24986         into register.
24988 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
24990         PR target/85321
24991         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24992         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
24993         and -mstring-compare-inline-limit.
24995 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24997         PR target/85287
24998         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
24999         for stack clash protection in a register whenever we need it to be in
25000         a register.
25002 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
25004         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
25005         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
25007 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
25009         PR target/85321
25010         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
25011         the help text.
25012         (mlong-double-): Ditto.
25013         * config/rs6000/sysv4.opt (msdata=): Ditto.
25014         (mtls-size=): Ditto.
25016 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25018         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
25019         erroneous entries for
25020         "vector int vec_ldl (int, long int *)", and
25021         "vector unsigned int vec_ldl (int, unsigned long int *)".
25022         Add comments and entries for
25023         "vector bool char vec_ldl (int, bool char *)",
25024         "vector bool short vec_ldl (int, bool short *)",
25025         "vector bool int vec_ldl (int, bool int *)",
25026         "vector bool long long vec_ldl (int, bool long long *)",
25027         "vector pixel vec_ldl (int, pixel *)",
25028         "vector long long vec_ldl (int, long long *)",
25029         "vector unsigned long long vec_ldl (int, unsigned long long *)".
25030         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
25031         type tree bool_long_long_type_node and correct definition of
25032         bool_V2DI_type_node to make reference to this new type tree.
25033         (rs6000_mangle_type): Replace erroneous reference to
25034         bool_long_type_node with bool_long_long_type_node.
25035         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
25036         comments to emphasize sign distinctions for char and int types and
25037         replace RS6000_BTI_bool_long constant with
25038         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
25039         use of RS6000_BTI_pixel.
25040         (bool_long_type_node): Remove this macro definition.
25041         (bool_long_long_type_node): New macro definition
25043 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
25045         PR rtl-optimization/85300
25046         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
25047         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
25048         simplify_unary_operation fails.
25050 2018-04-10  Martin Liska  <mliska@suse.cz>
25052         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
25053         cgraph_edge and ipa_ref.
25055 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
25057         PR target/85177
25058         PR target/85255
25059         * config/i386/sse.md
25060         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
25061         computation of the VEC_MERGE selector from mask.
25062         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
25063         Fix decoding of the VEC_MERGE selector into mask.
25065 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
25067         PR tree-optimization/85286
25068         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
25070 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
25072         * final.c (final_1): Set insn_last_address as well as
25073         insn_current_address.
25075 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25077         PR target/85173
25078         * explow.c (emit_stack_probe): Call validize_mem on memory location
25079         before passing it to gen_probe_stack.  Create address operand and
25080         legitimize it for the probe_stack_address case.
25082 2018-04-09  Jan Hubicka  <jh@suse.cz>
25084         PR lto/85078
25085         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
25086         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
25087         * tree.c (free_lang_data_in_type): Fix handling of binfos;
25088         walk basetypes.
25089         (free_lang_data): Rebuild type inheritance graph.
25091 2018-04-09  Martin Sebor  <msebor@redhat.com>
25093         * invoke.texi (-finline-small-functions): Mention other optimization
25094         options.
25095         (-findirect-inlining, -fpartial-inlining): Same.
25096         (-finline-functions-called-once): Same.
25097         (-freorder-blocks-and-partition): Same.
25099 2018-04-09  Jan Hubicka  <jh@suse.cz>
25101         PR rtl/84058
25102         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
25103         jumps; choose last target that matches the criteria (i.e.
25104         no partition changes for non-crossing jumps).
25105         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
25106         support for redirecting crossing jumps to non-crossing.
25108 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
25110         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
25111         also for naked functions.
25113 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
25115         * config/arc/arc.md (add_shift): New pattern.
25116         (add_shift2): Likewise.
25117         (sub_shift): Likewise.
25118         (sub_shift_cmp0_noout): Likewise.
25119         (compare_si_ashiftsi): Likewise.
25120         (xbfu_cmp0_noout): New combine pattern.
25121         (xbfu_cmp0"): Likewise.
25122         (movsi_set_cc_insn): Place the predicable variant first.
25123         (commutative_binary_cmp0_noout): Remove clobber.
25124         (commutative_binary_cmp0): New pattern.
25125         (noncommutative_binary_cmp0): Likewise.
25126         (noncommutative_binary_cmp0_noout): Likewise.
25127         (noncommutative_binary_comparison_result_used): Removed.
25128         (rsub_cmp0): New pattern.
25129         (rsub_cmp0_noout): Likewise.
25130         (extzvsi): Changed, keep only meaningful variants.
25131         (SQH, SEZ): New iterators.
25132         (SQH_postfix): New mode attribute.
25133         (SEZ_prefix): New code attribute.
25134         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
25135         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
25136         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
25137         of numerical value.
25138         (noncommutative_operator): Check the availability of barrel
25139         shifter option.
25141 2018-04-09  Richard Biener  <rguenther@suse.de>
25143         PR tree-optimization/85284
25144         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
25145         Only use the niter constraining form of simple_iv when the exit
25146         is always executed.
25148 2018-04-09  Tom de Vries  <tom@codesourcery.com>
25150         PR target/84041
25151         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
25152         (define_expand "*memory_barrier"): New define_expand.
25153         (define_insn "memory_barrier"): New insn.
25155 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25157         PR rtl-optimization/80463
25158         PR rtl-optimization/83972
25159         PR rtl-optimization/83480
25161         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
25162         correct producer for the insn.
25163         (tidy_control_flow): Fixup seqnos in case of debug insns.
25165 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25167         PR rtl-optimization/83913
25169         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
25170         different sched-times when merging exprs.
25172 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25174         PR rtl-optimization/83962
25176         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
25177         tidy_fallthru_edge and tidy_control_flow.
25179 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
25181         PR rtl-optimization/83530
25183         * sel-sched.c (force_next_insn): New global variable.
25184         (remove_insn_for_debug): When force_next_insn is true, also leave only
25185         next insn in the ready list.
25186         (sel_sched_region): When the region wasn't scheduled, make another pass
25187         over it with force_next_insn set to 1.
25189 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
25191         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
25192         into tm_file.
25193         * config/nds32/constants.md (unspec_volatile_element): Add enum values
25194         for interrupt control.
25195         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25196         functions for interrupt control.
25197         * config/nds32/nds32-intrinsic.md: Likewise.
25198         * config/nds32/nds32_intrinsic.h: Likewise.
25199         * config/nds32/nds32.h (nds32_builtins): Likewise.
25201 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
25203         * config/nds32/nds32.c (nds32_init_machine_status,
25204         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
25205         strict_aligned_p field.
25206         (nds32_expand_to_rtl_hook): New function.
25207         (TARGET_EXPAND_TO_RTL_HOOK): Define.
25208         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
25210 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
25211             Chung-Ju Wu  <jasonwucj@gmail.com>
25213         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
25214         * config/nds32/nds32-n7.md: New file.
25215         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
25216         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
25217         pipeline.
25218         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
25219         * config/nds32/nds32.md (pipeline_model): Add n7.
25220         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
25221         * config/nds32/pipelines.md: Include n7 settings.
25223 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
25224             Chung-Ju Wu  <jasonwucj@gmail.com>
25226         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
25227         * config/nds32/nds32-e8.md: New file.
25228         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
25229         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
25230         pipeline.
25231         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
25232         * config/nds32/nds32.md (pipeline_model): Add e8.
25233         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
25234         * config/nds32/pipelines.md: Include e8 settings.
25236 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
25237             Chung-Ju Wu  <jasonwucj@gmail.com>
25239         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
25240         * config/nds32/nds32-n8.md: New file.
25241         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
25242         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
25243         pipeline.
25244         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
25245         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
25246         * config/nds32/nds32.md (pipeline_model): Add n8.
25247         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
25248         * config/nds32/pipelines.md: Include n8 settings.
25250 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
25251             Chung-Ju Wu  <jasonwucj@gmail.com>
25253         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
25254         * config/nds32/nds32-n9-2r1w.md: New file.
25255         * config/nds32/nds32-n9-3r2w.md: New file.
25256         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
25257         nds32_register_ports): New or modify for cpu n9.
25258         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
25259         pipeline.
25260         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
25261         * config/nds32/nds32-utils.c: New file.
25262         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
25263         TARGET_MUL_SLOW): Define.
25264         * config/nds32/nds32.md (pipeline_model): New attribute.
25265         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
25266         New options that support cpu n9.
25267         * config/nds32/pipelines.md: Include n9 settings.
25268         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
25270 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
25272         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
25273         information if necessary.
25274         (output_cond_branch_compare_zero): Likewise.
25275         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
25276         (nds32_target_alignment): Refine for alignment.
25277         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
25278         (FUNCTION_BOUNDARY): Modify.
25279         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
25280         align case.
25281         * config/nds32/nds32.opt (malways-align, malign-functions): New.
25283 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
25285         * config/nds32/constants.md (unspec_volatile_element): Add values for
25286         TLB operation and data prefetch.
25287         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25288         functions for TLB operation and data prefetch.
25289         * config/nds32/nds32-intrinsic.md: Likewise.
25290         * config/nds32/nds32_intrinsic.h: Likewise.
25291         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
25292         (nds32_print_operand): Likewise.
25293         * config/nds32/nds32.h (nds32_builtins): Likewise.
25295 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
25296         Andrew Pinski <pinsika@gcc.gnu.org>
25298         PR middle-end/82976
25299         * match.pd: Use constant_boolean_node of correct type instead of
25300         boolean_true_node or boolean_false_node for simplifying
25301         pointer comparisons to zero.
25303 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
25305         PR tree-optimization/80021
25306         * tree.c (verify_type_variant): Make error call in verify_variant_match
25307         translatable and remove final full stop.
25309 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
25311         * config/nds32/constants.md (unspec_volatile_element): Add
25312         UNSPEC_VOLATILE_EH_RETURN.
25313         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
25314         nds32_output_stack_pop): Support dwarf exception handling process.
25315         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
25316         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
25317         exception handling process.
25318         (nds32_compute_stack_frame): Likewise.
25319         (nds32_return_addr_rtx): Likewise.
25320         (nds32_initial_elimination_offset): Likewise.
25321         (nds32_expand_prologue): Likewise.
25322         (nds32_expand_epilogue): Likewise.
25323         (nds32_dynamic_chain_address): New function.
25324         * config/nds32/nds32.h (machine_function): Add fields for dwarf
25325         exception handling.
25326         (DYNAMIC_CHAIN_ADDRESS): Define.
25327         (EH_RETURN_DATA_REGNO): Define.
25328         (EH_RETURN_STACKADJ_RTX): Define.
25329         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
25330         patterns for dwarf exception handling.
25332 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
25334         * config/nds32/nds32.h: Clean up obsolete macros.
25336 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
25338         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25339         Add enum values for particular instructions.
25340         * config/nds32/nds32-intrinsic.c: Implementation of expanding
25341         particular intrinsic functions.
25342         * config/nds32/nds32-intrinsic.md: Likewise.
25343         * config/nds32/nds32_intrinsic.h: Likewise.
25344         * config/nds32/nds32.h (nds32_builtins): Likewise.
25345         * config/nds32/nds32.md (type): Add pbsad and pbsada.
25346         (btst, ave): New patterns for particular instructions.
25348 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
25350         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25351         Add enum values for atomic load/store and memory sync.
25352         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
25353         and memory sync.
25354         * config/nds32/nds32-intrinsic.md: Likewise.
25355         * config/nds32/nds32_intrinsic.h: Likewise.
25356         * config/nds32/nds32.h (nds32_builtins): Likewise.
25358 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
25360         PR tree-optimization/85257
25361         * fold-const.c (native_encode_vector): If not all elts could fit
25362         and off is -1, return 0 rather than offset.
25363         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
25364         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
25365         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
25366         adjust buffer in native_interpret_expr call.
25368 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
25370         * config/nds32/constants.md (unspec_volatile_element): Add cache
25371         control enum values.
25372         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
25373         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
25374         * config/nds32/nds32.c (nds32_cctl_names): New.
25375         (nds32_print_operand): Handle cache control register names.
25376         * config/nds32/nds32.h (nds32_builtins): New enum values.
25377         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
25378         macros.
25379         * config/nds32/nds32.md (type): Add mmu.
25380         * config/nds32/pipelines.md (simple_insn): Add mmu.
25382 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
25384         * config/nds32/nds32.md (type): Remove call.
25385         * config/nds32/pipelines.md (simple_insn): Likewise.
25387 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
25389         * config/nds32/constants.md (unspec_volatile_element): Add
25390         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
25391         UNSPEC_VOLATILE_FMFCFG.
25392         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
25393         description for fmfcfg and fmfcsr.
25394         (bdesc_1arg): Add fmtcsr.
25395         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
25396         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
25397         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
25398         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
25399         unspec_fmfcfg): New patterns.
25400         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
25401         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
25402         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
25403         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
25404         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
25405         __nds32__fmfcfg): Define.
25407 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
25409         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
25410         intrinsic register names.
25411         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
25412         intrinsic register enum values and macros.
25414 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
25416         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
25417         for load/store addressing form.
25418         (nds32_print_operand_address): Likewise.
25420 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
25422         PR target/85196
25423         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
25424         based on LABEL_REF.  Remove useless assertion.
25425         (pic_address_needs_scratch): Fix formatting.
25426         (sparc_legitimize_pic_address): Minor tweaks.
25427         (sparc_delegitimize_address): Adjust assertion accordingly.
25428         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
25429         into symbolic_operand.
25430         (movsi_high_pic_label_ref): Likewise.
25431         (movsi_lo_sum_pic_label_ref): Likewise.
25432         (movdi_pic_label_ref): Likewise.
25433         (movdi_high_pic_label_ref): Likewise.
25434         (movdi_lo_sum_pic_label_ref): Likewise.
25436 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
25438         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
25439         custom LIB_SPEC setup.
25441 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
25442             Kito Cheng  <kito.cheng@gmail.com>
25444         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
25445         * config/riscv/freebsd.h: New.
25447 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
25449         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
25450         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
25451         file.
25453 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
25454             Kito Cheng  <kito.cheng@gmail.com>
25456         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
25457         nds32_output_call, nds32_symbol_binds_local_p): New functions.
25458         * config/nds32/nds32-protos.h (nds32_output_call,
25459         nds32_output_return): Declare.
25460         * config/nds32/nds32.md: Refine all the call and return patterns.
25462 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
25464         PR debug/85252
25465         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
25466         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
25468         PR rtl-optimization/84872
25469         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
25470         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
25471         EDGE_CROSSING edge.
25473 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
25475         * expr.c (copy_blkmode_to_reg): Revert 254862.
25476         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
25478 2018-04-06  Richard Biener  <rguenther@suse.de>
25480         PR middle-end/85244
25481         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
25482         after seeing a component reference with an adjacent field.  Treat
25483         refs to arrays at struct end of external decls similar to
25484         refs to unconstrained commons.
25486 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
25488         PR sanitizer/85213
25489         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
25490         look through SAVE_EXPRs with non-side-effects argument.  Adjust
25491         recursive calls.
25492         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
25493         save_p here.
25495 2018-04-06  Richard Biener  <rguenther@suse.de>
25497         PR middle-end/85180
25498         * alias.c (find_base_term): New wrapper around find_base_term
25499         unwinding CSELIB_VAL_PTR changes.
25500         (find_base_term): Do not restore CSELIB_VAL_PTR during the
25501         recursion.
25503 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25505         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
25506         instructions.
25507         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
25508         constant definitions.
25509         ("nop"): lr 0,0 -> nopr r0
25510         ("nop_lr0", "nop_lr1"): New insn definitions.
25512 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
25514         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
25515         NDS32_V3PUSH_AVAILABLE_P macro.
25517 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
25518             Chung-Ju Wu  <jasonwucj@gmail.com>
25520         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
25521         (nds32*-*-*): Add float and fpu_config into supported_defaults.
25522         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
25523         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
25524         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
25525         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
25526         * config/nds32/constraints.md: New constraints and checking for hard
25527         float configuration.
25528         * config/nds32/iterators.md: New mode iterator and attribute for hard
25529         float configuration.
25530         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
25531         patterns.
25532         * config/nds32/nds32-fpu.md: New file.
25533         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
25534         deal with hard float code generation.
25535         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
25536         ARCH_V3S.
25537         (abi_type, float_reg_number): New enum type.
25538         * config/nds32/nds32-predicates.c: New predicates for hard float.
25539         * config/nds32/nds32-protos.h: Declare functions for hard float.
25540         * config/nds32/nds32.c: Implementation for hard float configuration.
25541         * config/nds32/nds32.h: Definitions for hard float configuration.
25542         * config/nds32/nds32.md: Include hard float machine description and
25543         modify patterns for hard float configuration.
25544         * config/nds32/nds32.opt: New options for hard float configuration.
25545         * config/nds32/predicates.md: New predicates for hard float
25546         configuration.
25548 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
25550         * common/config/nds32/nds32-common.c
25551         (nds32_option_optimization_table): Enable -mreleax-hint by default.
25553 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
25555         PR middle-end/85195
25556         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
25557         CONSTRUCTOR_ELT (ctor, ...)->value.
25559 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
25561         PR target/85193
25562         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
25564 2018-04-05  Tom de Vries  <tom@codesourcery.com>
25566         PR target/85204
25567         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
25568         cond jump.
25570 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
25571             Kito Cheng  <kito.cheng@gmail.com>
25573         * config/nds32/constraints.md (U33): Fine-tune checking condition.
25574         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
25575         * config/nds32/nds32.h (nds32_16bit_address_type): Add
25576         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
25578 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
25579             Kito Cheng  <kito.cheng@gmail.com>
25581         * config/nds32/constraints.md (Ufe): New memory constraint.
25582         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
25583         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
25584         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
25585         operands.
25586         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
25587         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
25589 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25591         * config/nds32/nds32.md: Use optimize_size in the condition for
25592         alu-shift instructions.
25594 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25596         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
25598 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25600         * config/nds32/nds32.md (negsi2): Refine pattern.
25602 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
25603             Chung-Ju Wu  <jasonwucj@gmail.com>
25605         * config/nds32/iterators.md (shift_rotate): New code iterator.
25606         (shift): New code attribute.
25607         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
25608         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
25609         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
25610         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
25611         bit-wise operations.
25612         (andsi3, *andsi3): Ditto.
25613         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
25614         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
25615         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
25616         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
25617         nds32_ior_operand, nds32_xor_operand): New predicates.
25619 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25621         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
25622         (addsi3, subsi3): ... this.
25624 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25626         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
25628 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25630         * config/nds32/nds32.md: Adjust indention.
25632 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
25634         * config/nds32/nds32.md (feature): New attribute.
25636 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
25638         * config/nds32/nds32.md (subtype): New attribute.
25640 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25642         PR target/85203
25643         * config/arm/arm-builtins.c (arm_expand_builtin): Change
25644         expansion to perform a bitwise AND of the argument followed by a
25645         boolean negation of the result.
25647 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
25649         PR rtl-optimization/84878
25650         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
25651         the basic block.  Assert the use reference is not artificial and that
25652         it has an associated insn.
25654 2018-04-04  Michael Matz  <matz@suse.de>
25656         * builtins.c (compute_objsize): Pass correct operand
25657         to array_at_struct_end_p.
25659 2018-04-04  Richard Biener  <rguenther@suse.de>
25661         PR lto/85176
25662         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
25663         from contexts for DINFO_LEVEL_TERSE and below.
25665 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25667         * config/nds32/nds32-doubleword.md (move_<mode>): Require
25668         resiter_operand condition.
25669         * config/nds32/nds32.md (*move<mode>): Ditto.
25671 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25672             Monk Chiang  <sh.chiang04@gmail.com>
25674         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
25676 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25678         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
25680 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25681             Kito Cheng  <kito.cheng@gmail.com>
25683         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
25684         nds32_cond_code_str, output_cond_branch,
25685         output_cond_branch_compare_zero, nds32_expand_cbranch,
25686         nds32_expand_cstore, nds32_expand_movcc,
25687         nds32_output_cbranchsi4_equality_zero,
25688         nds32_output_cbranchsi4_equality_reg,
25689         nds32_output_cbranchsi4_equality_reg_or_const_int,
25690         nds32_output_cbranchsi4_greater_less_zero: New functions.
25691         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
25692         nds32_expand_cstore, nds32_expand_movcc,
25693         nds32_output_cbranchsi4_equality_zero,
25694         nds32_output_cbranchsi4_equality_reg,
25695         nds32_output_cbranchsi4_equality_reg_or_const_int,
25696         nds32_output_cbranchsi4_greater_less_zero): Declare.
25697         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
25698         nds32_rimm11s_operand): New predicates.
25699         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
25700         * config/nds32/nds32.md: Rewrite all the branch and conditional move
25701         patterns.
25703 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25705         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
25706         * config/nds32/nds32.md: Ditto.
25707         * config/nds32/pipelines.md: Ditto.
25709 2018-04-04  Richard Biener  <rguenther@suse.de>
25711         PR tree-optimization/85168
25712         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
25713         propagating abnormals.
25715 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25717         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
25719 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25720             Kito Cheng  <kito.cheng@gmail.com>
25722         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
25723         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
25724         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
25725         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
25726         * config/nds32/nds32.md (sibcall_internal): New.
25727         (sibcall_register): Remove.
25728         (sibcall_immediate): Remove.
25729         (sibcall_value_internal): New.
25730         (sibcall_value_register): Remove.
25731         (sibcall_value_immediate): Remove.
25732         * config/nds32/predicates.md (nds32_general_register_operand): New.
25733         (nds32_call_address_operand): New.
25735 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25737         PR rtl-optimization/85167
25738         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
25739         bb_defs if *split_p, instead preinitialize it to NULL.
25741         PR tree-optimization/85156
25742         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
25743         evaluating the argument multiple times.
25745 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
25747         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
25748         than vector.
25749         (_mm_cvtpd_ps): Likewise.
25750         (_mm_cvttpd_epi32): Likewise.
25751         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
25752         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
25753         vector, pixel, and bool following altivec.h include.
25755 2018-04-03  Martin Sebor  <msebor@redhat.com>
25757         * doc/extend.texi (Common Function Attributes): Clarify.
25758         (const attribute): Likewise.
25759         (pure attribute): Likewise.
25761 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25763         PR target/85169
25764         * config/i386/i386.c (ix86_expand_vector_set): Use
25765         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
25767 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
25769         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
25770         instructions when changing rounding bits to preserve precision bits
25771         in the x87 control word.
25773 2018-04-03  Martin Liska  <mliska@suse.cz>
25775         PR tree-optimization/82491
25776         * rtl.h (strip_offset_and_add): Replace += suboffset with
25777         poly_uint64 () + suboffset.
25779 2018-03-29  Martin Liska  <mliska@suse.cz>
25780             Martin Jambor  <mjambor@suse.cz>
25782         PR ipa/84947
25783         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
25784         param_type is not an integral or pointer type.
25786 2018-04-03  Richard Biener  <rguenther@suse.de>
25788         * sese.h (recompute_all_dominators): Remove.
25790 2018-04-02  Martin Sebor  <msebor@redhat.com>
25792         * doc/invoke.texi (-Wrestrict): Fix typos.
25794 2018-04-02  Jim Wilson  <jimw@sifive.com>
25796         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
25797         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
25798         (<optab>di3, <optab>si3_extend): Likewise.
25799         (<optab>si3_mask, <optab>si3_mask_1): New.
25800         (<optab>di3_mask, <optab>di3_mask_1): New.
25801         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
25802         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
25803         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
25805 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
25807         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
25808         example.
25810 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
25812         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
25813         (nds32_canonicalize_comparison): New function.
25815 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25816             Kito Cheng  <kito.cheng@gmail.com>
25817             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
25819         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
25820         * config/nds32/constants.md (unspec_volatile_element): Add
25821         UNSPEC_VOLATILE_RELAX_GROUP.
25822         * config/nds32/nds32-relax-opt.c: New file.
25823         * config/nds32/nds32-predicates.c
25824         (nds32_symbol_load_store_p): New function.
25825         * config/nds32/nds32-protos.h
25826         (nds32_symbol_load_store_p): Declare function.
25827         (make_pass_nds32_relax_opt): Declare new rtl pass function.
25828         * config/nds32/nds32.c
25829         (nds32_register_pass): New function to register pass.
25830         (nds32_register_passes): New function to register passes.
25831         * config/nds32/nds32.md (relax_group): New pattern.
25832         * config/nds32/nds32.opt (mrelax-hint): New option.
25833         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
25835 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
25837         * config/nds32/t-nds32: Modify files dependency.
25839 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25841         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
25842         (PROFILE_HOOK): Define its implementation.
25844 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25846         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
25847         type and 32-bit size.
25849 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
25851         PR middle-end/85090
25852         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
25853         (V_128_256): New mode iterator.
25854         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
25855         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
25856         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
25857         of V.
25858         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
25859         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
25861 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
25863         PR target/83315
25864         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
25865         NaN inputs correctly.
25867 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
25869         PR target/80546
25870         * config/rs6000/vsx.md (??r): New mode attribute.
25871         (*vsx_mov<mode>_64bit): Use it.
25872         (*vsx_mov<mode>_32bit): Likewise.
25874 2018-03-30  Martin Sebor  <msebor@redhat.com>
25876         PR tree-optimization/84818
25877         * builtins.c (check_access): Use warning_n.
25879 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25881         PR target/83822
25882         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
25883         condition.
25884         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
25885         condition.
25887 2018-03-30  Julia Koval  <julia.koval@intel.com>
25889         PR target/84413
25890         * x86-tune.def (movx, partial_reg_dependency): Enable for
25891         m_SKYLAKE_AVX512.
25893 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
25895         PR inline-asm/84985
25896         * lra-constraints.c (process_alt_operands): Move setting
25897         this_alternative_matches below.
25899 2018-03-29  Martin Liska  <mliska@suse.cz>
25901         PR lto/84995.
25902         * doc/invoke.texi: Document how LTO works with debug info.
25903         Describe auto-load support of binutils.  Mention 'x86-64'
25904         as valid option value of -march option.
25906 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
25908         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
25910         PR c/85094
25911         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
25912         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
25913         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
25914         checking.
25916 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25918         PR target/84912
25919         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
25920         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
25921         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
25922         for RS6000_BTM_POWERPC64.
25923         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
25924         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
25925         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
25926         definition.
25927         (DIVDE): Use it.
25928         (DIVDEU): Likewise.
25930 2018-03-28  Carl Love  <cel@us.ibm.com>
25932         Revert
25933         2017-09-27  Carl Love  <cel@us.ibm.com>
25935         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
25936         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
25937         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
25938         fctiw instruction.
25940 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25942         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
25943         instead of __vector bool.
25944         (_mm_max_pu8): Likewise.
25945         (_mm_min_pi16): Likewise.
25947 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25949         PR target/84912
25950         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
25951         (DIVWEUO): Likewise.
25952         (DIVDEO): Likewise.
25953         (DIVDEUO): Likewise.
25954         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
25955         DIVWEUO and DIVDEUO.
25956         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
25957         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
25958         (div_extend): Likewise.
25959         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
25960         builtin function.
25961         (__builtin_divweuo): Likewise.
25962         (__builtin_divdeo): Likewise.
25963         (__builtin_divdeuo): Likewise.
25965 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25967         PR target/85095
25968         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
25969         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
25971         PR tree-optimization/82004
25972         * gimple-match-head.c (optimize_pow_to_exp): New function.
25973         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
25974         Don't fold to exp if optimize_pow_to_exp is false.
25976 2018-03-28  Martin Liska  <mliska@suse.cz>
25978         PR other/84819
25979         * calls.c (initialize_argument_information): Fix trailing space.
25980         * common.opt: Fix typo and provide better explanation for
25981         -fsanitize-coverage option.
25982         * config/i386/i386.opt: Fix typo.
25984 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25985             Martin Liska  <mliska@suse.cz>
25987         PR sanitizer/85081
25988         * gimplify.c (asan_poison_variable): Don't do the check for
25989         gimplify_omp_ctxp here.
25990         (gimplify_decl_expr): Do it here.
25991         (gimplify_target_expr): Likewise.
25993 2018-03-28  Martin Liska  <mliska@suse.cz>
25995         PR target/84988
25996         * config/i386/i386.c (ix86_function_arg_advance): Do not call
25997         chkp_type_bounds_count if MPX is not enabled.
25999 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
26001         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
26003 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
26005         PR target/84914
26006         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
26007         function to create the function decl for complex long double
26008         multiply and divide for -mabi=ieeelongdouble.
26009         (init_float128_ieee): Call it.
26011 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
26013         PR target/85044
26014         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
26015         -fcf-protection=branch -mibt.
26016         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
26018 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26020         PR target/81863
26021         * config/arm/arm.c (arm_valid_symbolic_address): Handle
26022         arm_word_relocations.
26024 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
26026         PR target/85056
26027         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
26028         extern array declarations.
26030 2018-03-27  Richard Biener  <rguenther@suse.de>
26032         PR middle-end/84067
26033         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
26034         explicit single_use checks.
26036 2018-03-27  Richard Biener  <rguenther@suse.de>
26038         PR tree-optimization/85082
26039         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
26040         Valueize the VUSE.
26042 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26044         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
26045         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
26046         Turn on fasynchronous-unwind-tables and funwind-tables.
26048 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
26050         PR target/85073
26051         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
26052         (*bmi_blsr_<mode>_ccz): Ditto.
26054 2018-03-26  Tom de Vries  <tom@codesourcery.com>
26056         PR tree-optimization/85063
26057         * omp-general.c (offloading_function_p): New function.  Factor out
26058         of ...
26059         * omp-offload.c (pass_omp_target_link::gate): ... here.
26060         * omp-general.h (offloading_function_p): Declare.
26061         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
26062         with attribute omp declare target for offloading functions.
26064 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
26066         PR tree-optimization/84005
26067         * tree-data-ref.h (get_base_for_alignment): Declare.
26068         * tree-data-ref.c (get_base_for_alignment_1): New function.
26069         (get_base_for_alignment): Likewise.
26070         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26071         get_base_for_alignment to find a suitable base object, instead
26072         of always using drb->base_address.
26074 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
26076         PR inline-asm/85022
26077         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
26078         known size by default.
26080 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
26082         PR inline-asm/85030
26083         * lra-constraints.c (process_alt_operands): Don't match BLKmode
26084         and non BLKmode operands.
26086 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26088         PR target/85026
26089         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
26090         Clean up attributes.
26092 2018-03-23  Richard Biener  <rguenther@suse.de>
26094         PR debug/85020
26095         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
26096         we are going to emit early debug for LTO.
26098 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
26100         PR inline-asm/85034
26101         * function.c (match_asm_constraints_1): Don't optimize if input
26102         doesn't satisfy general_operand predicate for output's mode.
26104         PR inline-asm/85022
26105         * alias.c (write_dependence_p): Don't require for x_canonicalized
26106         non-VOIDmode if x has VOIDmode.
26108         PR sanitizer/85029
26109         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
26110         just don't try to optimize it rather than assert it never happens.
26112 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26114         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
26115         macro expansions for definition of ST_INTERNAL_<mode> and
26116         LD_INTERNAL_<mode> builtins.
26117         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
26118         Remove prototype.
26119         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
26120         function.
26121         (altivec_expand_st_builtin): Likewise.
26122         (altivec_expand_builtin): Remove calls to deleted functions.
26123         (rs6000_address_for_altivec): Delete this function.
26124         * config/rs6000/vector.md: Remove expands for
26125         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
26127 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
26129         PR target/84826
26130         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
26131         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
26132         re-computing once computed.
26133         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
26134         (arm_init_machine_status): Initialize
26135         machine->static_chain_stack_bytes.
26137 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26139         PR target/84760
26140         * doc/extend.texi: Add four new prototypes for vec_ld.
26141         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
26142         definitions for more logical presentation.
26143         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
26144         entries for V1TI variants of __builtin_altivec_ld builtin.
26145         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
26146         handling of V1TI variant of LVX icode pattern.
26147         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
26148         (rs6000_gimple_fold_builtin): Likewise.
26149         (altivec_init_builtins): Add code to define
26150         __builtin_altivec_lvx_v1ti function.
26152 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
26154         PR inline-asm/84941
26155         * function.c (match_asm_constraints_1): Don't do the optimization
26156         if input isn't a REG, SUBREG, MEM or constant.
26158 2018-03-22  Tom de Vries  <tom@codesourcery.com>
26160         PR tree-optimization/84956
26161         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
26162         bb_has_abnormal_pred.
26164 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
26166         PR sanitizer/85018
26167         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
26168         DECL_INITIAL (decl) to decl at the end.
26169         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
26170         adjust the comment.
26172 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
26174         * doc/extend.texi (__builtin_tgmath): Document when complex
26175         integer types are treated as _Complex _Float64.
26177 2018-03-21  Tom de Vries  <tom@codesourcery.com>
26179         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
26181 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
26183         PR tree-optimization/84960
26184         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
26185         if it is ENTRY block, move them into single succ of ENTRY in that case.
26187 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
26189         PR tree-optimization/84811
26190         * poly-int.h (poly_span_traits): Remove the T3 parameter and
26191         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
26192         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
26193         (known_subrange_p): Update accordingly.  Cast each value involved
26194         in the size comparison, rather than casting the result of the
26195         subtraction.
26197 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
26199         PR tree-optimization/84982
26200         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
26201         by flipping the least significant bit rather than all bits from
26202         bitpos to bitpos + bitsize - 1.
26204 2018-03-21  Nathan Sidwell  <nathan@acm.org>
26206         * doc/extend.texi (Deprecated Features): Remove mention of
26207         long-deleted deprecations.
26209 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26211         PR jit/84288
26212         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
26213         * configure: Regenerate.
26215 2018-03-21  Tom de Vries  <tom@codesourcery.com>
26217         PR tree-optimization/83126
26218         * tree-parloops.c (num_phis): New function.
26219         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
26221 2018-03-21  Nathan Sidwell  <nathan@acm.org>
26223         * doc/extend.texi (Deprecated Features): Update deprecated flags,
26224         mention anon-struct/union members and trailing attributes.
26226 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
26228         PR tree-optimization/84969
26229         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
26230         builtin memset partitions if they set different rhs values.
26232 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
26234         PR rtl-optimization/84989
26235         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
26236         VEC_DUPLICATE with scalar result mode.
26238 2018-03-21  Martin Liska  <mliska@suse.cz>
26240         PR ipa/84963
26241         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
26242         not intended return statement.
26244 2018-03-21  Martin Liska  <mliska@suse.cz>
26246         PR target/84988
26247         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
26248         (chkp_find_bound_slots_1): Limit number of iterations.
26250 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
26252         PR target/84838
26253         * Minor grammar fixes for x86 options.
26255 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
26257         PR debug/84875
26258         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
26259         holding REG_CFA_RESTORE notes, instead turn them into a USE.
26261 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
26263         PR target/83789
26264         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
26265         (altivec_lvx_<mode>_1op): Likewise.
26266         (altivec_stvx_<mode>_2op): Likewise.
26267         (altivec_stvx_<mode>_1op): Likewise.
26268         (altivec_lvx_<VM2:mode>): New define_expand.
26269         (altivec_stvx_<VM2:mode>): Likewise.
26270         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
26271         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26272         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
26273         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26274         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
26275         (rs6000_gen_lvx): Likewise.
26276         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
26277         (altivec_expand_stv_builtin): Likewise.
26278         (altivec_expand_builtin): Likewise.
26279         * config/rs6000/vector.md: Likewise.
26281 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26283         PR target/82518
26284         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
26285         BYTES_BIG_ENDIAN.
26287 2018-03-20  Richard Biener  <rguenther@suse.de>
26289         PR target/84986
26290         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
26291         sign-conversions as zero, fall back to standard scalar_stmt
26292         cost for the rest.
26294 2018-03-20  Martin Liska  <mliska@suse.cz>
26296         PR ipa/84825
26297         * predict.c (rebuild_frequencies): Handle case when we have
26298         PROFILE_ABSENT, but flag_guess_branch_prob is false.
26300 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
26302         PR target/84990
26303         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
26304         flag_section_anchors.
26305         * varasm.c (use_blocks_for_decl_p): Remove hack for
26306         dw2_force_const_mem.
26308         PR target/84845
26309         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
26310         to ...
26311         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
26312         be created, use lowpart_subreg of operands[0] rather than operands[0]
26313         itself.
26314         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
26315         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
26316         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
26317         and n constraint instead of aarch64_shift_imm_di and Usd.
26318         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
26319         (*aarch64_<optab>_reg_minus<mode>3): ... this.
26321 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
26323         PR target/82989
26324         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
26325         to favor GPR over NEON registers.
26326         (<shift>di3_neon): Likewise.
26328 2018-03-20  Tom de Vries  <tom@codesourcery.com>
26330         PR target/84952
26331         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
26332         (nvptx_process_pars): Emit bar.sync asap and alap.
26334 2018-03-20  Tom de Vries  <tom@codesourcery.com>
26336         PR target/84954
26337         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
26338         seen_label if seen_label is already set.
26340 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
26342         PR target/84945
26343         * config/i386/i386.c (fold_builtin_cpu): For features above 31
26344         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
26345         Use 1U instead of 1.  Formatting fixes.
26347         PR c/84953
26348         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
26349         instead of TREE_TYPE (s1) for the return value.
26351 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
26353         PR tree-optimization/84946
26354         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
26355         bitsize + bitsize in poly_uint64 rather than poly_int64.
26357         PR sanitizer/78651
26358         * dwarf2asm.c: Include fold-const.c.
26359         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
26360         of decl rather than decl itself.
26362         PR rtl-optimization/84643
26363         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
26365 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
26367         PR sanitizer/78651
26368         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
26369         calling assemble_variable.
26371 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
26373         PR target/81647
26374         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
26375         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
26377 2018-03-19  Jim Wilson  <jimw@sifive.com>
26379         PR bootstrap/84856
26380         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
26381         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
26382         (riscv_first_stack_step): Likewise.
26383         (riscv_option_override): Use STACK_BOUNDARY instead of
26384         MIN_STACK_BOUNDARY.
26385         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
26386         MIN_STACK_BOUNDARY.
26387         (BIGGEST_ALIGNMENT): Set to 128.
26388         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
26389         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
26390         STACK_BOUNDARY.
26392 2018-03-19  Richard Biener  <rguenther@suse.de>
26394         PR tree-optimization/84933
26395         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
26396         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
26398 2018-03-19  Richard Biener  <rguenther@suse.de>
26400         PR tree-optimization/84859
26401         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
26402         (cond_if_else_store_replacement): Perform sinking operation on
26403         single-store BBs regardless of MAX_STORES_TO_SINK setting.
26404         Generalize what a BB with a single eligible store is.
26406 2018-03-19  Richard Biener  <rguenther@suse.de>
26408         PR tree-optimization/84929
26409         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
26410         chrec_is_positive against non-chrec arg.
26412 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
26414         PR target/84711
26415         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
26417 2018-03-18  Martin Liska  <mliska@suse.cz>
26419         PR rtl-optimization/84635
26420         * regrename.c (build_def_use): Use matches_mode only when
26421         matches >= 0.
26423 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
26425         PR tree-optimization/84913
26426         * tree-vect-loop.c (vectorizable_reduction): Don't try to
26427         vectorize chains of COND_EXPRs.
26429 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
26431         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
26433 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
26435         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
26437 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
26439         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
26441 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
26442             Kito Cheng  <kito.cheng@gmail.com>
26444         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
26445         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
26446         (nds32_adjust_reg_alloc_order): New function.
26447         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
26449 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
26451         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
26452         nds32_print_operand, nds32_print_operand_address): Use
26453         HOST_WIDE_INT_PRINT_DEC instead.
26455 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
26457         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
26459 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
26461         PR target/84902
26462         * config/i386/i386.c (initial_ix86_tune_features,
26463         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
26464         unsigned long long.
26465         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
26466         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
26467         rather than 1u << ix86_tune.  Formatting fix.
26468         (ix86_option_override_internal): Change ix86_arch_mask from
26469         unsigned int to unsigned HOST_WIDE_INT, initialize to
26470         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
26471         (ix86_function_specific_restore): Likewise.
26473 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
26475         PR target/84899
26476         * postreload.c (reload_combine_recognize_pattern): Perform
26477         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
26478         truncate_int_for_mode the result for the destination's mode.
26480         PR c/84909
26481         * hsa-gen.c (mem_type_for_type): Fix comment typo.
26482         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
26483         Likewise.
26484         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
26485         Likewise.
26487 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
26489         PR target/84876
26490         * lra-assigns.c (lra_split_hard_reg_for): Don't use
26491         regno_allocno_class_array and sorted_pseudos.
26492         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
26493         insns where regno is used.
26495 2018-03-16  Martin Liska  <mliska@suse.cz>
26497         PR ipa/84833
26498         * multiple_target.c (create_dispatcher_calls): Redirect
26499         reference in the symbol table.
26501 2018-03-16  Martin Liska  <mliska@suse.cz>
26503         PR ipa/84722
26504         * multiple_target.c (create_dispatcher_calls): Redirect also
26505         an alias.
26507 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
26509         PR c++/79937
26510         PR c++/82410
26511         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
26512         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
26513         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
26515 2018-03-16  Julia Koval  <julia.koval@intel.com>
26517         * doc/invoke.texi (Skylake Server): Add CLWB.
26518         Cannonlake): Remove CLWB.
26520 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
26522         PR tree-optimization/84841
26523         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
26524         1 << 3.
26525         (FLOAT_ONE_CONST_TYPE): Define.
26526         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
26527         (sort_by_operand_rank): Put entries with higher constant_type last
26528         rather than first to match comments.
26530 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
26532         * config/nios2/nios2.md (movsi_internal): Fix thinko in
26533         split predicate.
26535 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
26537         PR c++/79085
26538         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
26539         check and use address of target always.
26541 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
26543         PR target/84574
26544         * config/i386/i386.c (indirect_thunk_needed): Update comments.
26545         (indirect_thunk_bnd_needed): Likewise.
26546         (indirect_thunks_used): Likewise.
26547         (indirect_thunks_bnd_used): Likewise.
26548         (indirect_return_needed): New.
26549         (indirect_return_bnd_needed): Likewise.
26550         (output_indirect_thunk_function): Add a bool argument for
26551         function return.
26552         (output_indirect_thunk_function): Don't generate alias for
26553         function return thunk.
26554         (ix86_code_end): Call output_indirect_thunk_function to generate
26555         function return thunks.
26556         (ix86_output_function_return): Set indirect_return_bnd_needed
26557         and indirect_return_needed instead of indirect_thunk_bnd_needed
26558         and indirect_thunk_needed.
26560 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
26562         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
26563         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
26564         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
26566 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
26567             Paul Hua <paul.hua.gm@gmail.com>
26569         PR c/84852
26570         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
26572 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
26574         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
26575         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
26576         resp. SFmode cases.
26578 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
26580         PR target/84711
26581         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
26582         instead of GET_MODE_SIZE when comparing Units.
26584 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
26586         PR target/68256
26587         * varasm.c (hash_section): Return an unchangeble hash value
26588         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
26589         Return !aarch64_can_use_per_function_literal_pools_p ().
26591 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
26593         PR target/84860
26594         * optabs.c (emit_conditional_move): Pass address of cmode's copy
26595         rather than address of cmode as last argument to prepare_cmp_insn.
26597 2018-03-15  Julia Koval  <julia.koval@intel.com>
26599         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
26600         F_AVX512VNNI, F_AVX512BITALG): New.
26602 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
26604         PR target/83451
26605         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
26606         insn for floating-point loads and stores.
26608 2018-03-14  Carl Love  <cel@us.ibm.com>
26610         * config/rs6000/rs6000-c.c: Add macro definitions for
26611         ALTIVEC_BUILTIN_VEC_PERMXOR.
26612         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
26613         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
26614         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
26615         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
26616         UNSPEC_VPERMXOR.
26617         * config/doc/extend.texi: Add prototypes for vec_permxor.
26619 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
26621         PR c/84852
26622         * diagnostic-show-locus.c (class layout_point): Convert m_line
26623         from int to linenum_type.
26624         (line_span::comparator): Use linenum "compare" function when
26625         comparing line numbers.
26626         (test_line_span): New function.
26627         (layout_range::contains_point): Convert param "row" from int to
26628         linenum_type.
26629         (layout_range::intersects_line_p): Likewise.
26630         (layout::will_show_line_p): Likewise.
26631         (layout::print_source_line): Likewise.
26632         (layout::should_print_annotation_line_p): Likewise.
26633         (layout::print_annotation_line): Likewise.
26634         (layout::print_leading_fixits): Likewise.
26635         (layout::annotation_line_showed_range_p): Likewise.
26636         (struct line_corrections): Likewise for field m_row.
26637         (line_corrections::line_corrections): Likewise for param "row".
26638         (layout::print_trailing_fixits): Likewise.
26639         (layout::get_state_at_point): Likewise.
26640         (layout::get_x_bound_for_row): Likewise.
26641         (layout::print_line): Likewise.
26642         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
26643         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
26644         * input.c (selftest::test_linenum_comparisons): New function.
26645         (selftest::input_c_tests): Call it.
26646         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
26647         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
26648         * selftest.h (ASSERT_GT): New macro.
26649         (ASSERT_GT_AT): New macro.
26650         (ASSERT_LT): New macro.
26651         (ASSERT_LT_AT): New macro.
26653 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
26655         PR rtl-optimization/84780
26656         * combine.c (distribute_links): Don't make a link based on pc_rtx.
26658 2018-03-14  Martin Liska  <mliska@suse.cz>
26660         * tree.c (record_node_allocation_statistics): Use
26661         get_stats_node_kind.
26662         (get_stats_node_kind): New function extracted from
26663         record_node_allocation_statistics.
26664         (free_node): Use get_stats_node_kind.
26666 2018-03-14  Richard Biener  <rguenther@suse.de>
26668         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
26669         that the value-set of ANTIC_IN doesn't grow.
26671         Revert
26672         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
26673         member.
26674         (BB_VISITED_WITH_VISITED_SUCCS): New define.
26675         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
26677 2018-03-14  Julia Koval  <julia.koval@intel.com>
26679         * config.gcc (icelake-client, icelake-server): New.
26680         (icelake): Remove.
26681         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
26682         (initial_ix86_arch_features): Ditto.
26683         (PTA_SKYLAKE): Add SGX.
26684         (PTA_ICELAKE): Remove.
26685         (PTA_ICELAKE_CLIENT): New.
26686         (PTA_ICELAKE_SERVER): New.
26687         (ix86_option_override_internal): Split up icelake on icelake client and
26688         icelake server.
26689         (get_builtin_code_for_version): Ditto.
26690         (fold_builtin_cpu): Ditto.
26691         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
26692         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
26693         * config/i386/i386.h (processor_type): Ditto.
26694         * doc/invoke.texi: Ditto.
26696 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
26698         PR sanitizer/83392
26699         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
26700         INTEGER_CST offset, add it together with bitpos / 8 and
26701         sign extend based on POINTER_SIZE.
26703         PR target/84844
26704         Revert
26705         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
26707         PR target/78090
26708         * config/i386/constraints.md (Yc): New register constraint.
26709         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
26710         Use Yc constraint for alternative 2 of operand 0.  Remove
26711         preferred_for_speed attribute.
26713 2018-03-14  Richard Biener  <rguenther@suse.de>
26715         PR tree-optimization/84830
26716         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
26717         with the old one to avoid oscillations.
26719 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
26721         PR target/83712
26722         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
26723         pseudos.
26724         (assign_by_spills): Return a flag of reload assignment failure.
26725         Do not process the reload assignment failures.  Do not spill other
26726         reload pseudos if they has the same reg class.  Update n if
26727         necessary.
26728         (lra_assign): Add a return arg.  Set up from the result of
26729         assign_by_spills call.
26730         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26731         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26732         usage_insns if it is not NULL.
26733         (spill_hard_reg_in_range): New function.
26734         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26735         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26736         function prototypes.
26737         (lra_assign): Change prototype.
26738         * lra.c (lra): Add code to deal with fails by splitting hard reg
26739         live ranges.
26741 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
26743         * config/riscv/riscv.opt (mrelax): New option.
26744         * config/riscv/riscv.c (riscv_file_start): Emit ".option
26745         "norelax" when riscv_mrelax is disabled.
26746         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
26748 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26750         PR target/84743
26751         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
26752         reassociation for int modes.
26754 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26756         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
26757         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
26758         for big-endian.
26759         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
26760         * config/aarch64/aarch64-sve.md
26761         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
26762         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
26763         (*extend<mode><Vwide>2): Rename to...
26764         (aarch64_sve_extend<mode><Vwide>2): ...this.
26765         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
26766         renaming the old pattern to...
26767         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
26768         unsigned packs.
26769         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
26770         define_expand, renaming the old pattern to...
26771         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
26772         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
26773         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
26774         account when deciding which SVE instruction the optab should use.
26775         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
26777 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26779         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
26780         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
26781         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
26782         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
26783         (tlsdesc_small_<mode>): Turn a define_expand and use
26784         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
26785         (tlsdesc_small_advsimd_<mode>): ...this.
26786         (tlsdesc_small_sve_<mode>): New pattern.
26788 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26790         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
26791         (UNSPEC_UMUL_HIGHPART): New constants.
26792         (MUL_HIGHPART): New int iteraor.
26793         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
26794         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
26795         define_expand.
26796         (*<su>mul<mode>3_highpart): New define_insn.
26798 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
26800         PR lto/84805
26801         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
26802         incomplete types.
26804 2018-03-13  Martin Liska  <mliska@suse.cz>
26806         PR ipa/84658.
26807         * (sem_item_optimizer::sem_item_optimizer): Initialize new
26808         vector.
26809         (sem_item_optimizer::~sem_item_optimizer): Release it.
26810         (sem_item_optimizer::merge_classes): Register variable aliases.
26811         (sem_item_optimizer::fixup_pt_set): New function.
26812         (sem_item_optimizer::fixup_points_to_sets): Likewise.
26813         * ipa-icf.h: Declare new variables and functions.
26815 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
26817         PR middle-end/84834
26818         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
26819         integer_pow2p@2 and test integer_pow2p in condition.
26820         (A < 0 ? C : 0): Similarly for @1.
26822         PR middle-end/84831
26823         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
26824         characters starting at p contain '\0' character, don't look beyond
26825         that.
26827         PR target/84827
26828         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
26829         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
26831         PR target/84828
26832         * reg-stack.c (change_stack): Change update_end var from int to
26833         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
26834         also call set_block_for_insn on the newly added insns and rescan.
26836         PR target/84786
26837         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
26838         on the last operand.
26840         PR c++/84704
26841         * tree.c (stabilize_reference_1): Return save_expr (e) for
26842         STATEMENT_LIST even if it doesn't have side-effects.
26844 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
26846         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
26848 2018-03-12  Renlin Li  <renlin.li@arm.com>
26850         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
26851         aarch64_output_scalar_simd_mov_immediate.
26853 2018-03-12  Martin Sebor  <msebor@redhat.com>
26855         PR tree-optimization/83456
26856         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
26857         for perfectly overlapping calls to memcpy.
26858         (gimple_fold_builtin_memory_chk): Same.
26859         (gimple_fold_builtin_strcpy): Handle no-warning.
26860         (gimple_fold_builtin_stxcpy_chk): Same.
26861         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
26863 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26865         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
26866         parameter.  Use it for SFmode.
26867         (rs6000_function_arg_advance_1): Adjust.
26868         (rs6000_function_arg): Adjust.
26869         (rs6000_gimplify_va_arg): Pass false for that new parameter.
26871 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26873         PR rtl-optimization/84169
26874         PR rtl-optimization/84780
26875         * combine.c (can_combine_p): Check for a 2-insn combination whether
26876         the destination register is used between the two insns, too.
26878 2018-03-12  Richard Biener  <rguenther@suse.de>
26880         PR tree-optimization/84803
26881         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
26882         for refs DR analysis didn't process.
26884 2018-03-12  Richard Biener  <rguenther@suse.de>
26886         PR tree-optimization/84777
26887         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
26888         force-vectorize loops ignore whether we are optimizing for size.
26890 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26892         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
26893         (TARGET_MD_ASM_ADJUST): Define.
26895 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
26896             Kito Cheng  <kito.cheng@gmail.com>
26897             Chung-Ju Wu  <jasonwucj@gmail.com>
26899         * config/nds32/nds32.c (nds32_compute_stack_frame,
26900         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
26901         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
26902         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
26903         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
26904         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
26905         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
26906         * config/nds32/nds32.md (prologue, epilogue): Use macro
26907         NDS32_V3PUSH_AVAILABLE_P to do checking.
26909 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
26911         PR debug/58150
26912         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
26913         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
26914         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
26915         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
26916         addition of most attributes on !orig_type_die or the attribute not
26917         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
26919 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26920             Chung-Ju Wu  <jasonwucj@gmail.com>
26922         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
26923         __NDS32_VH__ macro.
26924         * config/nds32/nds32.opt (mvh): New option.
26926 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26927             Chung-Ju Wu  <jasonwucj@gmail.com>
26929         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
26930         function.
26931         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
26932         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
26933         definition.
26935 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26936             Chung-Ju Wu  <jasonwucj@gmail.com>
26938         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
26939         function.
26940         * config/nds32/nds32-multiple.md (strlensi): New pattern.
26941         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
26943 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
26944             Kito Cheng  <kito.cheng@gmail.com>
26945             Chung-Ju Wu  <jasonwucj@gmail.com>
26947         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
26948         UNSPEC_FFMISM and UNSPEC_FLMISM.
26949         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
26950         for ffb, ffmism and flmism.
26951         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
26952         (unspec_ffmism): Ditto.
26953         (unspec_flmism): Ditto.
26954         (nds32_expand_builtin_impl): Check if string extension is available.
26955         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
26956         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
26958 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
26960         Reverting patch:
26961         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
26963         PR target/83712
26964         * lra-assigns.c (assign_by_spills): Return a flag of reload
26965         assignment failure.  Do not process the reload assignment
26966         failures.  Do not spill other reload pseudos if they has the same
26967         reg class.
26968         (lra_assign): Add a return arg.  Set up from the result of
26969         assign_by_spills call.
26970         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26971         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26972         usage_insns if it is not NULL.
26973         (spill_hard_reg_in_range): New function.
26974         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26975         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26976         function prototypes.
26977         (lra_assign): Change prototype.
26978         * lra.c (lra): Add code to deal with fails by splitting hard reg
26979         live ranges.
26981 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
26983         PR target/84807
26984         * config/i386/i386.opt: Replace Enforcment with Enforcement.
26986 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
26988         PR debug/84620
26989         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
26990         (dw_val_node): Add val_symbolic_view.
26991         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
26992         (symview_upper_bound): New.
26993         (new_line_info_table): Initialize symviews_since_reset.
26994         (dwarf2out_source_line): Count symviews_since_reset and set
26995         symview_upper_bound.
26996         (dw_val_equal_p): Handle symview.
26997         (add_AT_symview): New.
26998         (print_dw_val): Handle symview.
26999         (attr_checksum, attr_checksum_ordered): Likewise.
27000         (same_dw_val_p, size_of_die): Likewise.
27001         (value_format, output_die): Likewise.
27002         (add_high_low_attributes): Use add_AT_symview for entry_view.
27003         (dwarf2out_finish): Reset symview_upper_bound, clear
27004         zero_view_p.
27006 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
27008         PR target/83969
27009         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
27010         Add strict argument and use it.
27011         (rs6000_split_multireg_move): Update for new strict argument.
27012         (mem_operand_gpr): Disallow all non-offsettable addresses.
27013         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
27015 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
27017         PR target/84772
27018         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
27019         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
27020         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
27022         PR c++/84767
27023         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
27024         decl, use remap_type if we want to use the type.
27026 2018-03-09  Martin Sebor  <msebor@redhat.com>
27028         PR tree-optimization/84526
27029         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27030         Remove dead code.
27031         (builtin_access::generic_overlap): Be prepared to handle non-array
27032         base objects.
27034 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
27036         PR rtl-optimization/84682
27037         * lra-constraints.c (process_address_1): Check is_address flag
27038         for address constraints.
27039         (process_alt_operands): Likewise.
27040         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
27041         preprocess_constraints.
27042         * recog.h (preprocess_constraints): Add oploc parameter.
27043         Adjust callers.
27044         * recog.c (preprocess_constraints): Test address_operand for
27045         CT_ADDRESS constraints.
27047 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
27049         PR target/83712
27050         * lra-assigns.c (assign_by_spills): Return a flag of reload
27051         assignment failure.  Do not process the reload assignment
27052         failures.  Do not spill other reload pseudos if they has the same
27053         reg class.
27054         (lra_assign): Add a return arg.  Set up from the result of
27055         assign_by_spills call.
27056         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27057         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
27058         usage_insns if it is not NULL.
27059         (spill_hard_reg_in_range): New function.
27060         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27061         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27062         function prototypes.
27063         (lra_assign): Change prototype.
27064         * lra.c (lra): Add code to deal with fails by splitting hard reg
27065         live ranges.
27067 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27069         PR target/83193
27070         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
27071         Accept complain bool parameter.  Only emit errors if it is true.
27072         (arm_parse_cpu_option_name): Likewise.
27073         (arm_target_thumb_only): Adjust callers of the above.
27074         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
27075         prototype to take a default true bool parameter.
27076         (arm_parse_arch_option_name): Likewise.
27078 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
27079             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
27081         PR jit/64089
27082         PR jit/84288
27083         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
27084         * configure: Regenerate.
27085         * configure.ac ("linker --version-script option"): New.
27086         ("linker soname option"): New.
27088 2018-03-09  Richard Biener  <rguenther@suse.de>
27090         PR tree-optimization/84775
27091         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
27092         immediate uses of predicate stmts and mark them modified.
27094         Revert
27095         PR tree-optimization/84178
27096         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27097         to caller.
27098         (version_loop_for_if_conversion): Delay update_ssa call.
27099         (tree_if_conversion): Delay update_ssa until after predicate
27100         insertion.
27102 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
27104         PR target/84763
27105         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
27106         when the function accesses prior frames.
27108 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
27110         PR debug/84456
27111         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
27112         gen_llsym, otherwise call maybe_gen_llsym.
27114         PR inline-asm/84742
27115         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
27116         has ',' character inside of it.
27118 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27120         PR target/84748
27121         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
27122         as clobbering CC_REGNUM.
27124 2018-03-08  Richard Biener  <rguenther@suse.de>
27126         PR middle-end/84552
27127         * tree-scalar-evolution.c: Include tree-into-ssa.h.
27128         (follow_copies_to_constant): Do not follow SSA names registered
27129         for update.
27131 2018-03-08  Richard Biener  <rguenther@suse.de>
27133         PR tree-optimization/84178
27134         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27135         to caller.
27136         (version_loop_for_if_conversion): Delay update_ssa call.
27137         (tree_if_conversion): Delay update_ssa until after predicate
27138         insertion.
27140 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
27142         PR tree-optimization/84178
27143         * tree-if-conv.c (release_bb_predicate): Remove the
27144         the assertion that the stmts have NULL use_ops.
27145         Discard the statements, asserting that they haven't
27146         yet been added to a BB.
27148 2018-03-08  Richard Biener  <rguenther@suse.de>
27150         PR tree-optimization/84746
27151         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
27152         (phi_translate): Pass in destination ANTIC_OUT set.
27153         (phi_translate_1): Likewise.  For a simplified result lookup
27154         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
27155         (phi_translate_set): Adjust.
27156         (do_pre_regular_insertion): Likewise.
27157         (do_pre_partial_partial_insertion): Likewise.
27159 2018-03-08  Martin Liska  <mliska@suse.cz>
27161         PR gcov-profile/84735
27162         * doc/gcov.texi: Document usage of profile files.
27163         * gcov-io.h: Document changes in the format.
27165 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
27167         PR debug/84404
27168         PR debug/84408
27169         * dwarf2out.c (struct dw_line_info_table): Update comments for
27170         view == -1.
27171         (FORCE_RESET_NEXT_VIEW): New.
27172         (FORCE_RESETTING_VIEW_P): New.
27173         (RESETTING_VIEW_P): Check for -1 too.
27174         (ZERO_VIEW_P): Likewise.
27175         (new_line_info_table): Force-reset next view.
27176         (dwarf2out_begin_function): Likewise.
27177         (dwarf2out_source_line): Simplify zero_view_p initialization.
27178         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
27179         view directly.  Omit view when omitting .loc at line 0.
27181 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
27183         PR tree-optimization/84740
27184         * tree-switch-conversion.c (process_switch): Call build_constructors
27185         only if info.phi_count is non-zero.
27187         PR tree-optimization/84739
27188         * tree-tailcall.c (find_tail_calls): Check call arguments against
27189         DECL_ARGUMENTS (current_function_decl) rather than
27190         DECL_ARGUMENTS (func) when checking for tail recursion.
27192 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
27194         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
27195         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
27196         Volker Reichelt's entry and add entries for people that perform
27197         GCC fuzzy testing and report numerous bugs.
27199 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
27201         PR target/82411
27202         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
27203         readonly data in sdata, if that is disabled.
27204         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
27205         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
27206         -mreadonly-in-sdata option.
27208 2018-03-07  Martin Sebor  <msebor@redhat.com>
27210         PR tree-optimization/84468
27211         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
27212         basic block when looking for nul assignment.
27214 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
27216         PR target/84277
27217         * except.h (output_function_exception_table): Adjust prototype.
27218         * except.c (output_function_exception_table): Remove FNNAME parameter
27219         and add SECTION parameter.  Ouput one part of the table at a time.
27220         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
27221         the first part of the exception table and emit unwind directives.
27222         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
27223         (i386_pe_seh_cold_init): Likewise.
27224         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
27225         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
27226         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
27227         (ix86_output_call_insn): Emit a nop in one more case for SEH.
27228         * config/i386/winnt.c: Include except.h.
27229         (struct seh_frame_state): Add reg_offset, after_prologue and
27230         in_cold_section fields.
27231         (i386_pe_seh_end_prologue): Set seh->after_prologue.
27232         (i386_pe_seh_cold_init): New function.
27233         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
27234         to seh->in_cold_section.
27235         (seh_emit_push): Record the offset of the push.
27236         (seh_emit_save): Record the offet of the save.
27237         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
27238         Test seh->after_prologue to disregard the epilogue.
27239         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
27240         (i386_pe_end_cold_function): New function.
27242 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
27244         PR fortran/84565
27245         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
27246         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
27248         PR c++/84704
27249         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
27250         on tmp_var.
27251         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
27252         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
27254         PR middle-end/84723
27255         * multiple_target.c: Include tree-inline.h and intl.h.
27256         (expand_target_clones): Diagnose and fail if node->definition and
27257         !tree_versionable_function_p (node->decl).
27259 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
27261         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
27262         sprint_ul.
27263         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
27264         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
27265         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
27267 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
27269         PR target/84710
27270         * combine.c (try_combine): Use reg_or_subregno instead of handling
27271         just paradoxical SUBREGs and REGs.
27273 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
27275         * config/arc/arc.c (arc_finalize_pic): Remove function.
27276         (arc_must_save_register): We use single base PIC register, remove
27277         checks to save/restore the PIC register.
27278         (arc_expand_prologue): Likewise.
27279         * config/arc/arc-protos.h (arc_set_default_type_attributes):
27280         Remove.
27281         (arc_verify_short): Likewise.
27282         (arc_attr_type): Likewise.
27283         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
27284         (walk_stores): Likewise.
27285         (arc_address_cost): Make it static.
27286         (arc_verify_short): Likewise.
27287         (branch_dest): Likewise.
27288         (arc_attr_type): Likewise.
27289         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
27290         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
27291         (arc_final_prescan_insn): Remove inserting the nops due to
27292         hardware hazards.  It is done in reorg step.
27293         (insn_length_variant_t): Remove.
27294         (insn_length_parameters_t): Likewise.
27295         (arc_insn_length_parameters): Likewise.
27296         (arc_get_insn_variants): Likewise.
27297         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
27299 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
27301         PR inline-asm/84683
27302         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
27303         assertion failure.
27305         PR tree-optimization/84687
27306         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
27307         on new_node->decl.
27308         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
27310 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27312         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
27313         Rename to ppc_speculation_barrier.
27314         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
27315         __builtin_ppc_speculation_barrier.
27317 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
27319         PR target/84700
27320         * combine.c (combine_simplify_rtx): Don't try to simplify if
27321         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
27322         are equal to x.
27324 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
27326         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
27327         to 32 bytes when compiling for POWER9.
27329 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
27331         PR target/84564
27332         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
27333         regparm >= 3 with no arg reg available also for calls with
27334         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
27336         PR target/84524
27337         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
27338         orig,vex.
27339         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
27341 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
27343         PR target/84264
27344         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
27346 2018-03-05  Richard Biener  <rguenther@suse.de>
27348         PR tree-optimization/84486
27349         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
27350         When inserting a __builtin_assume_aligned call set the LHS
27351         SSA name alignment info accordingly.
27353 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
27355         PR tree-optimization/84114
27356         * config/aarch64/aarch64.c (aarch64_reassociation_width)
27357         Avoid reassociation of FLOAT_MODE addition.
27359 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
27361         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
27362         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
27363         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
27364         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
27365         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
27366         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
27367         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
27368         and -mwbnoinvd.
27369         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
27370         __builtin_ia32_wbinvd): New builtins.
27371         (SPECIAL_ARGS2): New.
27372         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
27373         (SPECIAL_ARGS2): New.
27374         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
27375         (ix86_valid_target_attribute_inner_p): Ditto.
27376         (ix86_init_mmx_sse_builtins): Add special_args2.
27377         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
27378         TARGET_WBNOINVD_P): New.
27379         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
27380         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
27381         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
27382         * config/i386/immintrin.h (_wbinvd): New intrinsic.
27383         * config/i386/pconfigintrin.h: New file.
27384         * config/i386/wbnoinvdintrin.h: Ditto.
27385         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
27386         wbnoinvdintrin.h.
27387         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
27389 2018-03-05  Richard Biener  <rguenther@suse.de>
27391         PR tree-optimization/84670
27392         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27393         member.
27394         (BB_VISITED_WITH_VISITED_SUCCS): New define.
27395         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27396         (compute_antic_aux): Only assert the number of values in ANTIC_IN
27397         doesn't grow if all successors (recursively) were visited at least
27398         once.
27400 2018-03-05  Richard Biener  <rguenther@suse.de>
27402         PR tree-optimization/84650
27403         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
27404         if executed in the loop pipeline.
27406 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
27408         * doc/configfiles.texi (Configuration Files): Move info about
27409         conditionalizing $target-protos.h to...
27410         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
27411         differs from $target-protos.h.
27413 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
27414             Chung-Ju Wu  <jasonwucj@gmail.com>
27416         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
27417         * config/nds32/nds32-multiple.md (setmemsi): Define.
27418         * config/nds32/nds32-memory-manipulation.c
27419         (nds32_gen_dup_4_byte_to_word_value): New.
27420         (emit_setmem_word_loop): New.
27421         (emit_setmem_byte_loop): New.
27422         (nds32_expand_setmem_loop): New.
27423         (nds32_expand_setmem_loop_v3m): New.
27424         (nds32_expand_setmem_unroll): New.
27425         (nds32_expand_setmem): New.
27427 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
27428             Chung-Ju Wu  <jasonwucj@gmail.com>
27430         * config/nds32/nds32-memory-manipulation.c
27431         (nds32_emit_load_store): New.
27432         (nds32_emit_post_inc_load_store): New.
27433         (nds32_emit_mem_move): New.
27434         (nds32_emit_mem_move_block): New.
27435         (nds32_expand_movmemsi_loop_unknown_size): New.
27436         (nds32_expand_movmemsi_loop_known_size): New.
27437         (nds32_expand_movmemsi_loop): New.
27438         (nds32_expand_movmemsi_unroll): New.
27439         (nds32_expand_movmemqi): Rename ...
27440         (nds32_expand_movmemsi): ... to this.
27441         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
27442         (movmemsi): ... to this.
27443         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
27444         (nds32_expand_movmemsi): ... to this.
27446 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
27447             Monk Chiang  <sh.chiang04@gmail.com>
27448             Chung-Ju Wu  <jasonwucj@gmail.com>
27450         * config/nds32/nds32-protos.h
27451         (nds32_expand_load_multiple): New arguments.
27452         (nds32_expand_store_multiple): Ditto.
27453         (nds32_valid_multiple_load_store): Rename ...
27454         (nds32_valid_multiple_load_store_p): ... to this.
27455         * config/nds32/nds32-memory-manipulation.c
27456         (nds32_expand_load_multiple): Refine implementation.
27457         (nds32_expand_store_multiple): Ditto.
27458         * config/nds32/nds32-multiple.md
27459         (load_multiple): Update nds32_expand_load_multiple interface.
27460         (store_multiple): Update nds32_expand_store_multiple interface.
27461         * config/nds32/nds32-predicates.c
27462         (nds32_valid_multiple_load_store): Rename ...
27463         (nds32_valid_multiple_load_store_p): ... to this and refine
27464         implementation.
27465         * config/nds32/predicates.md
27466         (nds32_load_multiple_and_update_address_operation): New predicate.
27467         (nds32_store_multiple_and_update_address_operation): New predicate.
27469 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
27470             Chung-Ju Wu  <jasonwucj@gmail.com>
27472         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
27473         (combo): New attribute.
27474         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
27476 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
27478         * config/nds32/nds32.opt: Change -mcmodel= default value.
27480 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
27481             Monk Chiang  <sh.chiang04@gmail.com>
27482             Chung-Ju Wu  <jasonwucj@gmail.com>
27484         * config/nds32/constants.md (unspec_element): New enum.
27485         * config/nds32/constraints.md (Umw): New constraint.
27486         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
27487         * config/nds32/nds32-intrinsic.md: Likewise.
27488         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
27489         (nds32_valid_smw_lwm_base_p): New.
27490         (nds32_output_smw_single_word): New.
27491         (nds32_output_lmw_single_word): New.
27492         (nds32_expand_unaligned_load): New.
27493         (nds32_expand_unaligned_store): New.
27494         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
27495         (nds32_output_smw_single_word): Declare.
27496         (nds32_output_lmw_single_word): Declare.
27497         (nds32_expand_unaligned_load): Declare.
27498         (nds32_expand_unaligned_store): Declare.
27499         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
27500         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
27501         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
27502         NDS32_BUILTIN_UASTORE_DW.
27503         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
27504         predicate.
27506 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
27507             Kito Cheng  <kito.cheng@gmail.com>
27508             Chung-Ju Wu  <jasonwucj@gmail.com>
27510         * config/nds32/nds32-intrinsic.c
27511         (nds32_expand_builtin_null_ftype_reg): Delete.
27512         (nds32_expand_builtin_reg_ftype_imm): Ditto.
27513         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
27514         (nds32_read_argument): New.
27515         (nds32_legitimize_target): Ditto.
27516         (nds32_legitimize_argument): Ditto.
27517         (nds32_check_constant_argument): Ditto.
27518         (nds32_expand_unop_builtin): Ditto.
27519         (nds32_expand_unopimm_builtin): Ditto.
27520         (nds32_expand_binop_builtin): Ditto.
27521         (nds32_builtin_decl_impl): Ditto.
27522         (builtin_description): Ditto.
27523         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
27524         (nds32_init_builtins_impl): Ditto.
27525         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
27526         (nds32_builtin_decl): New.
27527         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
27528         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
27530 2018-03-02  Jeff Law  <law@redhat.com>
27532         * reorg.c (stop_search_p): Handle DEBUG_INSN.
27533         (redundant_insn, fill_simple_delay_slots): Likewise.
27534         (fill_slots_from_thread): Likewise.
27535         * resource.c (mark_referenced_resources): Likewise.
27536         (mark_set_resources, find_dead_or_set_registers): Likewise.
27538 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
27540         * substring-locations.h (format_warning_va): Formatting fix for
27541         ATTRIBUTE_GCC_DIAG.
27542         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
27543         argument.
27544         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
27545         * substring-locations.c: Include intl.h.
27546         (format_warning_va): Turned into small wrapper around
27547         format_warning_n_va, renamed to ...
27548         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
27549         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
27550         use ngettext.
27551         (format_warning_at_substring_n): New function.
27552         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
27553         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
27554         format_warning_at_substring with just a shorter name instead of
27555         const function pointer.
27556         (fmtwarn_n): New function.
27557         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
27558         appropriate, get rid of all the fmtstr temporaries, move conditionals
27559         with G_() wrapped string literals directly into fmtwarn arguments,
27560         cast dir.len to (int), formatting fixes.
27562 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
27564         * doc/invoke.texi: Remove "Cilk Plus" references.
27566 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
27567             Richard Biener  <rguenther@suse.de>
27569         PR ipa/84628
27570         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
27571         for error or warning attributes if CALL_FROM_THUNK_P is set.
27572         Formatting fixes.
27574 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
27576         PR target/56540
27577         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
27578         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
27580         PR target/56540
27581         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
27582         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
27584         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
27585         instead of -1U in last predictors element's probability member.
27587 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
27589         PR ipa/83983
27590         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
27591         arguments if they are comparable.
27593 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
27595         PR tree-optimization/84634
27596         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
27597         masks and masked_loop_p with a single loop_masks, making sure it's
27598         null for bb vectorization.
27600 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
27602         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
27603         (vect_analyze_data_ref_access): Use loop->safe_len rather than
27604         loop->force_vectorize to check whether there is no alias.
27606 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
27608         PR target/84614
27609         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
27610         prototypes.
27611         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
27612         comments.
27613         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
27614         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
27615         instead of a loop around prev_real_insn.
27616         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
27617         prev_real_insn.
27619         PR inline-asm/84625
27620         * config/i386/i386.c (ix86_print_operand): Use conditional
27621         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
27622         zero vector.
27624 2018-03-02  Richard Biener  <rguenther@suse.de>
27626         PR tree-optimization/84427
27627         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
27628         (bitmap_set_subtract_values): Rewrite to handle multiple
27629         exprs per value.
27630         (clean): Likewise.
27631         (prune_clobbered_mems): Likewise.
27632         (phi_translate): Take edge instead of pred/phiblock.
27633         (phi_translate_1): Likewise.
27634         (phi_translate_set): Likewise.  Insert all translated
27635         exprs for a value into the set, keeping possibly multiple
27636         expressions per value.
27637         (compute_antic_aux): Adjust for phi_translate changes.
27638         When intersecting union the expressions and prune those
27639         not in the final value set, keeping possibly multiple
27640         expressions per value.  Do not use value-insertion
27641         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
27642         all expressions.  Add verification that the value-sets
27643         only shrink during iteration.
27644         (compute_partial_antic_aux): Adjust for the phi_translate changes.
27645         (do_pre_regular_insertion): Likewise.
27646         (do_pre_partial_partial_insertion): Likewise.
27648 2018-03-02  Richard Biener  <rguenther@suse.de>
27650         PR target/82005
27651         * config/darwin.c (saved_debug_info_level): New static global.
27652         (darwin_asm_lto_start): Disable debug info generation for LTO out.
27653         (darwin_asm_lto_end): Restore debug info generation settings.
27655 2018-03-01  Martin Liska  <mliska@suse.cz>
27657         PR sanitizer/82484
27658         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
27659         volatile arguments.
27661 2018-03-01  Richard Biener  <rguenther@suse.de>
27663         PR debug/84645
27664         * dwarf2out.c (gen_variable_die): Properly handle late VLA
27665         type annotation with LTO when debug was disabled at compile-time.
27667 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
27669         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
27670         XINT with INTVAL.
27671         (mips_final_postscan_insn): Likewise.
27673 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
27675         PR rtl-optimization/84528
27676         * alias.c (init_alias_target): Add commentary.
27677         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
27678         a unique base value if the frame pointer is not eliminated
27679         to the stack pointer.
27681 2018-03-01  Tom de Vries  <tom@codesourcery.com>
27683         PR rtl-optimization/83327
27684         * lra-int.h (hard_regs_spilled_into): Declare.
27685         * lra.c (hard_regs_spilled_into): Define.
27686         (init_reg_info): Init hard_regs_spilled_into.
27687         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
27688         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
27689         (process_bb_lives): Handle hard_regs_spilled_into.
27690         (lra_create_live_ranges_1): Before doing liveness propagation, clear
27691         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
27693 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
27695         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
27696         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
27697         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
27698         * config/rs6000/aix72.h: New file.
27700 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
27702         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
27703         instead of warning_at with conditional singular and plural messages
27704         where possible.
27706         PR target/52991
27707         * stor-layout.c (update_alignment_for_field): For
27708         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
27709         && !DECL_PACKED (field), do the alignment update, just use
27710         only desired_align instead of MAX (type_align, desired_align)
27711         as the alignment.
27712         (place_field): Don't do known_align < desired_align handling
27713         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
27714         is non-NULL, instead do it after rli->prev_field handling and
27715         only if not within a bitfield word.  For DECL_PACKED (field)
27716         use type_align of BITS_PER_UNIT.
27718 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
27720         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
27721         superfluous parentheses and trailing spaces.
27723 2018-02-28  Richard Biener  <rguenther@suse.de>
27725         PR tree-optimization/84584
27726         * graphite-scop-detection.c (scop_detection::add_scop): Discard
27727         SCoPs with fake exit edge.
27729 2018-02-28  Martin Liska  <mliska@suse.cz>
27731         PR testsuite/84597
27732         * timevar.c (timer::print): Fix format to properly print 100%
27733         values.
27735 2018-02-28  Richard Biener  <rguenther@suse.de>
27737         PR middle-end/84607
27738         * genmatch.c (capture_info::walk_match): Do not mark
27739         captured expressions without operands as expr_p given
27740         they act more like predicates and should be subject to
27741         "lost tail" side-effect preserving.
27743 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
27745         PR rtl-optimization/81611
27746         * auto-inc-dec.c (attempt_change): Move dead note from
27747         mem_insn if it's the next use of regno
27748         (find_address): Take address use of reg holding
27749         non-incremented value.  Add parm to limit search to the named
27750         reg only.
27751         (merge_in_block): Attempt to use a mem insn that is the next
27752         use of the original regno.
27754 2018-02-27  Martin Sebor  <msebor@redhat.com>
27756         PR c++/83871
27757         * doc/invoke.texi (-Wmissing-attributes): New option.
27758         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
27760 2018-02-27  Martin Sebor  <msebor@redhat.com>
27762         PR translation/84207
27763         * diagnostic-core.h (warning_n, error_n, inform_n): Change
27764         n argument to unsigned HOST_WIDE_INT.
27765         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
27766         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
27767         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
27768         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
27770 2018-02-27  Richard Biener  <rguenther@suse.de>
27772         PR tree-optimization/84512
27773         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
27774         Do not use the estimate returned from record_stmt_cost for
27775         the scalar iteration cost but sum properly using add_stmt_cost.
27777 2018-02-27  Richard Biener  <rguenther@suse.de>
27779         PR tree-optimization/84466
27780         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
27781         Adjust last change to less strictly validate use operands.
27783 2018-02-27  Martin Liska  <mliska@suse.cz>
27785         PR gcov-profile/84548
27786         * gcov.c (process_file): Allow partial overlap and consider it
27787         also as group functions.
27788         (output_lines): Properly calculate range of lines for a group.
27790 2018-02-27  Martin Liska  <mliska@suse.cz>
27792         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
27793         'ggc' suffixes.  Change first column width.
27794         (timer::print): Fix formatting of the column.
27796 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
27798         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
27799         preserve inline entry blocks for the sake of debug inline
27800         entry point markers alone.
27801         (remove_unused_locals): Suggest in comments a better place to
27802         force the preservation of inline entry blocks that are
27803         otherwise unused, but do not preserve them.
27805 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27807         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
27809 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27811         PR target/84039
27812         * config/i386/constraints.md (Bs): Replace
27813         ix86_indirect_branch_register with
27814         TARGET_INDIRECT_BRANCH_REGISTER.
27815         (Bw): Likewise.
27816         * config/i386/i386.md (indirect_jump): Likewise.
27817         (tablejump): Likewise.
27818         (*sibcall_memory): Likewise.
27819         (*sibcall_value_memory): Likewise.
27820         Peepholes of indirect call and jump via memory: Likewise.
27821         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
27822         (*sibcall_value_GOT_32): Likewise.
27823         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27824         (GOT_memory_operand): Likewise.
27825         (call_insn_operand): Likewise.
27826         (sibcall_insn_operand): Likewise.
27827         (GOT32_symbol_operand): Likewise.
27828         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
27830 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27832         PR rtl-optimization/83496
27833         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
27834         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
27835         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
27836         redundant insn, if any.
27837         (relax_delay_slots): Likewise.
27838         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
27840 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
27842         PR tree-optimization/83965
27843         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
27844         that grouped statements are part of a reduction chain.  Return
27845         true if the statement is not marked as a reduction itself but
27846         is part of a group.
27847         (vect_recog_dot_prod_pattern): Don't check whether the statement
27848         is part of a group here.
27849         (vect_recog_sad_pattern): Likewise.
27850         (vect_recog_widen_sum_pattern): Likewise.
27852 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27854         PR debug/84545
27855         * final.c (rest_of_clean_state): Also look for calls inside sequences.
27857 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27859         PR target/84530
27860         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
27861         the bool argument.
27862         (ix86_output_indirect_function_return): New prototype.
27863         (ix86_split_simple_return_pop_internal): Likewise.
27864         * config/i386/i386.c (indirect_return_via_cx): New.
27865         (indirect_return_via_cx_bnd): Likewise.
27866         (indirect_thunk_name): Handle return va CX_REG.
27867         (output_indirect_thunk_function): Create alias for
27868         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
27869         (ix86_output_indirect_jmp): Remove the bool argument.
27870         (ix86_output_indirect_function_return): New function.
27871         (ix86_split_simple_return_pop_internal): Likewise.
27872         * config/i386/i386.md (*indirect_jump): Don't pass false
27873         to ix86_output_indirect_jmp.
27874         (*tablejump_1): Likewise.
27875         (simple_return_pop_internal): Change it to define_insn_and_split.
27876         Call ix86_split_simple_return_pop_internal to split it for
27877         -mfunction-return=.
27878         (simple_return_indirect_internal): Call
27879         ix86_output_indirect_function_return instead of
27880         ix86_output_indirect_jmp.
27882 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
27884         PR bootstrap/84405
27885         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
27886         memset and value initialization afterwards.
27888 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
27890         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
27892 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27894         PR target/84521
27895         * common/config/aarch64/aarch64-common.c
27896         (aarch_option_optimization_table[]): Switch
27897         off fomit-frame-pointer
27899 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27900             Chung-Ju Wu  <jasonwucj@gmail.com>
27902         * config/nds32/nds32-multiple.md (load_multiple): Disallow
27903         volatile memory.
27904         (store_multiple): Ditto.
27906 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27908         * config.gcc: Add --with-cpu support for nds32 target.
27909         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
27910         * config/nds32/nds32.opt: Add -mcpu= option.
27912 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
27914         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
27915         isel=yes): Warn for these deprecated options.
27917 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
27919         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
27920         ISA_2_5_MASKS_EMBEDDED.
27922 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
27924         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
27925         p->max as pointers rather than using iterative_hash_expr.
27927 2018-02-23  Carl Love  <cel@us.ibm.com>
27929         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
27930         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
27931         BU_P8V_OVERLOAD_2.
27932         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
27933         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
27934         P8V_BUILTIN_VEC_VUNSIGNED2.
27936 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
27938         PR target/81572
27939         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
27940         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
27941         LRA_UNKNOWN_ALT.
27942         * lra-constraints.c (curr_insn_transform): Set up
27943         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
27944         LRA_UNKNOWN_ALT.
27945         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
27946         * lra-eliminations.c (spill_pseudos): Ditto.
27947         (process_insn_for_elimination): Ditto.
27948         * lra-lives.c (reg_early_clobber_p): Use the new macros.
27949         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
27950         LRA_NON_CLOBBERED_ALT.
27952 2018-02-22  Martin Sebor  <msebor@redhat.com>
27954         PR tree-optimization/84480
27955         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
27956         to maybe_diag_stxncpy_trunc.  Call it.
27957         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
27958         from gimple_fold_builtin_strcpy.  Print inlining stack.
27959         (handle_builtin_stxncpy): Print inlining stack.
27960         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
27962 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
27964         PR target/84176
27965         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
27966         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
27967         and -fcheck-pointer-bounds are used together.
27968         (indirect_thunk_prefix): New enum.
27969         (indirect_thunk_need_prefix): New function.
27970         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
27971         "_nt" instead of "_bnd" for NOTRACK prefix.
27972         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
27973         (output_indirect_thunk_function): Likewise.
27974         (): Likewise.
27975         (ix86_code_end): Update output_indirect_thunk_function calls.
27976         (ix86_output_indirect_branch_via_reg): Replace
27977         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
27978         (ix86_output_indirect_branch_via_push): Likewise.
27979         (ix86_output_function_return): Likewise.
27980         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
27981         incompatible with -fcf-protection=branch and
27982         -fcheck-pointer-bounds.
27984 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27986         PR target/83335
27987         * config/aarch64/aarch64.c (aarch64_print_address_internal):
27988         Change gcc_assert call to output_operand_lossage.
27990 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27992         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
27994 2018-02-22  DJ Delorie  <dj@redhat.com>
27995             Sebastian Perta  <sebastian.perta@renesas.com>
27996             Oleg Endo  <olegendo@gcc.gnu.org>
27998         * config/rx/rx.c (rx_rtx_costs): New function.
27999         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
28001 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
28003         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
28005 2018-02-22  Martin Liska  <mliska@suse.cz>
28007         PR driver/83193
28008         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
28009         Add "native" as a possible value.
28011 2018-02-22  Martin Liska  <mliska@suse.cz>
28013         PR driver/83193
28014         * config/i386/i386.c (ix86_option_override_internal):
28015         Add "native" as a possible value for -march and -mtune.
28017 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
28019         PR target/84502
28020         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
28021         to all type variants.
28023         PR tree-optimization/84503
28024         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
28025         width as info->bitpos + info->bitsize - start.
28026         (merged_store_group::merge_overlapping): Simplify width computation.
28027         (check_no_overlap): New function.
28028         (imm_store_chain_info::try_coalesce_bswap): Compute expected
28029         start + width and last_order of the group, fail if check_no_overlap
28030         fails.
28031         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
28032         to group if check_no_overlap fails.
28034 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
28036         * config/rs6000/altivec.md: Delete contraint arguments to
28037         define_expand, define_split, and define_peephole2, and in
28038         define_insn_and_split if always unused.
28039         * config/rs6000/darwin.md: Ditto.
28040         * config/rs6000/dfp.md: Ditto.
28041         * config/rs6000/rs6000.md: Ditto.
28042         * config/rs6000/sync.md: Ditto.
28043         * config/rs6000/vector.md: Ditto.
28044         * config/rs6000/vsx.md: Ditto.
28046 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
28048         * config/rs6000/altivec.md: Write output control strings as braced
28049         blocks instead of double-quoted strings.
28050         * config/rs6000/darwin.md: Ditto.
28051         * config/rs6000/rs6000.md: Ditto.
28052         * config/rs6000/vector.md: Ditto.
28053         * config/rs6000/vsx.md: Ditto.
28055 2018-02-21  Jason Merrill  <jason@redhat.com>
28057         PR c++/84314 - ICE with templates and fastcall attribute.
28058         * attribs.c (build_type_attribute_qual_variant): Remove assert.
28060 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
28062         * ipa-cp.c (determine_versionability): Fix comment typos.
28064 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
28066         PR c/84229
28067         * ipa-cp.c (determine_versionability): Do not version functions caling
28068         va_arg_pack.
28070 2018-02-21  Martin Liska  <mliska@suse.cz>
28072         PR driver/83193
28073         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
28074         Add "native" as a possible value.
28075         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
28076         the macro when native cpu detection is available.
28078 2018-02-21  Martin Liska  <mliska@suse.cz>
28080         PR driver/83193
28081         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
28082         Add "native" as a possible value.
28083         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
28084         when native cpu detection is available.
28086 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
28087             Martin Sebor  <msebor@redhat.com>
28089         PR tree-optimization/84478
28090         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
28091         false.
28092         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
28093         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
28094         support which is conservatively correct, for 2 only stay conservative
28095         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
28096         argument to the 2 argument get_range_strlen, adjust 6 arg
28097         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
28098         false.
28099         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
28100         (gimple_fold_builtin_strlen): Pass true as last argument to
28101         get_range_strlen.
28103 2018-02-20  Martin Sebor  <msebor@redhat.com>
28105         PR middle-end/84095
28106         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
28107         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
28108         (builtin_memref::builtin_memref): Factor out parts into
28109         set_base_and_offset and call it.
28111 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
28113         PR middle-end/84406
28114         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
28115         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
28116         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
28117         search at the associated MODE_INT.
28119 2018-02-20  Jeff Law  <law@redhat.com>
28121         PR middle-end/82123
28122         PR tree-optimization/81592
28123         PR middle-end/79257
28124         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
28125         for range data rather than using global data.
28126         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
28127         range data rather than using global data.
28128         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
28129         pass it to children as needed.
28130         (struct directive::fmtresult): Similarly.
28131         (struct directive::set_width): Similarly.
28132         (struct directive::set_precision): Similarly.
28133         (format_integer, format_directive, parse_directive): Similarly.
28134         (format_none): Accept unnamed vr_values parameter.
28135         (format_percent, format_floating, format_character): Similarly.
28136         (format_string, format_plain): Similarly.
28137         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
28138         the EVRP range analyzer for range data rather than using global data.
28139         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
28140         gimple-ssa-evrp-analyze.h
28141         (class sprintf_dom_walker): Add after_dom_children member function.
28142         Add evrp_range_analyzer member.
28143         (sprintf_dom_walker::before_dom_children): Call into the EVRP
28144         range analyzer as needed.
28145         (sprintf_dom_walker::after_dom_children): New member function.
28146         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
28147         if not optimizing.
28148         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
28149         (evrp_range_analyzer::pop_to_marker): Likewise.
28151 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
28153         PR tree-optimization/84419
28154         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
28155         with the required type if its current type is compatible but
28156         different.
28158 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
28160         PR middle-end/82004
28161         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
28162         after vectorization.
28164 2018-02-20  Martin Liska  <mliska@suse.cz>
28166         PR driver/83193
28167         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
28168         possible values if we don't have a hint.
28170 2018-02-20  Martin Liska  <mliska@suse.cz>
28172         PR c/84310
28173         PR target/79747
28174         * final.c (shorten_branches): Build align_tab array with one
28175         more element.
28176         * opts.c (finish_options): Add alignment option limit check.
28177         (MAX_CODE_ALIGN): Likewise.
28178         (MAX_CODE_ALIGN_VALUE): Likewise.
28179         * doc/invoke.texi: Document maximum allowed option value for
28180         all -falign-* options.
28182 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
28184         PR target/84146
28185         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
28186         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
28187         * var-tracking.c (emit_note_insn_var_location): Remove all references
28188         to NOTE_INSN_CALL_ARG_LOCATION.
28189         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
28190         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
28191         Use copy_rtx_if_shared.
28192         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
28193         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
28194         (dwarf2out_var_location): Remove handling of
28195         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
28196         on call_insn.
28197         * final.c (final_scan_insn): Remove all references to
28198         NOTE_INSN_CALL_ARG_LOCATION.
28199         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
28200         before dumping final insns.
28201         * except.c (emit_note_eh_region_end): Remove all references to
28202         NOTE_INSN_CALL_ARG_LOCATION.
28203         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
28204         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
28205         * config/arc/arc.c (hwloop_optimize): Likewise.
28206         * config/arm/arm.c (create_fix_barrier): Likewise.
28207         * config/s390/s390.c (s390_chunkify_start): Likewise.
28208         * config/sh/sh.c (find_barrier): Likewise.
28209         * config/i386/i386.c (rest_of_insert_endbranch,
28210         ix86_seh_fixup_eh_fallthru): Likewise.
28211         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
28212         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
28213         * config/frv/frv.c (frv_function_prologue): Likewise.
28214         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
28215         reg note.
28216         (note_outside_basic_block_p): Remove all references to
28217         NOTE_INSN_CALL_ARG_LOCATION.
28218         * gengtype.c (adjust_field_rtx_def): Likewise.
28219         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
28220         Likewise.
28221         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
28222         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
28224         PR c++/84444
28225         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
28226         is ADDR_EXPR.
28228         PR tree-optimization/84452
28229         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
28230         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
28231         is NULL.
28233 2018-02-19  Martin Liska  <mliska@suse.cz>
28235         PR sanitizer/82183
28236         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
28238 2018-02-19  Martin Liska  <mliska@suse.cz>
28239             Richard Sandiford  <richard.sandiford@linaro.org>
28241         PR tree-optimization/82491
28242         * gimple-fold.c (get_base_constructor): Make earlier bail out
28243         to prevent ubsan.
28245 2018-02-19  Carl Love  <cel@us.ibm.com>
28247         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
28248         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
28249         BU_P8V_OVERLOAD_1.
28250         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
28251         P8V_BUILTIN_VEC_NEG.
28253 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
28255         * config/rl78/rl78.md (movdf): New define expand.
28257 2018-02-19  Martin Liska  <mliska@suse.cz>
28259         PR other/80589
28260         * doc/invoke.texi: Fix typo.
28261         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
28263 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
28265         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
28266         handle rs6000_single_float and rs6000_double_float specially for
28267         e500 family CPUs.
28269 2018-02-16  Jeff Law  <law@redhat.com>
28271         * config/rx/rx.c (add_pop_cfi_notes): New function.;
28272         (pop_regs): Use it.
28274 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
28276         PR ipa/84425
28277         * ipa-inline.c (inline_small_functions): Fix a typo.
28279 2018-02-16  Nathan Sidwell  <nathan@acm.org>
28281         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
28283 2018-02-16  Carl Love  <cel@us.ibm.com>
28285         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
28286         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
28287         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
28288         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
28289         expansion to P8V_BUILTIN_VEC_FLOAT2.
28291 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
28293         PR rtl-optimization/70023
28294         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
28295         src_regno into account.
28297 2018-02-16  Carl Love  <cel@us.ibm.com>
28299         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
28300         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
28301         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
28302         * config/rs6000/rs6000.c: Remove case statements for
28303         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
28304         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
28305         and P9V_BUILTIN_VEC_VINSERT4B.
28306         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
28307         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
28308         * config/rs6000/vsx.md:
28309         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
28310         vec_insert4b.
28312 2018-02-16  Carl Love  <cel@us.ibm.com>
28314         * config/rs6000/altivec.h: Add builtin names vec_extract4b
28315         vec_insert4b.
28316         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
28317         definitions.
28318         * config/rs6000/rs6000-c.c: Add the definitions for
28319         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
28320         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
28321         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
28322         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
28323         definition for insert4b and define insn *insert3b_internal.
28324         * doc/extend.texi: Add documentation for vec_extract4b.
28326 2018-02-16  Nathan Sidwell  <nathan@acm.org>
28328         * doc/extend.texi (Backwards Compatibility): Mention friend
28329         injection.  Note for-scope is deprecated.
28330         * doc/invoke.texi (-ffriend-injection): Deprecate.
28332 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
28334         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
28335         that moved to I2, also allow destinations that are a paradoxical
28336         subreg (instead of a normal reg).
28338 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
28340         PR target/83831
28341         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
28342         to QImode.
28344 2018-02-16  Richard Biener  <rguenther@suse.de>
28346         PR tree-optimization/84037
28347         PR tree-optimization/84016
28348         PR target/82862
28349         * config/i386/i386.c (ix86_builtin_vectorization_cost):
28350         Adjust vec_construct for the fact we need additional higher latency
28351         128bit inserts for AVX256 and AVX512 vector builds.
28352         (ix86_add_stmt_cost): Scale vector construction cost for
28353         elementwise loads.
28355 2018-02-16  Richard Biener  <rguenther@suse.de>
28357         PR tree-optimization/84417
28358         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
28359         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
28360         (non_rewritable_lvalue_p): Likewise, use poly-ints.
28362 2018-02-16  Martin Liska  <mliska@suse.cz>
28364         PR sanitizer/84307
28365         * internal-fn.def (ASAN_CHECK): Set proper flags.
28366         (ASAN_MARK): Likewise.
28368 2018-02-16  Julia Koval  <julia.koval@intel.com>
28370         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
28371         from PTA_CANNONLAKE.
28373 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
28375         PR target/84272
28376         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
28377         Use ++iter rather than iter++ for std::list iterators.
28378         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
28379         defer deleting them until all nodes in the forest are processed.  Do
28380         free even leaf nodes.  Change to_process into auto_vec.
28382         PR bootstrap/84405
28383         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
28384         * vec.h (vec_default_construct): Use memset instead of placement new
28385         if BROKEN_VALUE_INITIALIZATION is defined.
28386         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
28387         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
28388         is defined.
28390         PR rtl-optimization/83723
28391         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
28392         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
28393         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
28394         recursive calls.
28395         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
28396         callers.
28397         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
28399 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
28401         PR rtl-optimization/81443
28402         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
28403         from inner REGs to paradoxical SUBREGs.
28405 2018-02-16  Richard Biener  <rguenther@suse.de>
28407         PR tree-optimization/84399
28408         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28409         For operands we can analyze at their definition make sure we can
28410         analyze them at each use as well.
28412 2018-02-16  Richard Biener  <rguenther@suse.de>
28414         PR tree-optimization/84190
28415         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
28416         volatile accesses if the decl isn't volatile.
28418 2018-02-15  Jason Merrill  <jason@redhat.com>
28420         PR c++/84314 - ICE with templates and fastcall attribute.
28421         * attribs.c (build_type_attribute_qual_variant): Don't clobber
28422         TYPE_CANONICAL on an existing type.
28424 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
28426         PR tree-optimization/84383
28427         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
28428         dstoff nor call operand_equal_p if dstbase is NULL.
28430         PR tree-optimization/84334
28431         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
28432         also a CONSTANT_CLASS_P, punt.
28434 2018-02-14  Jim Wilson  <jimw@sifive.com>
28436         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
28437         first SMALL_OPERAND check.  New local min_second_step.  Move assert
28438         to where locals are set.  Add TARGET_RVC support.
28439         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
28441 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
28443         * doc/invoke.texi: Correct -Wformat-overflow code sample.
28445 2018-02-14  Martin Sebor  <msebor@redhat.com>
28447         PR tree-optimization/83698
28448         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
28449         arrays constrain the offset range to their bounds.
28450         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
28451         (builtin_access::overlap): Avoid setting the size of overlap if it's
28452         already been set.
28453         (maybe_diag_overlap): Also consider arrays when deciding what values
28454         of offsets to include in diagnostics.
28456 2018-02-14  Martin Sebor  <msebor@redhat.com>
28458         PR c/84108
28459         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
28460         that correspond to the kind of a declaration.
28462 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
28464         PR target/83984
28465         * config/pa/pa.md: Load address of PIC label using the linkage table
28466         if the label is nonlocal.
28468 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28470         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
28471         warning message if user requests -maltivec=be.
28472         * doc/invoke.texi: Document deprecation of -maltivec=be.
28474 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
28476         PR target/84220
28477         * config/rs6000/rs6000-c.c: Update definitions for
28478         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
28479         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
28481 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
28483         PR target/84239
28484         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
28485         add _get_ssp intrinsics. Remove argument from
28486         __builtin_ia32_rdssp[d|q].
28487         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
28488         * config/i386/i386-builtin.def: Remove argument from
28489         __builtin_ia32_rdssp[d|q].
28490         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
28491         ix86_expand_special_args_builtin for _rdssp[d|q].
28492         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
28493         Clear register before usage.
28494         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
28495         Add documentation for new _get_ssp and _inc_ssp intrinsics.
28497 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
28499         PR tree-optimization/84357
28500         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
28501         operand 1 of an ARRAY_REF too.
28503 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
28505         PR target/83831
28506         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
28507         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
28508         declarations.
28509         (set_of_reg): New struct.
28510         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
28511         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
28512         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
28513         functions.
28514         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
28515         Split into bitclr, bitset, bitinvert patterns if appropriate.
28516         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
28517         use rx_fuse_in_memory_bitop.
28518         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
28519         to named insn, correct maximum insn length.
28521 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
28523         PR target/79242
28524         * machmode.def: Define a complex mode for PARTIAL_INT.
28525         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
28526         MODE_PARTIAL_INT.
28527         * doc/rtl.texi: Document CSPImode.
28528         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
28529         handling.
28530         (msp430_hard_regno_nregs_with_padding): Likewise.
28532 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
28534         PR target/84279
28535         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
28537 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
28539         PR rtl-optimization/84169
28540         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
28541         we generated a parallel as new i3 and we split that to new i2 and i3
28542         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
28543         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
28544         those to i2, not i1.  Partially rewrite this scan code.
28546 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
28548         PR c/82210
28549         * stor-layout.c (place_field): For variable length fields, adjust
28550         offset_align afterwards not just based on the field's alignment,
28551         but also on the size.
28553         PR middle-end/84309
28554         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
28555         of exps and logs in the use_exp2 case.
28557 2018-02-13  Jeff Law  <law@redhat.com>
28559         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
28560         entry for "vector".
28562         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
28563         ARGS as unused.
28565 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
28567         PR debug/84342
28568         PR debug/84319
28569         * common.opt (gas-loc-support, gas-locview-support): New.
28570         (ginline-points, ginternal-reset-location-views): New.
28571         * doc/invoke.texi: Document them.  Use @itemx where intended.
28572         (gvariable-location-views): Adjust.
28573         * target.def (reset_location_view): New.
28574         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
28575         (TARGET_RESET_LOCATION_VIEW): New.
28576         * doc/tm.texi: Rebuilt.
28577         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
28578         (dwarf2out_default_as_locview_support): New.
28579         (output_asm_line_debug_info): Use option variables.
28580         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
28581         (output_loc_list): Likewise.
28582         (add_high_low_attributes): Check option variables.
28583         Don't output entry view attribute in strict mode.
28584         (gen_inlined_subroutine_die): Check option variables.
28585         (dwarf2out_inline_entry): Likewise.
28586         (init_sections_and_labels): Likewise.
28587         (dwarf2out_early_finish): Likewise.
28588         (maybe_reset_location_view): New, from...
28589         (dwarf2out_var_location): ... here.  Call it.
28590         * debug.h (dwarf2out_default_as_loc_support): Declare.
28591         (dwarf2out_default_as_locview_support): Declare.
28592         * hooks.c (hook_int_rtx_insn_0): New.
28593         * hooks.h (hook_int_rtx_insn_0): Declare.
28594         * toplev.c (process_options): Take -gas-loc-support and
28595         -gas-locview-support from dwarf2out.  Enable
28596         -gvariable-location-views by default only with locview
28597         assembler support.  Enable -ginternal-reset-location-views by
28598         default only if the target defines the corresponding hook.
28599         Enable -ginline-points by default if location views are
28600         enabled; force it disabled if statement frontiers are
28601         disabled.
28602         * tree-inline.c (expand_call_inline): Check option variables.
28603         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
28605 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
28607         PR tree-optimization/84321
28608         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
28609         handling.  Also check whether the anti-range contains any values
28610         that satisfy the mask; switch to a VR_RANGE if not.
28612 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
28614         PR sanitizer/84340
28615         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
28617 2018-02-13  Martin Jambor  <mjambor@suse.cz>
28619         PR c++/83990
28620         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
28621         of call statements, also set location of a load to a temporary.
28623 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
28625         * config/rl78/rl78.c (add_vector_labels): New function.
28626         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
28627         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
28628         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
28629         which checks that no arguments are passed.
28630         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
28631         * doc/extend.texi: Documentation for the new attribute.
28633 2018-02-13  Andreas Schwab  <schwab@suse.de>
28635         * config/riscv/linux.h (CPP_SPEC): Define.
28637 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
28639         PR target/84335
28640         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
28641         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
28642         OPTION_MASK_ISA_AES as first argument to def_builtin_const
28643         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
28644         instead of OPTION_MASK_ISA_PCLMUL as first argument to
28645         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
28646         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
28647         temporarily for AES and PCLMUL builtins.
28649         PR tree-optimization/84339
28650         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
28651         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
28652         Formatting fixes.
28654         PR middle-end/84309
28655         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
28656         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
28657         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
28658         inline function.
28659         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
28660         inline function.
28661         * omp-simd-clone.h: New file.
28662         * omp-simd-clone.c: Include omp-simd-clone.h.
28663         (expand_simd_clones): No longer static.
28664         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
28665         cgraph.h and omp-simd-clone.h.
28666         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
28667         (vect_recog_widen_shift_pattern): Formatting fix.
28668         (vect_pattern_recog_1): Don't check optab for calls.
28670         PR target/84336
28671         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
28672         operands[2] into a REG before using gen_lowpart on it.
28674 2018-02-12  Jeff Law  <law@redhat.com>
28676         PR target/83760
28677         * config/sh/sh.c (find_barrier): Consider a sibling call
28678         a barrier as well.
28680         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
28681         successfully back substituting a reg.
28683 2018-02-12  Richard Biener  <rguenther@suse.de>
28685         PR tree-optimization/84037
28686         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
28687         parameter, move visited init to caller.
28688         (vect_slp_analyze_operations): Separate cost from validity
28689         check, initialize visited once for all instances.
28690         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
28691         for all instances.
28692         * tree-vect-stmts.c (vect_model_simple_cost): Make early
28693         out an assert.
28694         (vect_model_promotion_demotion_cost): Likewise.
28695         (vectorizable_bswap): Guard cost modeling with !slp_node
28696         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
28697         SLP stmts.
28698         (vectorizable_call): Likewise.
28699         (vectorizable_conversion): Likewise.
28700         (vectorizable_assignment): Likewise.
28701         (vectorizable_shift): Likewise.
28702         (vectorizable_operation): Likewise.
28703         (vectorizable_store): Likewise.
28704         (vectorizable_load): Likewise.
28705         (vectorizable_condition): Likewise.
28706         (vectorizable_comparison): Likewise.
28708 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
28710         PR sanitizer/84307
28711         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
28712         (ASAN_MARK): Fix fnspec to account for return value, change pointer
28713         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
28715 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
28717         PR middle-end/83665
28718         * params.def (inline-min-speedup): Increase from 8 to 15.
28719         (max-inline-insns-auto): Decrease from 40 to 30.
28720         * ipa-split.c (consider_split): Add some buffer for function to
28721         be considered inlining candidate.
28722         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
28723         default values.
28725 2018-02-12  Richard Biener  <rguenther@suse.de>
28727         PR tree-optimization/84037
28728         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
28729         matched stmts if we cannot swap the non-matched ones.
28731 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28733         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
28734         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
28735         _mm_maskz_scalef_round_ss): New intrinsics.
28736         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
28737         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
28738         __builtin_ia32_scalefss_round): Remove.
28739         (__builtin_ia32_scalefsd_mask_round,
28740         __builtin_ia32_scalefss_mask_round): New intrinsics.
28741         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
28742         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
28743         ((match_operand:VF_128 2 "<round_nimm_predicate>"
28744         "<round_constraint>")): Changed to ...
28745         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
28746         "<round_scalar_constraint>")): ... this.
28747         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
28748         %0, %1, %2<round_op3>}"): Changed to ...
28749         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
28750         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
28751         %2<round_scalar_mask_op3>}"): ... this.
28752         * config/i386/subst.md (round_scalar_nimm_predicate): New.
28754 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28756         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
28757         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
28758         (_mm_maskz_sqrt_round_ss): New intrinsics.
28759         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
28760         (__builtin_ia32_sqrtsd_mask_round)
28761         (__builtin_ia32_sqrtss_mask_round): New builtins.
28762         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
28763         (__builtin_ia32_sqrtss_round): Remove.
28764         (__builtin_ia32_sqrtsd_mask_round)
28765         (__builtin_ia32_sqrtss_mask_round): New builtins.
28766         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
28767         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
28768         ((match_operand:VF_128 1 "vector_operand"
28769         "xBm,<round_constraint>")): Changed to ...
28770         ((match_operand:VF_128 1 "vector_operand"
28771         "xBm,<round_scalar_constraint>")): ... this.
28772         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
28773         %0, %2, %<iptr>1<round_op3>}): Changed to ...
28774         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
28775         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
28776         %<iptr>1<round_scalar_mask_op3>}): ... this.
28777         ((set_attr "prefix" "<round_prefix>")): Changed to ...
28778         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
28780 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
28782         PR target/84266
28783         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
28784         Cast vec_cmpeq result to correct type.
28785         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
28786         Cast vec_cmpgt result to correct type.
28788 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
28790         * final.c (final_scan_insn_1): Renamed from...
28791         (final_scan_insn): ... this.  New wrapper, to recover
28792         seen from the outermost call in recursive ones.
28793         * config/sparc/sparc.c (output_return): Drop seen from call.
28794         (output_sibcall): Likewise.
28795         * config/visium/visium.c (output_branch): Likewise.
28797 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
28799         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
28800         function label.
28802 2018-02-10  Alan Modra  <amodra@gmail.com>
28804         PR target/84300
28805         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
28806         Specify LR as an input.
28808 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
28810         PR sanitizer/83987
28811         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
28812         remove_member_access_dummy_vars): New functions.
28813         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
28814         lower_omp_1, execute_lower_omp): Use them.
28816         PR rtl-optimization/84308
28817         * shrink-wrap.c (spread_components): Release todo vector.
28819 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
28821         PR rtl-optimization/57193
28822         * ira-color.c (struct allocno_color_data): Add member
28823         conflict_allocno_hard_prefs.
28824         (update_conflict_allocno_hard_prefs): New.
28825         (bucket_allocno_compare_func): Add a preference based on
28826         conflict_allocno_hard_prefs.
28827         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
28828         (color_allocnos): Remove a dead code.  Initiate
28829         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
28831 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28833         PR target/84226
28834         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
28835         constraint from =wa to wa.  Avoid a subreg on the output operand,
28836         instead use a pseudo and subreg it in a move.
28837         (p9_xxbrd_<mode>): Changed to ...
28838         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
28839         (p9_xxbrd_v2df): New expander.
28840         (p9_xxbrw_<mode>): Changed to ...
28841         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
28842         (p9_xxbrw_v4sf): New expander.
28844 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28846         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
28848 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
28850         PR target/83926
28851         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
28852         multiply in 32-bit mode.
28853         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
28854         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
28855         mode.
28857 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28859         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
28860         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
28861         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
28862         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
28864 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
28866         PR lto/84213
28867         * dwarf2out.c (is_trivial_indirect_ref): New function.
28868         (dwarf2out_late_global_decl): Do not generate a location
28869         attribute for variables that have a non-trivial DECL_VALUE_EXPR
28870         and that are not defined in the current unit.
28872 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28874         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
28875         instead of a libcall for UNORDERED.
28877 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
28879         PR target/82641
28880         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
28881         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
28883 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28885         PR target/PR84295
28886         * config/s390/s390.c (s390_set_current_function): Invoke
28887         s390_indirect_branch_settings also if fndecl didn't change.
28889 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28891         * config/rs6000/rs6000.md (blockage): Set length to zero.
28893 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28895         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
28897 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28899         PR sanitizer/84285
28900         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
28901         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
28902         -static-lib*san.
28904         PR debug/84252
28905         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
28906         PARALLEL incoming that failed vt_get_decl_and_offset check.
28908         PR middle-end/84237
28909         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
28910         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
28911         TREE_READONLY bit.
28912         (get_variable_section): For decls in named .bss* sections pass true as
28913         second argument to bss_initializer_p.
28915 2018-02-09  Marek Polacek  <polacek@redhat.com>
28916             Jakub Jelinek  <jakub@redhat.com>
28918         PR c++/83659
28919         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
28920         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
28921         Sync some changes from cxx_fold_indirect_ref.
28923 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28925         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
28926         markers.
28927         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
28928         (BLOCK_INLINE_ENTRY_LABEL): New.
28929         (dwarf2out_var_location): Disregard inline entry markers.
28930         (inline_entry_data): New struct.
28931         (inline_entry_data_hasher): New hashtable type.
28932         (inline_entry_data_hasher::hash): New.
28933         (inline_entry_data_hasher::equal): New.
28934         (inline_entry_data_table): New variable.
28935         (add_high_low_attributes): Add DW_AT_entry_pc and
28936         DW_AT_GNU_entry_view attributes if a pending entry is found
28937         in inline_entry_data_table.  Add old entry_pc attribute only
28938         if debug nonbinding markers are disabled.
28939         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
28940         markers are enabled.
28941         (block_within_block_p, dwarf2out_inline_entry): New.
28942         (dwarf2out_finish): Check that no entries remained in
28943         inline_entry_data_table.
28944         * final.c (reemit_insn_block_notes): Handle inline entry notes.
28945         (final_scan_insn, notice_source_line): Likewise.
28946         (rest_of_clean_state): Skip inline entry markers.
28947         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
28948         markers.
28949         * gimple.c (gimple_build_debug_inline_entry): New.
28950         * gimple.h (enum gimple_debug_subcode): Add
28951         GIMPLE_DEBUG_INLINE_ENTRY.
28952         (gimple_build_debug_inline_entry): Declare.
28953         (gimple_debug_inline_entry_p): New.
28954         (gimple_debug_nonbind_marker_p): Adjust.
28955         * insn-notes.def (INLINE_ENTRY): New.
28956         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
28957         inline entry marker notes.
28958         (print_insn): Likewise.
28959         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
28960         (INSN_DEBUG_MARKER_KIND): Likewise.
28961         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
28962         * tree-inline.c (expand_call_inline): Build and insert
28963         debug_inline_entry stmt.
28964         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
28965         inline entry blocks early, if nonbind markers are enabled.
28966         (dump_scope_block): Dump fragment info.
28967         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
28968         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
28969         (gimple_build_debug_inline_entry): New.
28970         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
28971         Enable/disable inline entry points too.
28972         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
28973         (DEBUG_INSN): Describe inline entry markers.
28975         * common.opt (gvariable-location-views): New.
28976         (gvariable-location-views=incompat5): New.
28977         * config.in: Rebuilt.
28978         * configure: Rebuilt.
28979         * configure.ac: Test assembler for view support.
28980         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
28981         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
28982         * dwarf2out.c (var_loc_view): New typedef.
28983         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
28984         (dwarf2out_locviews_in_attribute): New.
28985         (dwarf2out_locviews_in_loclist): New.
28986         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
28987         (enum dw_line_info_opcode): Add LI_adv_address.
28988         (struct dw_line_info_table): Add view.
28989         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
28990         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
28991         (zero_view_p): New variable.
28992         (ZERO_VIEW_P): New macro.
28993         (output_asm_line_debug_info): New.
28994         (struct var_loc_node): Add view.
28995         (add_AT_view_list, AT_loc_list): New.
28996         (add_var_loc_to_decl): Add view param.  Test it against last.
28997         (new_loc_list): Add view params.  Record them.
28998         (AT_loc_list_ptr): Handle loc and view lists.
28999         (view_list_to_loc_list_val_node): New.
29000         (print_dw_val): Handle dw_val_class_view_list.
29001         (size_of_die): Likewise.
29002         (value_format): Likewise.
29003         (loc_list_has_views): New.
29004         (gen_llsym): Set vl_symbol too.
29005         (maybe_gen_llsym, skip_loc_list_entry): New.
29006         (dwarf2out_maybe_output_loclist_view_pair): New.
29007         (output_loc_list): Output view list or entries too.
29008         (output_view_list_offset): New.
29009         (output_die): Handle dw_val_class_view_list.
29010         (output_dwarf_version): New.
29011         (output_compilation_unit_header): Use it.
29012         (output_skeleton_debug_sections): Likewise.
29013         (output_rnglists, output_line_info): Likewise.
29014         (output_pubnames, output_aranges): Update version comments.
29015         (output_one_line_info_table): Output view numbers in asm comments.
29016         (dw_loc_list): Determine current endview, pass it to new_loc_list.
29017         Call maybe_gen_llsym.
29018         (loc_list_from_tree_1): Adjust.
29019         (add_AT_location_description): Create view list attribute if
29020         needed, check it's absent otherwise.
29021         (convert_cfa_to_fb_loc_list): Adjust.
29022         (maybe_emit_file): Call output_asm_line_debug_info for test.
29023         (dwarf2out_var_location): Reset views as needed.  Precompute
29024         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
29025         attribute.  Set view.
29026         (new_line_info_table): Reset next view.
29027         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
29028         (dwarf2out_source_line): Likewise.  Output view resets and labels to
29029         the assembler, or select appropriate line info opcodes.
29030         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
29031         (optimize_string_length): Catch it.  Adjust.
29032         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
29033         dw_val_class_view_list, and remove it if no longer needed.
29034         (hash_loc_list): Hash view numbers.
29035         (loc_list_hasher::equal): Compare them.
29036         (optimize_location_lists): Check whether a view list symbol is
29037         needed, and whether the locview attribute is present, and
29038         whether they match.  Remove the locview attribute if no longer
29039         needed.
29040         (index_location_lists): Call skip_loc_list_entry for test.
29041         (dwarf2out_finish): Call output_asm_line_debug_info for test.
29042         Use output_dwarf_version.
29043         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
29044         (struct dw_val_node): Add val_view_list.
29045         * final.c (SEEN_NEXT_VIEW): New.
29046         (set_next_view_needed): New.
29047         (clear_next_view_needed): New.
29048         (maybe_output_next_view): New.
29049         (final_start_function): Rename to...
29050         (final_start_function_1): ... this.  Take pointer to FIRST,
29051         add SEEN parameter.  Emit param bindings in the initial view.
29052         (final_start_function): Reintroduce SEEN-less interface.
29053         (final): Rename to...
29054         (final_1): ... this.  Take SEEN parameter.  Output final pending
29055         next view at the end.
29056         (final): Reintroduce seen-less interface.
29057         (final_scan_insn): Output pending next view before switching
29058         sections or ending a block.  Mark the next view as needed when
29059         outputting variable locations.  Notify debug backend of section
29060         changes, and of location view changes.
29061         (rest_of_handle_final): Adjust.
29062         * toplev.c (process_options): Autodetect value for debug variable
29063         location views option.  Warn on incompat5 without -gdwarf-5.
29064         * doc/invoke.texi (gvariable-location-views): New.
29065         (gvariable-location-views=incompat5): New.
29066         (gno-variable-location-views): New.
29068 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
29070         PR tree-optimization/84136
29071         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
29072         that the result of find_edge is non-NULL.
29074 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
29076         PR target/83008
29077         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
29078         storing integer register in SImode.  Fix cost of 256 and 512
29079         byte aligned SSE register store.
29081 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
29083         * config/i386/i386.c (ix86_multiplication_cost): Fix
29084         multiplication cost for TARGET_AVX512DQ.
29086 2018-02-08  Marek Polacek  <polacek@redhat.com>
29088         PR tree-optimization/84238
29089         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
29090         get_range_strlen.
29092 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
29094         PR tree-optimization/84265
29095         * tree-vect-stmts.c (vectorizable_store): Don't treat
29096         VMAT_CONTIGUOUS accesses as grouped.
29097         (vectorizable_load): Likewise.
29099 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
29101         PR tree-optimization/81635
29102         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
29103         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
29104         (test_round_for_mask): New functions.
29105         (wide_int_cc_tests): Call test_round_for_mask.
29106         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
29107         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
29108         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
29109         range returned by get_range_info.
29111 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
29113         PR ipa/81360
29114         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
29115         * symtab.c: Include builtins.h
29116         (symtab_node::output_to_lto_symbol_table_p): Move here
29117         from lto-streamer-out.c:output_symbol_p.
29118         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
29119         (output_symbol_p): Move all logic to symtab.c
29120         (produce_symtab): Update.
29122 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29124         * config/s390/s390-opts.h (enum indirect_branch): Define.
29125         * config/s390/s390-protos.h (s390_return_addr_from_memory)
29126         (s390_indirect_branch_via_thunk)
29127         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
29128         (enum s390_indirect_branch_type): Define.
29129         * config/s390/s390.c (struct s390_frame_layout, struct
29130         machine_function): Remove.
29131         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
29132         (indirect_branch_table_label_no, indirect_branch_table_name):
29133         Define variables.
29134         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
29135         (enum s390_indirect_branch_option): Define.
29136         (s390_return_addr_from_memory): New function.
29137         (s390_handle_string_attribute): New function.
29138         (s390_attribute_table): Add new attribute handler.
29139         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
29140         (s390_indirect_branch_via_thunk): New function.
29141         (s390_indirect_branch_via_inline_thunk): New function.
29142         (s390_function_ok_for_sibcall): When jumping via thunk disallow
29143         sibling call optimization for non z10 compiles.
29144         (s390_emit_call): Force indirect branch target to be a single
29145         register.  Add r1 clobber for non-z10 compiles.
29146         (s390_emit_epilogue): Emit return jump via return_use expander.
29147         (s390_reorg): Handle JUMP_INSNs as execute targets.
29148         (s390_option_override_internal): Perform validity checks for the
29149         new command line options.
29150         (s390_indirect_branch_attrvalue): New function.
29151         (s390_indirect_branch_settings): New function.
29152         (s390_set_current_function): Invoke s390_indirect_branch_settings.
29153         (s390_output_indirect_thunk_function):  New function.
29154         (s390_code_end): Implement target hook.
29155         (s390_case_values_threshold): Implement target hook.
29156         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
29157         macros.
29158         * config/s390/s390.h (struct s390_frame_layout)
29159         (struct machine_function): Move here from s390.c.
29160         (TARGET_INDIRECT_BRANCH_NOBP_RET)
29161         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
29162         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
29163         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
29164         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
29165         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
29166         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
29167         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
29168         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
29169         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
29170         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
29171         (mnemonic attribute): Add values which aren't recognized
29172         automatically.
29173         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
29174         pattern for branch conversion.  Fix mnemonic attribute.
29175         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
29176         indirect branch via thunk if requested.
29177         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
29178         ("*indirect_jump"): Disable for branch conversion using out of
29179         line thunks.
29180         ("indirect_jump_via_thunk<mode>_z10")
29181         ("indirect_jump_via_thunk<mode>")
29182         ("indirect_jump_via_inlinethunk<mode>_z10")
29183         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
29184         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
29185         ("casesi_jump_via_inlinethunk<mode>_z10")
29186         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
29187         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
29188         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
29189         ("*indirect2_jump"): Disable for branch conversion.
29190         ("casesi_jump"): Turn into expander and expand patterns for branch
29191         conversion.
29192         ("return_use"): New expander.
29193         ("*return"): Emit return via thunk and rename it to ...
29194         ("*return<mode>"): ... this one.
29195         * config/s390/s390.opt: Add new options and and enum for the
29196         option values.
29198 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
29200         * lra-constraints.c (match_reload): Unconditionally use
29201         gen_lowpart_SUBREG, rather than selecting between that
29202         and equivalent gen_rtx_SUBREG code.
29204 2018-02-08  Richard Biener  <rguenther@suse.de>
29206         PR tree-optimization/84233
29207         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
29208         changed flag instead of boguously re-using phi_inserted.
29210 2018-02-08  Martin Jambor  <mjambor@suse.cz>
29212         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
29213         static local variables.
29215 2018-02-08  Richard Biener  <rguenther@suse.de>
29217         PR tree-optimization/84278
29218         * tree-vect-stmts.c (vectorizable_store): When looking for
29219         smaller vector types to perform grouped strided loads/stores
29220         make sure the mode is supported by the target.
29221         (vectorizable_load): Likewise.
29223 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
29225         * config/aarch64/aarch64.c (aarch64_components_for_bb):
29226         Increase LDP/STP opportunities by adding adjacent callee-saves.
29228 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
29230         PR rtl-optimization/84068
29231         PR rtl-optimization/83459
29232         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
29234 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
29236         PR tree-optimization/84224
29237         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
29238         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
29239         non-zero arguments.
29241 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
29243         PR target/84113
29244         * config/rs6000/altivec.md (*restore_world): Remove LR use.
29245         * config/rs6000/predicates.md (restore_world_operation): Adjust op
29246         count, remove one USE.
29248 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
29250         * doc/install.texi (Configuration): Document the
29251         --with-long-double-format={ibm,ieee} PowerPC configuration
29252         options.
29254         PR target/84154
29255         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
29256         Convert from define_expand to be define_insn_and_split.  Rework
29257         float/double/_Float128 conversions to QI/HI/SImode to work with
29258         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
29259         conversions to QI/HImode types did a store and then a load to
29260         truncate the value.  For conversions to VSX registers, don't split
29261         the insn, instead emit the code directly.  Use the code iterator
29262         any_fix to combine signed and unsigned conversions.
29263         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
29264         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
29265         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29266         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
29267         (fix_<mode>di2_hw): Likewise.
29268         (fixuns_<mode>di2_hw): Likewise.
29269         (fix_<mode>si2_hw): Likewise.
29270         (fixuns_<mode>si2_hw): Likewise.
29271         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
29272         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29273         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
29274         fix<uns>_trunc<SFDF:mode>si2_p8.
29275         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
29276         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
29277         (fix<uns>_<mode>_mem): Likewise.
29278         (fctiw<u>z_<mode>_mem): Likewise.
29279         (fix<uns>_<mode>_mem): Likewise.
29280         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
29281         the register allocator from doing a direct move to the GPRs to do
29282         a store, and instead use the ISA 3.0 store byte/half-word from
29283         vector register instruction.  For IEEE 128-bit floating point,
29284         also optimize stores of 32-bit ints.
29285         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
29287 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
29289         * genextract.c (push_pathstr_operand): New function to support
29290         [a-zA-Z].
29291         (walk_rtx): Call push_pathstr_operand.
29292         (print_path): Support [a-zA-Z].
29294 2018-02-07  Richard Biener  <rguenther@suse.de>
29296         PR tree-optimization/84037
29297         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
29298         (cse_and_gimplify_to_preheader): Declare.
29299         (vect_get_place_in_interleaving_chain): Likewise.
29300         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29301         ivexpr_map.
29302         (_loop_vec_info::~_loop_vec_info): Delete it.
29303         (cse_and_gimplify_to_preheader): New function.
29304         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
29305         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
29306         (vectorizable_load): Likewise.  For grouped stores always base
29307         the IV on the first element.
29308         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
29309         condition before gimplifying.
29311 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
29313         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
29314         *DIV_EXPR and *MOD_EXPR.
29316 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
29318         PR target/84248
29319         * config/i386/i386.c (ix86_option_override_internal): Mask out
29320         the CF_SET bit when checking -fcf-protection.
29322 2018-02-07  Tom de Vries  <tom@codesourcery.com>
29324         PR libgomp/84217
29325         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
29326         enough.
29328 2018-02-07  Richard Biener  <rguenther@suse.de>
29330         PR tree-optimization/84204
29331         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
29332         this place.
29334         PR tree-optimization/84205
29335         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
29336         special-case isl_ast_op_zdiv_r.
29338         PR tree-optimization/84223
29339         * graphite-scop-detection.c (gather_bbs::before_dom_children):
29340         Only add conditions from within the region.
29341         (gather_bbs::after_dom_children): Adjust.
29343 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
29345         PR target/84209
29346         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
29347         * config/avr/avr.md: Only post-reload split REG-REG moves if
29348         either register is GENERAL_REG_P.
29350 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
29352         PR tree-optimization/84235
29353         * tree-ssa-scopedtables.c
29354         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
29355         if the subtraction is performed in floating point type where NaNs are
29356         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
29357         build 1.  Formatting fix.
29359 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
29361         PR target/84146
29362         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
29363         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
29364         and skip it regardless of bb boundaries.  Use CALL_P macro,
29365         don't test INSN_P (insn) together with CALL_P or JUMP_P check
29366         unnecessarily, formatting fix.
29368 2018-02-06  Michael Collison  <michael.collison@arm.com>
29370         * config/arm/thumb2.md:
29371         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
29372         (*thumb_mov_notscc): Ditto.
29374 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
29376         PR target/84154
29377         * config/rs6000/rs6000.md (su code attribute): Use "u" for
29378         unsigned_fix, not "s".
29380 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29382         * configure.ac (gcc_fn_eh_frame_ro): New function.
29383         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
29384         correct .eh_frame permissions.
29385         * configure: Regenerate.
29387 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
29389         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
29390         irrelevant options.
29392 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29394         * config/rs6000/rs6000.c (rs6000_option_override_internal):
29395         Display warning message for -mno-speculate-indirect-jumps.
29397 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
29399         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
29400         Undocumented.
29401         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
29403 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
29405         PR tree-optimization/84225
29406         * tree-eh.c (find_trapping_overflow): Only call
29407         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
29409 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
29411         PR target/84145
29412         * config/i386/i386.c: Reimplement the check of possible options
29413         -mibt/-mshstk conbination. Change error messages.
29414         * doc/invoke.texi: Fix a typo: remove extra '='.
29416 2018-02-06  Marek Polacek  <polacek@redhat.com>
29418         PR tree-optimization/84228
29419         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
29421 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
29423         PR target/82641
29424         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
29425         emitted arch directives.
29426         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
29427         __ARM_FEATURE_COPROC before changing architectures.
29429 2018-02-06  Richard Biener  <rguenther@suse.de>
29431         * config/i386/i386.c (print_reg): Fix typo.
29432         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
29434 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
29436         * configure: Regenerate.
29438 2018-02-05  Martin Sebor  <msebor@redhat.com>
29440         PR tree-optimization/83369
29441         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
29442         inlining context.
29444 2018-02-05  Martin Liska  <mliska@suse.cz>
29446         * doc/invoke.texi: Cherry-pick upstream r323995.
29448 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
29450         * ira.c (ira_init_register_move_cost): Adjust comment.
29452 2018-02-05  Martin Liska  <mliska@suse.cz>
29454         PR gcov-profile/84137
29455         * doc/gcov.texi: Fix typo in documentation.
29457 2018-02-05  Martin Liska  <mliska@suse.cz>
29459         PR gcov-profile/83879
29460         * doc/gcov.texi: Document necessity of --dynamic-list-data when
29461         using dlopen functionality.
29463 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
29465         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
29466         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
29467         _mm_maskz_range_ss, _mm_mask_range_round_ss,
29468         _mm_maskz_range_round_ss): New intrinsics.
29469         (__builtin_ia32_rangesd128_round)
29470         (__builtin_ia32_rangess128_round): Remove.
29471         (__builtin_ia32_rangesd128_mask_round,
29472         __builtin_ia32_rangess128_mask_round): New builtins.
29473         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
29474         __builtin_ia32_rangess128_round): Remove.
29475         (__builtin_ia32_rangesd128_mask_round,
29476         __builtin_ia32_rangess128_mask_round): New builtins.
29477         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
29478         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
29479         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
29480         "<round_saeonly_constraint>")): Changed to ...
29481         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
29482         "<round_saeonly_scalar_constraint>")): ... this.
29483         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
29484         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
29485         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
29486         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
29487         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
29489 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
29491         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
29492         options.
29493         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
29494         Remove all values except native, 8540 and 8548.
29496 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
29498         * config/i386/i386.c (ix86_output_function_return): Pass
29499         INVALID_REGNUM, instead of -1, as invalid register number to
29500         indirect_thunk_name and output_indirect_thunk.
29502 2018-02-02  Julia Koval  <julia.koval@intel.com>
29504         * config.gcc: Add -march=icelake.
29505         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
29506         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
29507         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
29508         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
29509         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
29510         (processor_target_table): Add icelake.
29511         (ix86_option_override_internal): Handle new PTAs.
29512         (get_builtin_code_for_version): Handle icelake.
29513         (M_INTEL_COREI7_ICELAKE): New.
29514         (fold_builtin_cpu): Handle icelake.
29515         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
29516         * doc/invoke.texi: Add -march=icelake.
29518 2018-02-02  Julia Koval  <julia.koval@intel.com>
29520         * config/i386/i386.c (ix86_option_override_internal): Change flags type
29521         to wide_int_bitmask.
29522         * wide-int-bitmask.h: New.
29524 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
29526         PR target/84066
29527         * config/i386/i386.md: Replace Pmode with word_mode in
29528         builtin_setjmp_setup and builtin_longjmp to support x32.
29530 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
29532         PR target/56010
29533         PR target/83743
29534         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
29535         #include "opts.h".
29536         (rs6000_supported_cpu_names): New static variable.
29537         (linux_cpu_translation_table): Likewise.
29538         (elf_platform) <cpu>: Define new static variable and use it.
29539         Translate kernel AT_PLATFORM name to canonical name if needed.
29540         Error if platform name is unknown.
29542 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
29544         PR target/84089
29545         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
29547 2018-02-01  Jeff Law  <law@redhat.com>
29549         PR target/84128
29550         * config/i386/i386.c (release_scratch_register_on_entry): Add new
29551         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
29552         the scratch if RELEASE_VIA_POP is false.
29553         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
29554         If we have to save a temporary register, decrement SIZE appropriately.
29555         Pass new arguments to release_scratch_register_on_entry.
29556         (ix86_adjust_stack_and_probe): Likewise.
29557         (ix86_emit_probe_stack_range): Pass new arguments to
29558         release_scratch_register_on_entry.
29560 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
29562         PR rtl-optimization/84157
29563         * combine.c (change_zero_ext): Use REG_P predicate in
29564         front of HARD_REGISTER_P predicate.
29566 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
29568         * config/avr/avr.c (avr_option_override): Move disabling of
29569         -fdelete-null-pointer-checks to...
29570         * common/config/avr/avr-common.c (avr_option_optimization_table):
29571         ...here.
29573 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29575         PR tree-optimization/81635
29576         * tree-data-ref.c (split_constant_offset_1): For types that
29577         wrap on overflow, try to use range info to prove that wrapping
29578         cannot occur.
29580 2018-02-01  Renlin Li  <renlin.li@arm.com>
29582         PR target/83370
29583         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
29584         TAILCALL_ADDR_REGS.
29585         (aarch64_register_move_cost): Likewise.
29586         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
29587         TAILCALL_ADDR_REGS.
29588         (REG_CLASS_NAMES): Likewise.
29589         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
29590         TAILCALL_ADDR_REGS. Remove IP registers.
29591         * config/aarch64/aarch64.md (Ucs): Update register constraint.
29593 2018-02-01  Richard Biener  <rguenther@suse.de>
29595         * domwalk.h (dom_walker::dom_walker): Add additional constructor
29596         for specifying RPO order and allow NULL for that.
29597         * domwalk.c (dom_walker::dom_walker): Likewise.
29598         (dom_walker::walk): Handle NULL RPO order.
29599         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
29600         in RPO order.
29601         (rewrite_update_dom_walker): Likewise.
29602         (mark_def_dom_walker): Likewise.
29604 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29606         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
29607         (aarch64_maybe_expand_sve_subreg_move): Declare.
29608         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
29609         * config/aarch64/predicates.md (aarch64_any_register_operand): New
29610         predicate.
29611         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
29612         that are semantically a reverse operation.
29613         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
29614         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
29615         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
29616         functions.
29617         (aarch64_can_change_mode_class): For big-endian, forbid changes
29618         between two SVE modes if they have different element sizes.
29620 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29622         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
29623         the TImode handling for big-endian targets.
29625 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29627         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
29628         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
29629         not just bytes.
29630         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
29631         Remove BSWAP handing for big-endian targets and use the form of
29632         LD1RQ appropariate for the mode.
29634 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29636         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
29637         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
29638         duplicated element.
29640 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
29642         PR tearget/83845
29643         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
29644         check for operands that need to go through aarch64_sve_reload_be.
29646 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
29648         PR tree-optimization/81661
29649         PR tree-optimization/84117
29650         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
29651         * tree-eh.c: Include gimplify.h.
29652         (find_trapping_overflow, replace_trapping_overflow,
29653         rewrite_to_non_trapping_overflow): New functions.
29654         * tree-vect-loop.c: Include tree-eh.h.
29655         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
29656         * tree-data-ref.c: Include tree-eh.h.
29657         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
29659 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
29661         PR rtl-optimization/84123
29662         * combine.c (change_zero_ext): Check if hard register satisfies
29663         can_change_dest_mode before calling gen_lowpart_SUBREG.
29665 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
29667         PR target/82444
29668         * ira.c (ira_init_register_move_cost): Remove assert.
29670 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29672         PR rtl-optimization/84071
29673         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
29674         * doc/tm.texi: Regenerate.
29676 2018-01-31  Richard Biener  <rguenther@suse.de>
29678         PR tree-optimization/84132
29679         * tree-data-ref.c (analyze_miv_subscript): Properly
29680         check whether evolution_function_is_affine_multivariate_p
29681         before calling gcd_of_steps_may_divide_p.
29683 2018-01-31  Julia Koval  <julia.koval@intel.com>
29685         PR target/83618
29686         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
29687         * config/i386/i386.md (rdpid_rex64) New.
29688         (rdpid): Make 32bit only.
29690 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
29692         PR lto/84105
29693         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
29694         an IDENTIFIER_NODE for FUNCTION_TYPE's.
29696 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29698         Revert
29699         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29701         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29703 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29705         PR rtl-optimization/84071
29706         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
29707         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
29709 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29711         * config/arc/arc.c (arc_handle_aux_attribute): New function.
29712         (arc_attribute_table): Add 'aux' attribute.
29713         (arc_in_small_data_p): Consider aux like variables.
29714         (arc_is_aux_reg_p): New function.
29715         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
29716         (arc_get_aux_arg): New function.
29717         (prepare_move_operands): Handle aux-register access.
29718         (arc_handle_aux_attribute): New function.
29719         * doc/extend.texi (ARC Variable attributes): Add subsection.
29721 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29723         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
29724         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
29725         (arc_attribute_table): Add 'uncached' attribute.
29726         (arc_print_operand): Print '.di' flag for uncached memory
29727         accesses.
29728         (arc_in_small_data_p): Do not consider for small data the uncached
29729         types.
29730         (arc_is_uncached_mem_p): New function.
29731         * config/arc/predicates.md (compact_store_memory_operand): Check
29732         for uncached memory accesses.
29733         (nonvol_nonimm_operand): Likewise.
29734         * doc/extend.texi (ARC Type Attribute): New subsection.
29736 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
29738         PR c/84100
29739         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
29740         falign-loops=): Add Optimization flag.
29742 2018-01-30  Jeff Law  <law@redhat.com>
29744         PR target/84064
29745         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
29746         INT_REGISTERS_SAVED.  Check it prior to calling
29747         get_scratch_register_on_entry.
29748         (ix86_adjust_stack_and_probe): Similarly.
29749         (ix86_emit_probe_stack_range): Similarly.
29750         (ix86_expand_prologue): Corresponding changes.
29752 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29754         PR target/40411
29755         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
29756         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
29758 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
29760         PR target/84112
29761         * lra-constraints.c (curr_insn_transform): Process AND in the
29762         address.
29764 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
29766         PR rtl-optimization/83986
29767         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
29768         dependence against last_pending_memory_flush in addition to
29769         pending_jump_insns.
29771 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
29773         PR tree-optimization/81611
29774         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
29775         copies.
29777 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29779         PR target/83758
29780         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
29781         a reg rtx.
29783 2018-01-30  Richard Biener  <rguenther@suse.de>
29784             Jakub Jelinek  <jakub@redhat.com>
29786         PR tree-optimization/84111
29787         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
29788         inner loops added during recursion, as they don't have up-to-date
29789         SSA form.
29791 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29793         PR ipa/81360
29794         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
29795         (can_inline_edge_by_limits_p): ... here.
29796         (can_early_inline_edge_p, check_callers,
29797         update_caller_keys, update_callee_keys, recursive_inlining,
29798         add_new_edges_to_heap, speculation_useful_p,
29799         inline_small_functions,
29800         inline_small_functions, flatten_function,
29801         inline_to_all_callers_1): Update.
29803 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29805         * profile-count.c (profile_count::combine_with_ipa_count): Handle
29806         zeros correctly.
29808 2018-01-30  Richard Biener  <rguenther@suse.de>
29810         PR tree-optimization/83008
29811         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
29812         invariant and constant vector uses in stmts when they need
29813         more than one stmt.
29815 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29817         PR bootstrap/84017
29818         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
29819         * configure: Regenerate.
29821 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29823         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
29824         pattern.
29825         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
29826         Use gen_rtx_REG rather than gen_lowpart.
29828 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29830         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
29831         rather than 0 when creating partial subregs.
29833 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29835         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
29836         of usage.
29838 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
29840         PR target/81550
29841         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
29842         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
29843         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
29844         flags.  This restores the settings used before the 2017-07-24.
29845         Turning off pre increment/decrement/modify allows IVOPTS to
29846         optimize DF/SF loops where the index is an int.
29848 2018-01-29  Richard Biener  <rguenther@suse.de>
29849             Kelvin Nilsen  <kelvin@gcc.gnu.org>
29851         PR bootstrap/80867
29852         * tree-vect-stmts.c (vectorizable_call): Don't call
29853         targetm.vectorize_builtin_md_vectorized_function if callee is
29854         NULL.
29856 2018-01-22  Carl Love  <cel@us.ibm.com>
29858         * doc/extend.tex: Fix typo in second arg in
29859         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
29861 2018-01-29  Richard Biener  <rguenther@suse.de>
29863         PR tree-optimization/84086
29864         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
29865         (flush_ssaname_freelist): When SSA names were released reset
29866         the SCEV hash table.
29868 2018-01-29  Richard Biener  <rguenther@suse.de>
29870         PR tree-optimization/84057
29871         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
29872         removed paths when removing edges.
29874 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
29876         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
29877         -mfunction-return=@var{choice}.
29879 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29881         PR diagnostic/84034
29882         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
29883         Handle CR like TAB.
29884         (layout::print_source_line): Likewise.
29885         (test_get_line_width_without_trailing_whitespace): Add test cases.
29887 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
29889         PR middle-end/84040
29890         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
29891         debug insns.
29893 2018-01-26  Jim Wilson  <jimw@sifive.com>
29895         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
29897         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
29898         specified.
29900 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29902         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
29903         and CMP + SUB-immediate -> SUBS.
29905 2018-01-26  Martin Sebor  <msebor@redhat.com>
29907         PR tree-optimization/83896
29908         * tree-ssa-strlen.c (get_string_len): Rename...
29909         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
29910         Avoid assuming length is constant.
29911         (handle_char_store): Use HOST_WIDE_INT for string length.
29913 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
29915         PR target/81763
29916         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
29917         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
29919 2018-01-26  Richard Biener  <rguenther@suse.de>
29921         PR rtl-optimization/84003
29922         * dse.c (record_store): Only record redundant stores when
29923         the earlier store aliases at least all accesses the later one does.
29925 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
29927         PR rtl-optimization/83985
29928         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
29929         REG_CFA_RESTORE insns.
29930         (delete_unmarked_insns): Don't ignore separate shrink wrapping
29931         REG_CFA_RESTORE insns here.
29933         PR c/83989
29934         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
29935         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
29937 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29939         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
29940         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
29941         (arc_init): Likewise.
29942         (arc_override_options): Likewise.
29943         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
29944         value.
29945         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
29946         support.
29947         * config/arc/arc.h (TARGET_DBNZ): Define.
29948         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
29949         properly set the tune attribute.
29950         (dbnz): Use TARGET_DBNZ guard.
29951         * config/arc/arc.opt (mtune): Add core3 option.
29953 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29955         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
29956         recognize new pic like addresses.
29957         (arc_delegitimize_address): Clean up.
29959 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29961         * config/arc/arc-arches.def: Option mrf16 valid for all
29962         architectures.
29963         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
29964         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
29965         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
29966         * config/arc/arc-tables.opt: Regenerate.
29967         * config/arc/arc.c (arc_conditional_register_usage): Handle
29968         reduced register file case.
29969         (arc_file_start): Set must have build attributes.
29970         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
29971         mrf16 option value.
29972         * config/arc/arc.opt (mrf16): Add new option.
29973         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
29974         * config/arc/genmultilib.awk: Handle new mrf16 option.
29975         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
29976         * config/arc/t-multilib: Regenerate.
29977         * doc/invoke.texi (ARC Options): Document mrf16 option.
29979 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29981         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
29982         * config/arc/arc.c (arc_handle_secure_attribute): New function.
29983         (arc_attribute_table): Add 'secure_call' attribute.
29984         (arc_print_operand): Print secure call operand.
29985         (arc_function_ok_for_sibcall): Don't optimize tail calls when
29986         secure.
29987         (arc_is_secure_call_p): New function.  * config/arc/arc.md
29988         (call_i): Add support for sjli instruction.
29989         (call_value_i): Likewise.
29990         * config/arc/constraints.md (Csc): New constraint.
29992 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29993             John Eric Martin  <John.Martin@emmicro-us.com>
29995         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
29996         * config/arc/arc.c (_arc_jli_section): New struct.
29997         (arc_jli_section): New type.
29998         (rc_jli_sections): New static variable.
29999         (arc_handle_jli_attribute): New function.
30000         (arc_attribute_table): Add jli_always and jli_fixed attribute.
30001         (arc_file_end): New function.
30002         (TARGET_ASM_FILE_END): Define.
30003         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
30004         (arc_add_jli_section): New function.
30005         (jli_call_scan): Likewise.
30006         (arc_reorg): Call jli_call_scan.
30007         (arc_output_addsi): Remove 'S' from printing asm operand.
30008         (arc_is_jli_call_p): New function.
30009         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
30010         operand.
30011         (movhi_insn): Likewise.
30012         (movsi_insn): Likewise.
30013         (movsi_set_cc_insn): Likewise.
30014         (loadqi_update): Likewise.
30015         (load_zeroextendqisi_update): Likewise.
30016         (load_signextendqisi_update): Likewise.
30017         (loadhi_update): Likewise.
30018         (load_zeroextendhisi_update): Likewise.
30019         (load_signextendhisi_update): Likewise.
30020         (loadsi_update): Likewise.
30021         (loadsf_update): Likewise.
30022         (movsicc_insn): Likewise.
30023         (bset_insn): Likewise.
30024         (bxor_insn): Likewise.
30025         (bclr_insn): Likewise.
30026         (bmsk_insn): Likewise.
30027         (bicsi3_insn): Likewise.
30028         (cmpsi_cc_c_insn): Likewise.
30029         (movsi_ne): Likewise.
30030         (movsi_cond_exec): Likewise.
30031         (clrsbsi2): Likewise.
30032         (norm_f): Likewise.
30033         (normw): Likewise.
30034         (swap): Likewise.
30035         (divaw): Likewise.
30036         (flag): Likewise.
30037         (sr): Likewise.
30038         (kflag): Likewise.
30039         (ffs): Likewise.
30040         (ffs_f): Likewise.
30041         (fls): Likewise.
30042         (call_i): Remove 'S' asm letter, add jli instruction.
30043         (call_value_i): Likewise.
30044         * config/arc/arc.op (mjli-always): New option.
30045         * config/arc/constraints.md (Cji): New constraint.
30046         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
30047         operand.
30048         (subsf3_fpx): Likewise.
30049         (mulsf3_fpx): Likewise.
30050         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
30051         asm operand.
30052         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
30053         function attrbutes.
30054         * doc/invoke.texi (ARC): Document mjli-always option.
30056 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
30058         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
30059         avoid addition with 0 and use incw and decw where possible.
30061 2018-01-26  Richard Biener  <rguenther@suse.de>
30063         PR tree-optimization/81082
30064         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
30065         association if it requires casting to unsigned.
30066         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
30067         from fold_plusminus_mult_expr to catch important cases late when
30068         range info is available.
30070 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30072         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
30073         * configure.ac (hidden_linkonce): New test.
30074         * configure: Regenerate.
30075         * config.in: Regenerate.
30077 2018-01-26  Julia Koval  <julia.koval@intel.com>
30079         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
30080         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
30081         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
30082         _mm_mask_bitshuffle_epi64_mask): Fix type.
30083         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
30084         USI_FTYPE_V4DI_V4DI_USI): Remove.
30085         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
30086         __builtin_ia32_vpshufbitqmb256_mask,
30087         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
30088         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
30089         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
30091 2018-01-26  Alan Modra  <amodra@gmail.com>
30093         PR target/84033
30094         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
30095         UNSPEC_VBPERMQ.  Sort other unspecs.
30097 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
30099         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
30101 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
30103         PR middle-end/83055
30104         * predict.c (drop_profile): Do not push/pop cfun; update also
30105         node->count.
30106         (handle_missing_profiles): Fix logic looking for zero profiles.
30108 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
30110         PR middle-end/83977
30111         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
30112         on functions with #pragma omp declare simd or functions with simd
30113         attribute.
30114         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
30115         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
30116         Remove trailing \n from warning_at calls.
30118 2018-01-25  Tom de Vries  <tom@codesourcery.com>
30120         PR target/84028
30121         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30122         for neutered workers.
30124 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
30126         PR target/68467
30127         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
30128         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
30130 2018-01-24  Jeff Law  <law@redhat.com>
30132         PR target/83994
30133         * i386.c (get_probe_interval): Move to earlier point.
30134         (ix86_compute_frame_layout): If -fstack-clash-protection and
30135         the frame is larger than the probe interval, then use pushes
30136         to save registers rather than reg->mem moves.
30137         (ix86_expand_prologue): Remove conditional for int_registers_saved
30138         assertion.
30140 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
30142         PR target/84014
30143         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
30144         min/max for never referenced object.
30146 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
30148         PR middle-end/83977
30149         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
30150         here.
30151         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
30152         attributes from DECL_ATTRIBUTES (decl) without affecting
30153         DECL_ATTRIBUTES (current_function_decl).
30154         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
30155         functions with non-NULL DECL_ABSTRACT_ORIGIN.
30157 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
30159         PR tree-optimization/83979
30160         * fold-const.c (fold_comparison): Use constant_boolean_node
30161         instead of boolean_{true,false}_node.
30163 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
30165         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
30166         with zero counts.
30168 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30170         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30171         Simplify the clause that sets the length attribute.
30172         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30173         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
30174         clause that sets the length attribute.
30175         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30177 2018-01-24  Tom de Vries  <tom@codesourcery.com>
30179         PR target/83589
30180         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
30181         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
30182         Add strict parameter.
30183         (prevent_branch_around_nothing): Insert dummy insn between branch to
30184         label and label with no ptx insn inbetween.
30185         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
30187 2018-01-24  Tom de Vries  <tom@codesourcery.com>
30189         PR target/81352
30190         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30191         for neutered threads in warp.
30192         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
30194 2018-01-24  Richard Biener  <rguenther@suse.de>
30196         PR tree-optimization/83176
30197         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
30198         operands.
30200 2018-01-24  Richard Biener  <rguenther@suse.de>
30202         PR tree-optimization/82819
30203         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
30204         code generating pluses that are no-ops in the target precision.
30206 2018-01-24  Richard Biener  <rguenther@suse.de>
30208         PR middle-end/84000
30209         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
30211 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
30213         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
30214         to merge probabilities.
30215         * predict.c (probably_never_executed): Also mark as cold functions
30216         with global 0 profile and guessed local profile.
30217         * profile-count.c (profile_probability::combine_with_count): New
30218         member function.
30219         * profile-count.h (profile_probability::operator*,
30220         profile_probability::operator*=, profile_probability::operator/,
30221         profile_probability::operator/=): Reduce precision to adjusted
30222         and set value to guessed on contradictory divisions.
30223         (profile_probability::combine_with_freq): Remove.
30224         (profile_probability::combine_wiht_count): Declare.
30225         (profile_count::force_nonzero):: Set to adjusted.
30226         (profile_count::probability_in):: Set quality to adjusted.
30227         * tree-ssa-tail-merge.c (replace_block_by): Use
30228         combine_with_count.
30230 2018-01-23  Andrew Waterman  <andrew@sifive.com>
30231             Jim Wilson  <jimw@sifive.com>
30233         * config/riscv/riscv.c (riscv_stack_boundary): New.
30234         (riscv_option_override): Set riscv_stack_boundary.  Handle
30235         riscv_preferred_stack_boundary_arg.
30236         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
30237         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
30238         (STACK_BOUNDARY): Set to riscv_stack_boundary.
30239         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
30240         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
30241         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
30243 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
30245         PR target/83905
30246         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
30247         of struct ix86_frame.
30248         (ix86_expand_epilogue): Likewise.  Add a local variable for
30249         the reg_save_offset field in struct ix86_frame.
30251 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
30253         PR tree-optimization/82604
30254         * tree-loop-distribution.c (enum partition_kind): New enum item
30255         PKIND_PARTIAL_MEMSET.
30256         (partition_builtin_p): Support above new enum item.
30257         (generate_code_for_partition): Ditto.
30258         (compute_access_range): Differentiate cases that equality can be
30259         proven at all loops, the innermost loops or no loops.
30260         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
30261         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
30262         (finalize_partitions, distribute_loop): Don't fuse partition of
30263         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
30264         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
30265         parloop is enabled.
30267 2018-01-23  Martin Liska  <mliska@suse.cz>
30269         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
30270         order to ignore the predictor.
30271         (PRED_POLYMORPHIC_CALL): Likewise.
30272         (PRED_RECURSIVE_CALL): Likewise.
30274 2018-01-23  Martin Liska  <mliska@suse.cz>
30276         * tree-profile.c (tree_profiling): Print function header to
30277         aware reader which function we are working on.
30278         * value-prof.c (gimple_find_values_to_profile): Do not print
30279         not interesting value histograms.
30281 2018-01-23  Martin Liska  <mliska@suse.cz>
30283         * profile-count.h (enum profile_quality): Add
30284         profile_uninitialized as the first value. Do not number values
30285         as they are zero based.
30286         (profile_count::verify): Update sanity check.
30287         (profile_probability::verify): Likewise.
30289 2018-01-23  Nathan Sidwell  <nathan@acm.org>
30291         * doc/invoke.texi (ffor-scope): Deprecate.
30293 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
30295         PR tree-optimization/83510
30296         * domwalk.c (set_all_edges_as_executable): New function.
30297         (dom_walker::dom_walker): Convert bool param
30298         "skip_unreachable_blocks" to enum reachability.  Move setup of
30299         edge flags to set_all_edges_as_executable and only do it when
30300         reachability is REACHABLE_BLOCKS.
30301         * domwalk.h (enum dom_walker::reachability): New enum.
30302         (dom_walker::dom_walker): Convert bool param
30303         "skip_unreachable_blocks" to enum reachability.
30304         (set_all_edges_as_executable): New decl.
30305         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
30306         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
30307         "reachability".
30308         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
30309         but converting true to REACHABLE_BLOCKS.
30310         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
30311         * tree-vrp.c
30312         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
30313         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
30314         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
30315         REACHABLE_BLOCKS.
30316         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
30317         if check_all_array_refs will be called.
30319 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
30321         * tree.c (selftest::test_location_wrappers): Add more test
30322         coverage.
30324 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
30326         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
30327         (selftest::test_bit_in_range): Likewise.
30329 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
30331         PR testsuite/83888
30332         * doc/sourcebuild.texi (vect_float): Say that the selector
30333         only describes the situation when -funsafe-math-optimizations is on.
30334         (vect_float_strict): Document.
30336 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
30338         PR tree-optimization/83965
30339         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
30340         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
30341         instead of checking only for a reduction.
30342         (vect_recog_widen_sum_pattern): Likewise.
30344 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
30346         * predict.c (probably_never_executed): Only use precise profile info.
30347         (compute_function_frequency): Skip after inlining hack since we now
30348         have quality checking.
30350 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
30352         * profile-count.h (profile_probability::very_unlikely,
30353         profile_probability::unlikely, profile_probability::even): Set
30354         precision to guessed.
30356 2018-01-23  Richard Biener  <rguenther@suse.de>
30358         PR tree-optimization/83963
30359         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
30360         Properly terminate dominator walk when crossing the exit edge not
30361         when visiting its source block.
30363 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
30365         PR c++/83918
30366         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
30367         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
30369 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
30371         PR tree-optimization/83957
30372         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
30373         semicolon after for body surrounded by braces.
30375         PR tree-optimization/83081
30376         * profile-count.h (profile_probability::split): New method.
30377         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
30378         Use profile_probability::split.
30379         (do_compare_rtx_and_jump): Fix adjustment of probabilities
30380         when splitting a single conditional jump into 2.
30382 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
30384         PR tree-optimization/69452
30385         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
30386         decl.
30388 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30390         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
30391         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
30392         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
30394 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30396         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
30397         declaration.
30398         * config/rl78/rl78.md (movdi): New define_expand.
30399         * config/rl78/rl78.c (rl78_split_movdi): New function.
30401 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30403         PR target/83862
30404         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
30405         no longer used.
30406         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
30407         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
30408         128-bit to produce an UNSPEC move to get the double word with the
30409         signbit and then a shift directly to do signbit.
30410         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
30411         implementation with a new version that just does either a direct
30412         move or a regular move.  Move memory interface to separate insns.
30413         Move insns so they are next to the expander.
30414         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
30415         with signbit move.  Split big and little endian case.
30416         (signbit<mode>2_dm_mem_le): Likewise.
30417         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
30418         (signbit<mode>2_dm2): Likewise.
30420 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30422         * config/rl78/rl78.md (anddi3): New define_expand.
30424 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30426         * config/rl78/rl78.md (umindi3): New define_expand.
30428 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30430         * config/rl78/rl78.md (smindi3): New define_expand.
30432 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30434         * config/rl78/rl78.md (smaxdi3): New define_expand.
30436 2018-01-22  Carl Love  <cel@us.ibm.com>
30438         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
30439         LVX_V1TI): Add macro expansion.
30440         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
30441         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
30442         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
30443         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
30444         Change check to determine if the instruction is a byte reversing
30445         entry.  Fix typo in comment.
30446         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
30447         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
30448         Add def_builtin calls for new builtins.
30449         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
30450         Add define_insn expansion.
30452 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30454         * config/rl78/rl78.md (umaxdi3): New define_expand.
30456 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
30458         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
30459         for non-QImode registers.
30461 2018-01-22  Richard Biener  <rguenther@suse.de>
30463         PR tree-optimization/83963
30464         * graphite-scop-detection.c (scop_detection::get_sese): Delay
30465         including the loop exit block.
30466         (scop_detection::merge_sese): Likewise.
30467         (scop_detection::add_scop): Do it here instead.
30469 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30471         * doc/sourcebuild.texi (arm_softfloat): Document.
30473 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
30475         PR gcc/77734
30476         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
30477         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
30478         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
30480 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30481             David Edelsohn  <dje.gcc@gmail.com>
30483         PR target/83946
30484         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30485         Change "crset eq" to "crset 2".
30486         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30487         (*call_indirect_aix<mode>_nospec): Likewise.
30488         (*call_value_indirect_aix<mode>_nospec): Likewise.
30489         (*call_indirect_elfv2<mode>_nospec): Likewise.
30490         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
30491         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
30492         change assembly output from . to $.
30493         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30494         (indirect_jump<mode>_nospec): Change assembly output from . to $.
30495         (*tablejump<mode>_internal1_nospec): Likewise.
30497 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
30499         PR target/80870
30500         * config/sh/sh_optimize_sett_clrt.cc:
30501         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
30503 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
30505         PR tree-optimization/83940
30506         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
30507         offset_dt to vect_constant_def rather than vect_unknown_def_type.
30508         (vect_check_load_store_mask): Add a mask_dt_out parameter and
30509         use it to pass back the definition type.
30510         (vect_check_store_rhs): Likewise rhs_dt_out.
30511         (vect_build_gather_load_calls): Add a mask_dt argument and use
30512         it instead of a call to vect_is_simple_use.
30513         (vectorizable_store): Update calls to vect_check_load_store_mask
30514         and vect_check_store_rhs.  Use the dt returned by the latter instead
30515         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
30516         instead of calls to vect_is_simple_use.  Pass the scalar rather
30517         than the vector operand to vect_is_simple_use when handling
30518         second and subsequent copies of an rhs value.
30519         (vectorizable_load): Update calls to vect_check_load_store_mask
30520         and vect_build_gather_load_calls.  Use the cached mask_dt and
30521         gs_info.offset_dt instead of calls to vect_is_simple_use.
30523 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
30525         PR middle-end/83945
30526         * tree-emutls.c: Include gimplify.h.
30527         (lower_emutls_2): New function.
30528         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
30529         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
30530         it before further processing.
30532         PR target/83930
30533         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
30534         UINTVAL (trueop1) instead of INTVAL (op1).
30536 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
30538         PR debug/81570
30539         PR debug/83728
30540         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
30541         INCOMING_FRAME_SP_OFFSET if not defined.
30542         (scan_trace): Add ENTRY argument.  If true and
30543         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
30544         emit a note to adjust the CFA offset.
30545         (create_cfi_notes): Adjust scan_trace callers.
30546         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
30547         INCOMING_FRAME_SP_OFFSET in the CIE.
30548         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
30549         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
30550         Likewise.
30551         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
30552         * doc/tm.texi: Regenerated.
30554 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30556         PR rtl-optimization/83147
30557         * lra-constraints.c (remove_inheritance_pseudos): Use
30558         lra_substitute_pseudo_within_insn.
30560 2018-01-19  Tom de Vries  <tom@codesourcery.com>
30561             Cesar Philippidis  <cesar@codesourcery.com>
30563         PR target/83920
30564         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
30566 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
30568         PR target/83790
30569         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
30570         spaces for function labels.
30572 2018-01-19  Martin Liska  <mliska@suse.cz>
30574         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
30575         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
30576         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
30577         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
30578         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
30579         (PRED_CONST_RETURN): Change from 69 to 65.
30580         (PRED_NULL_RETURN): Change from 91 to 71.
30581         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
30582         (PRED_LOOP_GUARD): Change from 66 to 73.
30584 2018-01-19  Martin Liska  <mliska@suse.cz>
30586         * predict.c (predict_insn_def): Add new assert.
30587         (struct branch_predictor): Change type to signed integer.
30588         (test_prediction_value_range): Amend test to cover
30589         PROB_UNINITIALIZED.
30590         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
30591         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
30592         (PRED_LOOP_ITERATIONS_MAX): Likewise.
30593         (PRED_LOOP_IV_COMPARE): Likewise.
30594         * predict.h (PROB_UNINITIALIZED): Define new constant.
30596 2018-01-19  Martin Liska  <mliska@suse.cz>
30598         * predict.c (dump_prediction): Add new format for
30599         analyze_brprob.py script which is enabled with -details
30600         suboption.
30601         * profile-count.h (precise_p): New function.
30603 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
30605         PR tree-optimization/83922
30606         * tree-vect-loop.c (vect_verify_full_masking): Return false if
30607         there are no statements that need masking.
30608         (vect_active_double_reduction_p): New function.
30609         (vect_analyze_loop_operations): Use it when handling phis that
30610         are not in the loop header.
30612 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
30614         PR tree-optimization/83914
30615         * tree-vect-loop.c (vectorizable_induction): Don't convert
30616         init_expr or apply the peeling adjustment for inductions
30617         that are nested within the vectorized loop.
30619 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30621         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
30622         instead of NEG.
30624 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
30626         PR sanitizer/81715
30627         PR testsuite/83882
30628         * function.h (gimplify_parameters): Add gimple_seq * argument.
30629         * function.c: Include gimple.h and options.h.
30630         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
30631         for the added local temporaries if needed.
30632         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
30633         if there are any parameter cleanups, wrap whole body into a
30634         try/finally with the cleanups.
30636 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
30638         PR target/82964
30639         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
30640         Use GET_MODE_CLASS for scalar floating point.
30642 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
30644         PR ipa/82256
30645         patch by PaX Team
30646         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
30647         Fix call of call_cgraph_insertion_hooks.
30649 2018-01-18  Martin Sebor  <msebor@redhat.com>
30651         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
30653 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
30655         PR ipa/83619
30656         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
30657         frequencies.
30659 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
30661         PR other/70268
30662         * common.opt: (-ffile-prefix-map): New option.
30663         * opts.c (common_handle_option): Defer it.
30664         * opts-global.c (handle_common_deferred_options): Handle it.
30665         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
30666         * file-prefix-map.h: New file.
30667         (remap_debug_filename, add_debug_prefix_map): ...here.
30668         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
30669         * final.c (debug_prefix_map, add_debug_prefix_map
30670         remap_debug_filename): Move to...
30671         * file-prefix-map.c: New file.
30672         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
30673         generalize, get rid of alloca(), use strrchr() instead of strchr().
30674         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
30675         Implement in terms of add_prefix_map().
30676         (remap_macro_filename, remap_debug_filename): Implement in term of
30677         remap_filename().
30678         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
30679         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
30680         * dbxout.c: Include file-prefix-map.h.
30681         * varasm.c: Likewise.
30682         * vmsdbgout.c: Likewise.
30683         * xcoffout.c: Likewise.
30684         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
30685         * doc/cppopts.texi (-fmacro-prefix-map): Document.
30686         * doc/invoke.texi (-ffile-prefix-map): Document.
30687         (-fdebug-prefix-map): Update description.
30689 2018-01-18  Martin Liska  <mliska@suse.cz>
30691         * config/i386/i386.c (indirect_thunk_name): Document that also
30692         lfence is emitted.
30693         (output_indirect_thunk): Document why both instructions
30694         (pause and lfence) are generated.
30696 2018-01-18  Richard Biener  <rguenther@suse.de>
30698         PR tree-optimization/83887
30699         * graphite-scop-detection.c
30700         (scop_detection::get_nearest_dom_with_single_entry): Remove.
30701         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
30702         (scop_detection::merge_sese): Re-implement with a flood-fill
30703         algorithm that properly finds a SESE region if it exists.
30705 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
30707         PR c/61240
30708         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
30709         pointer_diff optimizations use view_convert instead of convert.
30711 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30713         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30714         Generate different code for -mno-speculate-indirect-jumps.
30715         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30716         (*call_indirect_aix<mode>): Disable for
30717         -mno-speculate-indirect-jumps.
30718         (*call_indirect_aix<mode>_nospec): New define_insn.
30719         (*call_value_indirect_aix<mode>): Disable for
30720         -mno-speculate-indirect-jumps.
30721         (*call_value_indirect_aix<mode>_nospec): New define_insn.
30722         (*sibcall_nonlocal_sysv<mode>): Generate different code for
30723         -mno-speculate-indirect-jumps.
30724         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30726 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
30728         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
30729         long double type, set the flags for noting the default long double
30730         type, even if we don't pass or return a long double type.
30732 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
30734         PR ipa/83051
30735         * ipa-inline.c (flatten_function): Do not overwrite final inlining
30736         failure.
30738 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
30740         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
30741         support for merge[hl].
30742         (fold_mergehl_helper): New helper function.
30743         (tree-vector-builder.h): New #include for tree_vector_builder usage.
30744         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
30745         (altivec_vmrglw_direct): Add xxmrglw insn.
30747 2018-01-17  Andrew Waterman  <andrew@sifive.com>
30749         * config/riscv/riscv.c (riscv_conditional_register_usage): If
30750         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
30752 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
30754         PR lto/83121
30755         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
30756         call the lto_location_cache before reading the
30757         DECL_SOURCE_LOCATION of the types.
30759 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
30760             Richard Sandiford  <richard.sandiford@linaro.org>
30762         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
30763         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
30764         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
30765         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
30766         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
30767         Add declaration.
30768         * config/aarch64/constraints.md (aarch64_movti_operand):
30769         Limit immediates.
30770         * config/aarch64/predicates.md (Uti): Add new constraint.
30772 2018-01-17  Carl Love  <cel@us.ibm.com>
30774         * config/rs6000/vsx.md (define_expand xl_len_r,
30775         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
30776         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
30777         lxvll.
30778         (define_expand, define_insn): Move the shift left from  the
30779         define_insn to the define_expand for lxvl and stxvl instructions.
30780         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
30781         and XL_LEN_R definitions to PURE.
30783 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
30785         * config/i386/i386.c (indirect_thunk_name): Declare regno
30786         as unsigned int.  Compare regno with INVALID_REGNUM.
30787         (output_indirect_thunk): Ditto.
30788         (output_indirect_thunk_function): Ditto.
30789         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
30790         in the call to output_indirect_thunk_function.
30792 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
30794         PR middle-end/83884
30795         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
30796         rather than the size of inner_type to determine the stack slot size
30797         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
30799 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
30801         PR target/83546
30802         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
30803         to PTA_SILVERMONT.
30805 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
30807         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
30808         endian Linux systems to optionally enable multilibs for selecting
30809         the long double type if the user configured an explicit type.
30810         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
30811         have no long double multilibs if not defined.
30812         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
30813         warn if the user used -mabi={ieee,ibm}longdouble and we built
30814         multilibs for long double.
30815         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
30816         appropriate multilib option.
30817         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
30818         multilib options.
30819         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
30820         for building long double multilibs.
30821         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
30823 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
30825         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
30826         copies.
30828         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
30829         64 bits.
30830         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
30831         128 bits.
30833         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
30834         variables.
30836         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
30837         return value.
30839 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30841         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
30842         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
30844 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
30846         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
30847         different rtl trees depending on TARGET_64BIT.
30848         (rs6000_gen_lvx): Likewise.
30850 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30852         * config/visium/visium.md (nop): Tweak comment.
30853         (hazard_nop): Likewise.
30855 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30857         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
30858         -mspeculate-indirect-jumps.
30859         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
30860         for -mno-speculate-indirect-jumps.
30861         (*call_indirect_elfv2<mode>_nospec): New define_insn.
30862         (*call_value_indirect_elfv2<mode>): Disable for
30863         -mno-speculate-indirect-jumps.
30864         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
30865         (indirect_jump): Emit different RTL for
30866         -mno-speculate-indirect-jumps.
30867         (*indirect_jump<mode>): Disable for
30868         -mno-speculate-indirect-jumps.
30869         (*indirect_jump<mode>_nospec): New define_insn.
30870         (tablejump): Emit different RTL for
30871         -mno-speculate-indirect-jumps.
30872         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
30873         (tablejumpsi_nospec): New define_expand.
30874         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
30875         (tablejumpdi_nospec): New define_expand.
30876         (*tablejump<mode>_internal1): Disable for
30877         -mno-speculate-indirect-jumps.
30878         (*tablejump<mode>_internal1_nospec): New define_insn.
30879         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
30880         option.
30882 2018-01-16  Artyom Skrobov tyomitch@gmail.com
30884         * caller-save.c (insert_save): Drop unnecessary parameter.  All
30885         callers updated.
30887 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30888             Richard Biener  <rguenth@suse.de>
30890         PR libgomp/83590
30891         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
30892         return early, inline manually is_gimple_sizepos.  Make sure if we
30893         call gimplify_expr we don't end up with a gimple constant.
30894         * tree.c (variably_modified_type_p): Don't return true for
30895         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
30896         * gimplify.h (is_gimple_sizepos): Remove.
30898 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30900         PR tree-optimization/83857
30901         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
30902         vectorizable_live_operation for pure SLP statements.
30903         (vectorizable_live_operation): Handle PHIs.
30905 2018-01-16  Richard Biener  <rguenther@suse.de>
30907         PR tree-optimization/83867
30908         * tree-vect-stmts.c (vect_transform_stmt): Precompute
30909         nested_in_vect_loop_p since the scalar stmt may get invalidated.
30911 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30913         PR c/83844
30914         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
30915         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
30916         If off is not INTEGER_CST, issue a may not be aligned warning
30917         rather than isn't aligned.  Use isn%'t rather than isn't.
30918         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
30919         into MULT_EXPR.
30920         <case MULT_EXPR>: Improve the case when bottom and one of the
30921         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
30922         operand, in that case check if the other operand is multiple of
30923         bottom divided by the INTEGER_CST operand.
30925 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30927         PR target/83858
30928         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
30929         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
30930         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
30931         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
30932         * config/pa/pa.c (pa_function_arg_advance): Likewise.
30933         (pa_function_arg, pa_arg_partial_bytes): Likewise.
30934         (pa_function_arg_size): New function.
30936 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30938         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
30939         in a separate statement.
30941 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30943         PR tree-optimization/83847
30944         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
30945         group gathers and scatters.
30947 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30949         PR rtl-optimization/86620
30950         * params.def (max-sched-ready-insns): Bump minimum value to 1.
30952         PR rtl-optimization/83213
30953         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
30954         to last if both are JUMP_INSNs.
30956         PR tree-optimization/83843
30957         * gimple-ssa-store-merging.c
30958         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
30959         store_immediate_info for bswap/nop orig_stores.
30961 2018-01-15  Andrew Waterman  <andrew@sifive.com>
30963         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
30964         !TARGET_MUL.
30965         <UDIV>: Increase cost if !TARGET_DIV.
30967 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
30969         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
30970         (define_attr "cr_logical_3op"): New.
30971         (cceq_ior_compare): Adjust.
30972         (cceq_ior_compare_complement): Adjust.
30973         (*cceq_rev_compare): Adjust.
30974         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
30975         (is_cracked_insn): Adjust.
30976         (insn_must_be_first_in_group): Adjust.
30977         * config/rs6000/40x.md: Adjust.
30978         * config/rs6000/440.md: Adjust.
30979         * config/rs6000/476.md: Adjust.
30980         * config/rs6000/601.md: Adjust.
30981         * config/rs6000/603.md: Adjust.
30982         * config/rs6000/6xx.md: Adjust.
30983         * config/rs6000/7450.md: Adjust.
30984         * config/rs6000/7xx.md: Adjust.
30985         * config/rs6000/8540.md: Adjust.
30986         * config/rs6000/cell.md: Adjust.
30987         * config/rs6000/e300c2c3.md: Adjust.
30988         * config/rs6000/e500mc.md: Adjust.
30989         * config/rs6000/e500mc64.md: Adjust.
30990         * config/rs6000/e5500.md: Adjust.
30991         * config/rs6000/e6500.md: Adjust.
30992         * config/rs6000/mpc.md: Adjust.
30993         * config/rs6000/power4.md: Adjust.
30994         * config/rs6000/power5.md: Adjust.
30995         * config/rs6000/power6.md: Adjust.
30996         * config/rs6000/power7.md: Adjust.
30997         * config/rs6000/power8.md: Adjust.
30998         * config/rs6000/power9.md: Adjust.
30999         * config/rs6000/rs64.md: Adjust.
31000         * config/rs6000/titan.md: Adjust.
31002 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31004         * config/i386/predicates.md (indirect_branch_operand): Rewrite
31005         ix86_indirect_branch_register logic.
31007 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31009         * config/i386/constraints.md (Bs): Update
31010         ix86_indirect_branch_register check.  Don't check
31011         ix86_indirect_branch_register with GOT_memory_operand.
31012         (Bw): Likewise.
31013         * config/i386/predicates.md (GOT_memory_operand): Don't check
31014         ix86_indirect_branch_register here.
31015         (GOT32_symbol_operand): Likewise.
31017 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31019         * config/i386/predicates.md (constant_call_address_operand):
31020         Rewrite ix86_indirect_branch_register logic.
31021         (sibcall_insn_operand): Likewise.
31023 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31025         * config/i386/constraints.md (Bs): Replace
31026         ix86_indirect_branch_thunk_register with
31027         ix86_indirect_branch_register.
31028         (Bw): Likewise.
31029         * config/i386/i386.md (indirect_jump): Likewise.
31030         (tablejump): Likewise.
31031         (*sibcall_memory): Likewise.
31032         (*sibcall_value_memory): Likewise.
31033         Peepholes of indirect call and jump via memory: Likewise.
31034         * config/i386/i386.opt: Likewise.
31035         * config/i386/predicates.md (indirect_branch_operand): Likewise.
31036         (GOT_memory_operand): Likewise.
31037         (call_insn_operand): Likewise.
31038         (sibcall_insn_operand): Likewise.
31039         (GOT32_symbol_operand): Likewise.
31041 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
31043         PR middle-end/83837
31044         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
31045         type rather than type addr's type points to.
31046         (expand_omp_atomic_mutex): Likewise.
31047         (expand_omp_atomic): Likewise.
31049 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
31051         PR target/83839
31052         * config/i386/i386.c (output_indirect_thunk_function): Use
31053         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
31054         for  __x86_return_thunk.
31056 2018-01-15  Richard Biener  <rguenther@suse.de>
31058         PR middle-end/83850
31059         * expmed.c (extract_bit_field_1): Fix typo.
31061 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31063         PR target/83687
31064         * config/arm/iterators.md (VF): New mode iterator.
31065         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
31066         Remove integer-related logic from pattern.
31067         (neon_vabd<mode>_3): Likewise.
31069 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
31071         PR middle-end/82694
31072         * common.opt (fstrict-overflow): No longer an alias.
31073         (fwrapv-pointer): New option.
31074         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
31075         also for pointer types based on flag_wrapv_pointer.
31076         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
31077         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
31078         opts->x_flag_wrapv got set.
31079         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
31080         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
31081         POINTER_TYPE_OVERFLOW_UNDEFINED.
31082         * match.pd: Likewise in address comparison pattern.
31083         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
31085 2018-01-15  Richard Biener  <rguenther@suse.de>
31087         PR lto/83804
31088         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
31089         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
31090         Reset type names to their identifier if their TYPE_DECL doesn't
31091         have linkage (and thus is used for ODR and devirt).
31092         (save_debug_info_for_decl): Remove.
31093         (save_debug_info_for_type): Likewise.
31094         (add_tree_to_fld_list): Adjust.
31095         * tree-pretty-print.c (dump_generic_node): Make dumping of
31096         type names more robust.
31098 2018-01-15  Richard Biener  <rguenther@suse.de>
31100         * BASE-VER: Bump to 8.0.1.
31102 2018-01-14  Martin Sebor  <msebor@redhat.com>
31104         PR other/83508
31105         * builtins.c (check_access): Avoid warning when the no-warning bit
31106         is set.
31108 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
31110         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
31111         * ira-color (allocno_hard_regs_compare): Likewise.
31113 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
31115         PR target/83013
31116         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
31117         Use .pushsection/.popsection.
31119 2018-01-14  Martin Sebor  <msebor@redhat.com>
31121         PR c++/81327
31122         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
31124 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
31126         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
31127         entry from extra_headers.
31128         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
31129         extra_headers, make the list bitwise identical to the i?86-*-* one.
31131 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31133         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
31134         -mcmodel=large with -mindirect-branch=thunk,
31135         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
31136         -mfunction-return=thunk-extern.
31137         * doc/invoke.texi: Document -mcmodel=large is incompatible with
31138         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
31139         -mfunction-return=thunk and -mfunction-return=thunk-extern.
31141 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31143         * config/i386/i386.c (print_reg): Print the name of the full
31144         integer register without '%'.
31145         (ix86_print_operand): Handle 'V'.
31146         * doc/extend.texi: Document 'V' modifier.
31148 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31150         * config/i386/constraints.md (Bs): Disallow memory operand for
31151         -mindirect-branch-register.
31152         (Bw): Likewise.
31153         * config/i386/predicates.md (indirect_branch_operand): Likewise.
31154         (GOT_memory_operand): Likewise.
31155         (call_insn_operand): Likewise.
31156         (sibcall_insn_operand): Likewise.
31157         (GOT32_symbol_operand): Likewise.
31158         * config/i386/i386.md (indirect_jump): Call convert_memory_address
31159         for -mindirect-branch-register.
31160         (tablejump): Likewise.
31161         (*sibcall_memory): Likewise.
31162         (*sibcall_value_memory): Likewise.
31163         Disallow peepholes of indirect call and jump via memory for
31164         -mindirect-branch-register.
31165         (*call_pop): Replace m with Bw.
31166         (*call_value_pop): Likewise.
31167         (*sibcall_pop_memory): Replace m with Bs.
31168         * config/i386/i386.opt (mindirect-branch-register): New option.
31169         * doc/invoke.texi: Document -mindirect-branch-register option.
31171 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31173         * config/i386/i386-protos.h (ix86_output_function_return): New.
31174         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
31175         set function_return_type.
31176         (indirect_thunk_name): Add ret_p to indicate thunk for function
31177         return.
31178         (output_indirect_thunk_function): Pass false to
31179         indirect_thunk_name.
31180         (ix86_output_indirect_branch_via_reg): Likewise.
31181         (ix86_output_indirect_branch_via_push): Likewise.
31182         (output_indirect_thunk_function): Create alias for function
31183         return thunk if regno < 0.
31184         (ix86_output_function_return): New function.
31185         (ix86_handle_fndecl_attribute): Handle function_return.
31186         (ix86_attribute_table): Add function_return.
31187         * config/i386/i386.h (machine_function): Add
31188         function_return_type.
31189         * config/i386/i386.md (simple_return_internal): Use
31190         ix86_output_function_return.
31191         (simple_return_internal_long): Likewise.
31192         * config/i386/i386.opt (mfunction-return=): New option.
31193         (indirect_branch): Mention -mfunction-return=.
31194         * doc/extend.texi: Document function_return function attribute.
31195         * doc/invoke.texi: Document -mfunction-return= option.
31197 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
31199         * config/i386/i386-opts.h (indirect_branch): New.
31200         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
31201         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
31202         with local indirect jump when converting indirect call and jump.
31203         (ix86_set_indirect_branch_type): New.
31204         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
31205         (indirectlabelno): New.
31206         (indirect_thunk_needed): Likewise.
31207         (indirect_thunk_bnd_needed): Likewise.
31208         (indirect_thunks_used): Likewise.
31209         (indirect_thunks_bnd_used): Likewise.
31210         (INDIRECT_LABEL): Likewise.
31211         (indirect_thunk_name): Likewise.
31212         (output_indirect_thunk): Likewise.
31213         (output_indirect_thunk_function): Likewise.
31214         (ix86_output_indirect_branch_via_reg): Likewise.
31215         (ix86_output_indirect_branch_via_push): Likewise.
31216         (ix86_output_indirect_branch): Likewise.
31217         (ix86_output_indirect_jmp): Likewise.
31218         (ix86_code_end): Call output_indirect_thunk_function if needed.
31219         (ix86_output_call_insn): Call ix86_output_indirect_branch if
31220         needed.
31221         (ix86_handle_fndecl_attribute): Handle indirect_branch.
31222         (ix86_attribute_table): Add indirect_branch.
31223         * config/i386/i386.h (machine_function): Add indirect_branch_type
31224         and has_local_indirect_jump.
31225         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
31226         to true.
31227         (tablejump): Likewise.
31228         (*indirect_jump): Use ix86_output_indirect_jmp.
31229         (*tablejump_1): Likewise.
31230         (simple_return_indirect_internal): Likewise.
31231         * config/i386/i386.opt (mindirect-branch=): New option.
31232         (indirect_branch): New.
31233         (keep): Likewise.
31234         (thunk): Likewise.
31235         (thunk-inline): Likewise.
31236         (thunk-extern): Likewise.
31237         * doc/extend.texi: Document indirect_branch function attribute.
31238         * doc/invoke.texi: Document -mindirect-branch= option.
31240 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
31242         PR ipa/83051
31243         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
31245 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
31247         * ipa-inline.c (want_inline_small_function_p): Return false if
31248         inlining has already failed with CIF_FINAL_ERROR.
31249         (update_caller_keys): Call want_inline_small_function_p before
31250         can_inline_edge_p.
31251         (update_callee_keys): Likewise.
31253 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
31255         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
31256         New function.
31257         (rs6000_quadword_masked_address_p): Likewise.
31258         (quad_aligned_load_p): Likewise.
31259         (quad_aligned_store_p): Likewise.
31260         (const_load_sequence_p): Add comment to describe the outer-most loop.
31261         (mimic_memory_attributes_and_flags): New function.
31262         (rs6000_gen_stvx): Likewise.
31263         (replace_swapped_aligned_store): Likewise.
31264         (rs6000_gen_lvx): Likewise.
31265         (replace_swapped_aligned_load): Likewise.
31266         (replace_swapped_load_constant): Capitalize argument name in
31267         comment describing this function.
31268         (rs6000_analyze_swaps): Add a third pass to search for vector loads
31269         and stores that access quad-word aligned addresses and replace
31270         with stvx or lvx instructions when appropriate.
31271         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
31272         New function prototype.
31273         (rs6000_quadword_masked_address_p): Likewise.
31274         (rs6000_gen_lvx): Likewise.
31275         (rs6000_gen_stvx): Likewise.
31276         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
31277         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
31278         when memory address is aligned.
31279         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
31280         this split to select lvx instruction when memory address is aligned.
31281         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
31282         instruction when memory address is aligned.
31283         (*vsx_le_perm_load_v16qi): Likewise.
31284         (four unnamed splitters): Modify to select the stvx instruction
31285         when memory is aligned.
31287 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
31289         * predict.c (determine_unlikely_bbs): Handle correctly BBs
31290         which appears in the queue multiple times.
31292 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31293             Alan Hayward  <alan.hayward@arm.com>
31294             David Sherwood  <david.sherwood@arm.com>
31296         * tree-vectorizer.h (vec_lower_bound): New structure.
31297         (_loop_vec_info): Add check_nonzero and lower_bounds.
31298         (LOOP_VINFO_CHECK_NONZERO): New macro.
31299         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
31300         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
31301         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
31302         fields.  Make seg_len the distance travelled, not including the
31303         access size.
31304         (dr_direction_indicator): Declare.
31305         (dr_zero_step_indicator): Likewise.
31306         (dr_known_forward_stride_p): Likewise.
31307         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
31308         tree-ssanames.h.
31309         (runtime_alias_check_p): Allow runtime alias checks with
31310         variable strides.
31311         (operator ==): Compare access_size and align.
31312         (prune_runtime_alias_test_list): Rework for new distinction between
31313         the access_size and seg_len.
31314         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
31315         segment lengths.
31316         (get_segment_min_max): New function.
31317         (create_intersect_range_checks): Use it.
31318         (dr_step_indicator): New function.
31319         (dr_direction_indicator): Likewise.
31320         (dr_zero_step_indicator): Likewise.
31321         (dr_known_forward_stride_p): Likewise.
31322         * tree-loop-distribution.c (data_ref_segment_size): Return
31323         DR_STEP * (niters - 1).
31324         (compute_alias_check_pairs): Update call to the dr_with_seg_len
31325         constructor.
31326         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
31327         (vect_preserves_scalar_order_p): New function, split out from...
31328         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
31329         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
31330         (vect_vfa_access_size): New function.
31331         (vect_vfa_align): Likewise.
31332         (vect_compile_time_alias): Take access_size_a and access_b arguments.
31333         (dump_lower_bound): New function.
31334         (vect_check_lower_bound): Likewise.
31335         (vect_small_gap_p): Likewise.
31336         (vectorizable_with_step_bound_p): Likewise.
31337         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
31338         depencies if the vectorization factor is 1.  Convert the checks
31339         for nonzero steps into checks on the bounds of DR_STEP.  Try using
31340         a bunds check for variable steps if the minimum required step is
31341         relatively small. Update calls to the dr_with_seg_len
31342         constructor and to vect_compile_time_alias.
31343         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
31344         function.
31345         (vect_loop_versioning): Call it.
31346         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
31347         when retrying.
31348         (vect_estimate_min_profitable_iters): Account for any bounds checks.
31350 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31351             Alan Hayward  <alan.hayward@arm.com>
31352             David Sherwood  <david.sherwood@arm.com>
31354         * doc/sourcebuild.texi (vect_scatter_store): Document.
31355         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
31356         optabs.
31357         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
31358         Document.
31359         * genopinit.c (main): Add supports_vec_scatter_store and
31360         supports_vec_scatter_store_cached to target_optabs.
31361         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
31362         IFN_MASK_SCATTER_STORE.
31363         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
31364         functions.
31365         * internal-fn.h (internal_store_fn_p): Declare.
31366         (internal_fn_stored_value_index): Likewise.
31367         * internal-fn.c (scatter_store_direct): New macro.
31368         (expand_scatter_store_optab_fn): New function.
31369         (direct_scatter_store_optab_supported_p): New macro.
31370         (internal_store_fn_p): New function.
31371         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
31372         IFN_MASK_SCATTER_STORE.
31373         (internal_fn_mask_index): Likewise.
31374         (internal_fn_stored_value_index): New function.
31375         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
31376         for scatter stores.
31377         * optabs-query.h (supports_vec_scatter_store_p): Declare.
31378         * optabs-query.c (supports_vec_scatter_store_p): New function.
31379         * tree-vectorizer.h (vect_get_store_rhs): Declare.
31380         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
31381         true for scatter stores.
31382         (vect_gather_scatter_fn_p): Handle scatter stores too.
31383         (vect_check_gather_scatter): Consider using scatter stores if
31384         supports_vec_scatter_store_p.
31385         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
31386         scatter stores too.
31387         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
31388         internal_fn_stored_value_index.
31389         (check_load_store_masking): Handle scatter stores too.
31390         (vect_get_store_rhs): Make public.
31391         (vectorizable_call): Use internal_store_fn_p.
31392         (vectorizable_store): Handle scatter store internal functions.
31393         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
31394         when deciding whether the end of the group has been reached.
31395         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
31396         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
31397         (mask_scatter_store<mode>): New insns.
31399 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31400             Alan Hayward  <alan.hayward@arm.com>
31401             David Sherwood  <david.sherwood@arm.com>
31403         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
31404         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
31405         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
31406         function.
31407         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
31408         Use vect_truncate_gather_scatter_offset if we can't treat the
31409         operation as a normal gather load or scatter store.
31410         (get_group_load_store_type): Take the gather_scatter_info
31411         as argument.  Try using a gather load or scatter store for
31412         single-element groups.
31413         (get_load_store_type): Update calls to get_group_load_store_type
31414         and vect_use_strided_gather_scatters_p.
31416 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31417             Alan Hayward  <alan.hayward@arm.com>
31418             David Sherwood  <david.sherwood@arm.com>
31420         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
31421         optional tree argument.
31422         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
31423         null target hooks.
31424         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
31425         but continue to use the current value as a fallback.
31426         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
31427         to compare the updates.
31428         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
31429         (get_load_store_type): Use it when handling a strided access.
31430         (vect_get_strided_load_store_ops): New function.
31431         (vect_get_data_ptr_increment): Likewise.
31432         (vectorizable_load): Handle strided gather loads.  Always pass
31433         a step to vect_create_data_ref_ptr and bump_vector_ptr.
31435 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31436             Alan Hayward  <alan.hayward@arm.com>
31437             David Sherwood  <david.sherwood@arm.com>
31439         * doc/md.texi (gather_load@var{m}): Document.
31440         (mask_gather_load@var{m}): Likewise.
31441         * genopinit.c (main): Add supports_vec_gather_load and
31442         supports_vec_gather_load_cached to target_optabs.
31443         * optabs-tree.c (init_tree_optimization_optabs): Use
31444         ggc_cleared_alloc to allocate target_optabs.
31445         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
31446         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
31447         functions.
31448         * internal-fn.h (internal_load_fn_p): Declare.
31449         (internal_gather_scatter_fn_p): Likewise.
31450         (internal_fn_mask_index): Likewise.
31451         (internal_gather_scatter_fn_supported_p): Likewise.
31452         * internal-fn.c (gather_load_direct): New macro.
31453         (expand_gather_load_optab_fn): New function.
31454         (direct_gather_load_optab_supported_p): New macro.
31455         (direct_internal_fn_optab): New function.
31456         (internal_load_fn_p): Likewise.
31457         (internal_gather_scatter_fn_p): Likewise.
31458         (internal_fn_mask_index): Likewise.
31459         (internal_gather_scatter_fn_supported_p): Likewise.
31460         * optabs-query.c (supports_at_least_one_mode_p): New function.
31461         (supports_vec_gather_load_p): Likewise.
31462         * optabs-query.h (supports_vec_gather_load_p): Declare.
31463         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
31464         and memory_type field.
31465         (NUM_PATTERNS): Bump to 15.
31466         * tree-vect-data-refs.c: Include internal-fn.h.
31467         (vect_gather_scatter_fn_p): New function.
31468         (vect_describe_gather_scatter_call): Likewise.
31469         (vect_check_gather_scatter): Try using internal functions for
31470         gather loads.  Recognize existing calls to a gather load function.
31471         (vect_analyze_data_refs): Consider using gather loads if
31472         supports_vec_gather_load_p.
31473         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
31474         (vect_get_gather_scatter_offset_type): Likewise.
31475         (vect_convert_mask_for_vectype): Likewise.
31476         (vect_add_conversion_to_patterm): Likewise.
31477         (vect_try_gather_scatter_pattern): Likewise.
31478         (vect_recog_gather_scatter_pattern): New pattern recognizer.
31479         (vect_vect_recog_func_ptrs): Add it.
31480         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
31481         internal_fn_mask_index and internal_gather_scatter_fn_p.
31482         (check_load_store_masking): Take the gather_scatter_info as an
31483         argument and handle gather loads.
31484         (vect_get_gather_scatter_ops): New function.
31485         (vectorizable_call): Check internal_load_fn_p.
31486         (vectorizable_load): Likewise.  Handle gather load internal
31487         functions.
31488         (vectorizable_store): Update call to check_load_store_masking.
31489         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
31490         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
31491         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
31492         (aarch64_gather_scale_operand_d): New predicates.
31493         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
31494         (mask_gather_load<mode>): New insns.
31496 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31497             Alan Hayward  <alan.hayward@arm.com>
31498             David Sherwood  <david.sherwood@arm.com>
31500         * optabs.def (fold_left_plus_optab): New optab.
31501         * doc/md.texi (fold_left_plus_@var{m}): Document.
31502         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
31503         * internal-fn.c (fold_left_direct): Define.
31504         (expand_fold_left_optab_fn): Likewise.
31505         (direct_fold_left_optab_supported_p): Likewise.
31506         * fold-const-call.c (fold_const_fold_left): New function.
31507         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
31508         * tree-parloops.c (valid_reduction_p): New function.
31509         (gather_scalar_reductions): Use it.
31510         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
31511         (vect_finish_replace_stmt): Declare.
31512         * tree-vect-loop.c (fold_left_reduction_fn): New function.
31513         (needs_fold_left_reduction_p): New function, split out from...
31514         (vect_is_simple_reduction): ...here.  Accept reductions that
31515         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
31516         (vect_force_simple_reduction): Also store the reduction type in
31517         the assignment's STMT_VINFO_REDUC_TYPE.
31518         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
31519         (merge_with_identity): New function.
31520         (vect_expand_fold_left): Likewise.
31521         (vectorize_fold_left_reduction): Likewise.
31522         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
31523         scalar phi in place for it.  Check for target support and reject
31524         cases that would reassociate the operation.  Defer the transform
31525         phase to vectorize_fold_left_reduction.
31526         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
31527         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
31528         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
31530 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31532         * tree-if-conv.c (predicate_mem_writes): Remove redundant
31533         call to ifc_temp_var.
31535 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31536             Alan Hayward  <alan.hayward@arm.com>
31537             David Sherwood  <david.sherwood@arm.com>
31539         * target.def (legitimize_address_displacement): Take the original
31540         offset as a poly_int.
31541         * targhooks.h (default_legitimize_address_displacement): Update
31542         accordingly.
31543         * targhooks.c (default_legitimize_address_displacement): Likewise.
31544         * doc/tm.texi: Regenerate.
31545         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
31546         as an argument, moving assert of ad->disp == ad->disp_term to...
31547         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
31548         Try calling targetm.legitimize_address_displacement before expanding
31549         the address rather than afterwards, and adjust for the new interface.
31550         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
31551         Match the new hook interface.  Handle SVE addresses.
31552         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
31553         new hook interface.
31555 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31557         * Makefile.in (OBJS): Add early-remat.o.
31558         * target.def (select_early_remat_modes): New hook.
31559         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
31560         * doc/tm.texi: Regenerate.
31561         * targhooks.h (default_select_early_remat_modes): Declare.
31562         * targhooks.c (default_select_early_remat_modes): New function.
31563         * timevar.def (TV_EARLY_REMAT): New timevar.
31564         * passes.def (pass_early_remat): New pass.
31565         * tree-pass.h (make_pass_early_remat): Declare.
31566         * early-remat.c: New file.
31567         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
31568         function.
31569         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
31571 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31572             Alan Hayward  <alan.hayward@arm.com>
31573             David Sherwood  <david.sherwood@arm.com>
31575         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
31576         vfm1 with a bound_epilog parameter.
31577         (vect_do_peeling): Update calls accordingly, and move the prologue
31578         call earlier in the function.  Treat the base bound_epilog as 0 for
31579         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
31580         this base when peeling for gaps.
31581         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
31582         with fully-masked loops.
31583         (vect_estimate_min_profitable_iters): Handle the single peeled
31584         iteration in that case.
31586 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31587             Alan Hayward  <alan.hayward@arm.com>
31588             David Sherwood  <david.sherwood@arm.com>
31590         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
31591         single-element interleaving even if the size is not a power of 2.
31592         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
31593         accesses for single-element interleaving if the group size is
31594         not a power of 2.
31596 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31597             Alan Hayward  <alan.hayward@arm.com>
31598             David Sherwood  <david.sherwood@arm.com>
31600         * doc/md.texi (fold_extract_last_@var{m}): Document.
31601         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
31602         * optabs.def (fold_extract_last_optab): New optab.
31603         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
31604         * internal-fn.c (fold_extract_direct): New macro.
31605         (expand_fold_extract_optab_fn): Likewise.
31606         (direct_fold_extract_optab_supported_p): Likewise.
31607         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
31608         * tree-vect-loop.c (vect_model_reduction_cost): Handle
31609         EXTRACT_LAST_REDUCTION.
31610         (get_initial_def_for_reduction): Do not create an initial vector
31611         for EXTRACT_LAST_REDUCTION reductions.
31612         (vectorizable_reduction): Leave the scalar phi in place for
31613         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
31614         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
31615         epilogue code for EXTRACT_LAST_REDUCTION and defer the
31616         transform phase to vectorizable_condition.
31617         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
31618         split out from...
31619         (vect_finish_stmt_generation): ...here.
31620         (vect_finish_replace_stmt): New function.
31621         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
31622         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
31623         pattern.
31624         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
31626 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31627             Alan Hayward  <alan.hayward@arm.com>
31628             David Sherwood  <david.sherwood@arm.com>
31630         * doc/md.texi (extract_last_@var{m}): Document.
31631         * optabs.def (extract_last_optab): New optab.
31632         * internal-fn.def (EXTRACT_LAST): New internal function.
31633         * internal-fn.c (cond_unary_direct): New macro.
31634         (expand_cond_unary_optab_fn): Likewise.
31635         (direct_cond_unary_optab_supported_p): Likewise.
31636         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
31637         loops using EXTRACT_LAST.
31638         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
31639         (extract_last_<mode>): ...this optab.
31640         (vec_extract<mode><Vel>): Update accordingly.
31642 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31643             Alan Hayward  <alan.hayward@arm.com>
31644             David Sherwood  <david.sherwood@arm.com>
31646         * target.def (empty_mask_is_expensive): New hook.
31647         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
31648         * doc/tm.texi: Regenerate.
31649         * targhooks.h (default_empty_mask_is_expensive): Declare.
31650         * targhooks.c (default_empty_mask_is_expensive): New function.
31651         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
31652         if the target says that empty masks are expensive.
31653         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
31654         New function.
31655         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
31657 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31658             Alan Hayward  <alan.hayward@arm.com>
31659             David Sherwood  <david.sherwood@arm.com>
31661         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
31662         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
31663         (vect_use_loop_mask_for_alignment_p): New function.
31664         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
31665         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
31666         niters_skip argument.  Make sure that the first niters_skip elements
31667         of the first iteration are inactive.
31668         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
31669         Update call to vect_set_loop_masks_directly.
31670         (get_misalign_in_elems): New function, split out from...
31671         (vect_gen_prolog_loop_niters): ...here.
31672         (vect_update_init_of_dr): Take a code argument that specifies whether
31673         the adjustment should be added or subtracted.
31674         (vect_update_init_of_drs): Likewise.
31675         (vect_prepare_for_masked_peels): New function.
31676         (vect_do_peeling): Skip prologue peeling if we're using a mask
31677         instead.  Update call to vect_update_inits_of_drs.
31678         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31679         mask_skip_niters.
31680         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
31681         alignment.  Do not include the number of peeled iterations in
31682         the minimum threshold in that case.
31683         (vectorizable_induction): Adjust the start value down by
31684         LOOP_VINFO_MASK_SKIP_NITERS iterations.
31685         (vect_transform_loop): Call vect_prepare_for_masked_peels.
31686         Take the number of skipped iterations into account when calculating
31687         the loop bounds.
31688         * tree-vect-stmts.c (vect_gen_while_not): New function.
31690 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31691             Alan Hayward  <alan.hayward@arm.com>
31692             David Sherwood  <david.sherwood@arm.com>
31694         * doc/sourcebuild.texi (vect_fully_masked): Document.
31695         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
31696         default value to 0.
31697         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
31698         split out from...
31699         (vect_analyze_loop_2): ...here. Don't check the vectorization
31700         factor against the number of loop iterations if the loop is
31701         fully-masked.
31703 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31704             Alan Hayward  <alan.hayward@arm.com>
31705             David Sherwood  <david.sherwood@arm.com>
31707         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
31708         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
31709         (dump_groups): Update accordingly.
31710         (iv_use::mem_type): New member variable.
31711         (address_p): New function.
31712         (record_use): Add a mem_type argument and initialize the new
31713         mem_type field.
31714         (record_group_use): Add a mem_type argument.  Use address_p.
31715         Remove obsolete null checks of base_object.  Update call to record_use.
31716         (find_interesting_uses_op): Update call to record_group_use.
31717         (find_interesting_uses_cond): Likewise.
31718         (find_interesting_uses_address): Likewise.
31719         (get_mem_type_for_internal_fn): New function.
31720         (find_address_like_use): Likewise.
31721         (find_interesting_uses_stmt): Try find_address_like_use before
31722         calling find_interesting_uses_op.
31723         (addr_offset_valid_p): Use the iv mem_type field as the type
31724         of the addressed memory.
31725         (add_autoinc_candidates): Likewise.
31726         (get_address_cost): Likewise.
31727         (split_small_address_groups_p): Use address_p.
31728         (split_address_groups): Likewise.
31729         (add_iv_candidate_for_use): Likewise.
31730         (autoinc_possible_for_pair): Likewise.
31731         (rewrite_groups): Likewise.
31732         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
31733         (determine_group_iv_cost): Update after split of USE_ADDRESS.
31734         (get_alias_ptr_type_for_ptr_address): New function.
31735         (rewrite_use_address): Rewrite address uses in calls that were
31736         identified by find_address_like_use.
31738 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31739             Alan Hayward  <alan.hayward@arm.com>
31740             David Sherwood  <david.sherwood@arm.com>
31742         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
31743         TARGET_MEM_REFs.
31744         * gimple-expr.h (is_gimple_addressable: Likewise.
31745         * gimple-expr.c (is_gimple_address): Likewise.
31746         * internal-fn.c (expand_call_mem_ref): New function.
31747         (expand_mask_load_optab_fn): Use it.
31748         (expand_mask_store_optab_fn): Likewise.
31750 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31751             Alan Hayward  <alan.hayward@arm.com>
31752             David Sherwood  <david.sherwood@arm.com>
31754         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
31755         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
31756         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
31757         (cond_umax@var{mode}): Document.
31758         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
31759         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
31760         (cond_umin_optab, cond_umax_optab): New optabs.
31761         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
31762         (COND_IOR, COND_XOR): New internal functions.
31763         * internal-fn.h (get_conditional_internal_fn): Declare.
31764         * internal-fn.c (cond_binary_direct): New macro.
31765         (expand_cond_binary_optab_fn): Likewise.
31766         (direct_cond_binary_optab_supported_p): Likewise.
31767         (get_conditional_internal_fn): New function.
31768         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
31769         Cope with reduction statements that are vectorized as calls rather
31770         than assignments.
31771         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
31772         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
31773         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
31774         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
31775         (UNSPEC_COND_EOR): New unspecs.
31776         (optab): Add mappings for them.
31777         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
31778         (sve_int_op, sve_fp_op): New int attributes.
31780 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31781             Alan Hayward  <alan.hayward@arm.com>
31782             David Sherwood  <david.sherwood@arm.com>
31784         * optabs.def (while_ult_optab): New optab.
31785         * doc/md.texi (while_ult@var{m}@var{n}): Document.
31786         * internal-fn.def (WHILE_ULT): New internal function.
31787         * internal-fn.h (direct_internal_fn_supported_p): New override
31788         that takes two types as argument.
31789         * internal-fn.c (while_direct): New macro.
31790         (expand_while_optab_fn): New function.
31791         (convert_optab_supported_p): Likewise.
31792         (direct_while_optab_supported_p): New macro.
31793         * wide-int.h (wi::udiv_ceil): New function.
31794         * tree-vectorizer.h (rgroup_masks): New structure.
31795         (vec_loop_masks): New typedef.
31796         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
31797         and fully_masked_p.
31798         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
31799         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
31800         (vect_max_vf): New function.
31801         (slpeel_make_loop_iterate_ntimes): Delete.
31802         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
31803         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
31804         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
31805         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
31806         internal-fn.h, stor-layout.h and optabs-query.h.
31807         (vect_set_loop_mask): New function.
31808         (add_preheader_seq): Likewise.
31809         (add_header_seq): Likewise.
31810         (interleave_supported_p): Likewise.
31811         (vect_maybe_permute_loop_masks): Likewise.
31812         (vect_set_loop_masks_directly): Likewise.
31813         (vect_set_loop_condition_masked): Likewise.
31814         (vect_set_loop_condition_unmasked): New function, split out from
31815         slpeel_make_loop_iterate_ntimes.
31816         (slpeel_make_loop_iterate_ntimes): Rename to..
31817         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
31818         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
31819         (vect_do_peeling): Update call accordingly.
31820         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
31821         loops.
31822         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31823         mask_compare_type, can_fully_mask_p and fully_masked_p.
31824         (release_vec_loop_masks): New function.
31825         (_loop_vec_info): Use it to free the loop masks.
31826         (can_produce_all_loop_masks_p): New function.
31827         (vect_get_max_nscalars_per_iter): Likewise.
31828         (vect_verify_full_masking): Likewise.
31829         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
31830         retries, and free the mask rgroups before retrying.  Check loop-wide
31831         reasons for disallowing fully-masked loops.  Make the final decision
31832         about whether use a fully-masked loop or not.
31833         (vect_estimate_min_profitable_iters): Do not assume that peeling
31834         for the number of iterations will be needed for fully-masked loops.
31835         (vectorizable_reduction): Disable fully-masked loops.
31836         (vectorizable_live_operation): Likewise.
31837         (vect_halve_mask_nunits): New function.
31838         (vect_double_mask_nunits): Likewise.
31839         (vect_record_loop_mask): Likewise.
31840         (vect_get_loop_mask): Likewise.
31841         (vect_transform_loop): Handle the case in which the final loop
31842         iteration might handle a partial vector.  Call vect_set_loop_condition
31843         instead of slpeel_make_loop_iterate_ntimes.
31844         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
31845         (check_load_store_masking): New function.
31846         (prepare_load_store_mask): Likewise.
31847         (vectorizable_store): Handle fully-masked loops.
31848         (vectorizable_load): Likewise.
31849         (supportable_widening_operation): Use vect_halve_mask_nunits for
31850         booleans.
31851         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
31852         (vect_gen_while): New function.
31853         * config/aarch64/aarch64.md (umax<mode>3): New expander.
31854         (aarch64_uqdec<mode>): New insn.
31856 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31857             Alan Hayward  <alan.hayward@arm.com>
31858             David Sherwood  <david.sherwood@arm.com>
31860         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
31861         (reduc_xor_scal_optab): New optabs.
31862         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
31863         (reduc_xor_scal_@var{m}): Document.
31864         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
31865         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
31866         internal functions.
31867         * fold-const-call.c (fold_const_call): Handle them.
31868         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
31869         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
31870         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
31871         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
31872         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
31873         (UNSPEC_XORV): New unspecs.
31874         (optab): Add entries for them.
31875         (BITWISEV): New int iterator.
31876         (bit_reduc_op): New int attributes.
31878 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31879             Alan Hayward  <alan.hayward@arm.com>
31880             David Sherwood  <david.sherwood@arm.com>
31882         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
31883         * internal-fn.def (VEC_SHL_INSERT): New internal function.
31884         * optabs.def (vec_shl_insert_optab): New optab.
31885         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
31886         (duplicate_and_interleave): Likewise.
31887         * tree-vect-loop.c: Include internal-fn.h.
31888         (neutral_op_for_slp_reduction): New function, split out from
31889         get_initial_defs_for_reduction.
31890         (get_initial_def_for_reduction): Handle option 2 for variable-length
31891         vectors by loading the neutral value into a vector and then shifting
31892         the initial value into element 0.
31893         (get_initial_defs_for_reduction): Replace the code argument with
31894         the neutral value calculated by neutral_op_for_slp_reduction.
31895         Use gimple_build_vector for constant-length vectors.
31896         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
31897         but the first group_size elements have a neutral value.
31898         Use duplicate_and_interleave otherwise.
31899         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
31900         Update call to get_initial_defs_for_reduction.  Handle SLP
31901         reductions for variable-length vectors by creating one vector
31902         result for each scalar result, with the elements associated
31903         with other scalar results stubbed out with the neutral value.
31904         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
31905         Require IFN_VEC_SHL_INSERT for double reductions on
31906         variable-length vectors, or SLP reductions that have
31907         a neutral value.  Require can_duplicate_and_interleave_p
31908         support for variable-length unchained SLP reductions if there
31909         is no neutral value, such as for MIN/MAX reductions.  Also require
31910         the number of vector elements to be a multiple of the number of
31911         SLP statements when doing variable-length unchained SLP reductions.
31912         Update call to vect_create_epilog_for_reduction.
31913         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
31914         and remove initial values.
31915         (duplicate_and_interleave): Make public.
31916         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
31917         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
31919 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31920             Alan Hayward  <alan.hayward@arm.com>
31921             David Sherwood  <david.sherwood@arm.com>
31923         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
31924         (can_duplicate_and_interleave_p): New function.
31925         (vect_get_and_check_slp_defs): Take the vector of statements
31926         rather than just the current one.  Remove excess parentheses.
31927         Restriction rejectinon of vect_constant_def and vect_external_def
31928         for variable-length vectors to boolean types, or types for which
31929         can_duplicate_and_interleave_p is false.
31930         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
31931         (duplicate_and_interleave): New function.
31932         (vect_get_constant_vectors): Use gimple_build_vector for
31933         constant-length vectors and suitable variable-length constant
31934         vectors.  Use duplicate_and_interleave for other variable-length
31935         vectors.  Don't defer the update when inserting new statements.
31937 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31938             Alan Hayward  <alan.hayward@arm.com>
31939             David Sherwood  <david.sherwood@arm.com>
31941         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
31942         min_profitable_iters doesn't go negative.
31944 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31945             Alan Hayward  <alan.hayward@arm.com>
31946             David Sherwood  <david.sherwood@arm.com>
31948         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
31949         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
31950         * optabs.def (vec_mask_load_lanes_optab): New optab.
31951         (vec_mask_store_lanes_optab): Likewise.
31952         * internal-fn.def (MASK_LOAD_LANES): New internal function.
31953         (MASK_STORE_LANES): Likewise.
31954         * internal-fn.c (mask_load_lanes_direct): New macro.
31955         (mask_store_lanes_direct): Likewise.
31956         (expand_mask_load_optab_fn): Handle masked operations.
31957         (expand_mask_load_lanes_optab_fn): New macro.
31958         (expand_mask_store_optab_fn): Handle masked operations.
31959         (expand_mask_store_lanes_optab_fn): New macro.
31960         (direct_mask_load_lanes_optab_supported_p): Likewise.
31961         (direct_mask_store_lanes_optab_supported_p): Likewise.
31962         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
31963         parameter.
31964         (vect_load_lanes_supported): Likewise.
31965         * tree-vect-data-refs.c (strip_conversion): New function.
31966         (can_group_stmts_p): Likewise.
31967         (vect_analyze_data_ref_accesses): Use it instead of checking
31968         for a pair of assignments.
31969         (vect_store_lanes_supported): Take a masked_p parameter.
31970         (vect_load_lanes_supported): Likewise.
31971         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
31972         vect_store_lanes_supported and vect_load_lanes_supported.
31973         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
31974         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
31975         parameter.  Don't allow gaps for masked accesses.
31976         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
31977         and vect_load_lanes_supported.
31978         (get_load_store_type): Take a masked_p parameter and update
31979         call to get_group_load_store_type.
31980         (vectorizable_store): Update call to get_load_store_type.
31981         Handle IFN_MASK_STORE_LANES.
31982         (vectorizable_load): Update call to get_load_store_type.
31983         Handle IFN_MASK_LOAD_LANES.
31985 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31986             Alan Hayward  <alan.hayward@arm.com>
31987             David Sherwood  <david.sherwood@arm.com>
31989         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
31990         modes for SVE.
31991         * config/aarch64/aarch64-protos.h
31992         (aarch64_sve_struct_memory_operand_p): Declare.
31993         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
31994         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
31995         (VPRED, vpred): Handle SVE structure modes.
31996         * config/aarch64/constraints.md (Utx): New constraint.
31997         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
31998         (aarch64_sve_struct_nonimmediate_operand): New predicates.
31999         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
32000         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
32001         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
32002         structure modes.  Split into pieces after RA.
32003         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
32004         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
32005         New patterns.
32006         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
32007         SVE structure modes.
32008         (aarch64_classify_address): Likewise.
32009         (sizetochar): Move earlier in file.
32010         (aarch64_print_operand): Handle SVE register lists.
32011         (aarch64_array_mode): New function.
32012         (aarch64_sve_struct_memory_operand_p): Likewise.
32013         (TARGET_ARRAY_MODE): Redefine.
32015 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32016             Alan Hayward  <alan.hayward@arm.com>
32017             David Sherwood  <david.sherwood@arm.com>
32019         * target.def (array_mode): New target hook.
32020         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
32021         * doc/tm.texi: Regenerate.
32022         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
32023         * hooks.c (hook_optmode_mode_uhwi_none): New function.
32024         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
32025         targetm.array_mode.
32026         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
32027         type sizes.
32029 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32030             Alan Hayward  <alan.hayward@arm.com>
32031             David Sherwood  <david.sherwood@arm.com>
32033         * fold-const.c (fold_binary_loc): Check the argument types
32034         rather than the result type when testing for a vector operation.
32036 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32038         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
32039         * doc/tm.texi: Regenerate.
32041 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32042             Alan Hayward  <alan.hayward@arm.com>
32043             David Sherwood  <david.sherwood@arm.com>
32045         * doc/invoke.texi (-msve-vector-bits=): Document new option.
32046         (sve): Document new AArch64 extension.
32047         * doc/md.texi (w): Extend the description of the AArch64
32048         constraint to include SVE vectors.
32049         (Upl, Upa): Document new AArch64 predicate constraints.
32050         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
32051         enum.
32052         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
32053         (msve-vector-bits=): New option.
32054         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
32055         SVE when these are disabled.
32056         (sve): New extension.
32057         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
32058         modes.  Adjust their number of units based on aarch64_sve_vg.
32059         (MAX_BITSIZE_MODE_ANY_MODE): Define.
32060         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
32061         aarch64_addr_query_type.
32062         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
32063         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
32064         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
32065         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
32066         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
32067         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
32068         (aarch64_simd_imm_zero_p): Delete.
32069         (aarch64_check_zero_based_sve_index_immediate): Declare.
32070         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32071         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32072         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32073         (aarch64_sve_float_mul_immediate_p): Likewise.
32074         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32075         rather than an rtx.
32076         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
32077         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
32078         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
32079         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
32080         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
32081         (aarch64_regmode_natural_size): Likewise.
32082         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
32083         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
32084         left one place.
32085         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
32086         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
32087         for VG and the SVE predicate registers.
32088         (V_ALIASES): Add a "z"-prefixed alias.
32089         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
32090         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
32091         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
32092         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
32093         (REG_CLASS_NAMES): Add entries for them.
32094         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
32095         and the predicate registers.
32096         (aarch64_sve_vg): Declare.
32097         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
32098         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
32099         (REGMODE_NATURAL_SIZE): Define.
32100         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
32101         SVE macros.
32102         * config/aarch64/aarch64.c: Include cfgrtl.h.
32103         (simd_immediate_info): Add a constructor for series vectors,
32104         and an associated step field.
32105         (aarch64_sve_vg): New variable.
32106         (aarch64_dbx_register_number): Handle VG and the predicate registers.
32107         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
32108         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
32109         (VEC_ANY_DATA, VEC_STRUCT): New constants.
32110         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
32111         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
32112         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
32113         (aarch64_get_mask_mode): New functions.
32114         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
32115         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
32116         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
32117         predicate modes and predicate registers.  Explicitly restrict
32118         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
32119         to store a vector mode if it is recognized by
32120         aarch64_classify_vector_mode.
32121         (aarch64_regmode_natural_size): New function.
32122         (aarch64_hard_regno_caller_save_mode): Return the original mode
32123         for predicates.
32124         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
32125         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
32126         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
32127         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
32128         functions.
32129         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
32130         does not overlap dest if the function is frame-related.  Handle
32131         SVE constants.
32132         (aarch64_split_add_offset): New function.
32133         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
32134         them aarch64_add_offset.
32135         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
32136         and update call to aarch64_sub_sp.
32137         (aarch64_add_cfa_expression): New function.
32138         (aarch64_expand_prologue): Pass extra temporary registers to the
32139         functions above.  Handle the case in which we need to emit new
32140         DW_CFA_expressions for registers that were originally saved
32141         relative to the stack pointer, but now have to be expressed
32142         relative to the frame pointer.
32143         (aarch64_output_mi_thunk): Pass extra temporary registers to the
32144         functions above.
32145         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
32146         IP0 and IP1 values for SVE frames.
32147         (aarch64_expand_vec_series): New function.
32148         (aarch64_expand_sve_widened_duplicate): Likewise.
32149         (aarch64_expand_sve_const_vector): Likewise.
32150         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
32151         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
32152         into the register, rather than emitting a SET directly.
32153         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
32154         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
32155         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
32156         (offset_9bit_signed_scaled_p): New functions.
32157         (aarch64_replicate_bitmask_imm): New function.
32158         (aarch64_bitmask_imm): Use it.
32159         (aarch64_cannot_force_const_mem): Reject expressions involving
32160         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
32161         (aarch64_classify_index): Handle SVE indices, by requiring
32162         a plain register index with a scale that matches the element size.
32163         (aarch64_classify_address): Handle SVE addresses.  Assert that
32164         the mode of the address is VOIDmode or an integer mode.
32165         Update call to aarch64_classify_symbol.
32166         (aarch64_classify_symbolic_expression): Update call to
32167         aarch64_classify_symbol.
32168         (aarch64_const_vec_all_in_range_p): New function.
32169         (aarch64_print_vector_float_operand): Likewise.
32170         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
32171         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
32172         and the FP immediates 1.0 and 0.5.
32173         (aarch64_print_address_internal): Handle SVE addresses.
32174         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
32175         (aarch64_regno_regclass): Handle predicate registers.
32176         (aarch64_secondary_reload): Handle big-endian reloads of SVE
32177         data modes.
32178         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
32179         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
32180         (aarch64_convert_sve_vector_bits): New function.
32181         (aarch64_override_options): Use it to handle -msve-vector-bits=.
32182         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32183         rather than an rtx.
32184         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
32185         Handle SVE vector and predicate modes.  Accept VL-based constants
32186         that need only one temporary register, and VL offsets that require
32187         no temporary registers.
32188         (aarch64_conditional_register_usage): Mark the predicate registers
32189         as fixed if SVE isn't available.
32190         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
32191         Return true for SVE vector and predicate modes.
32192         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
32193         rather than an unsigned int.  Handle SVE modes.
32194         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
32195         SVE modes.
32196         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
32197         if SVE is enabled.
32198         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32199         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32200         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32201         (aarch64_sve_float_mul_immediate_p): New functions.
32202         (aarch64_sve_valid_immediate): New function.
32203         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
32204         Explicitly reject structure modes.  Check for INDEX constants.
32205         Handle PTRUE and PFALSE constants.
32206         (aarch64_check_zero_based_sve_index_immediate): New function.
32207         (aarch64_simd_imm_zero_p): Delete.
32208         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
32209         vector modes.  Accept constants in the range of CNT[BHWD].
32210         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
32211         ask for an Advanced SIMD mode.
32212         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
32213         (aarch64_simd_vector_alignment): Handle SVE predicates.
32214         (aarch64_vectorize_preferred_vector_alignment): New function.
32215         (aarch64_simd_vector_alignment_reachable): Use it instead of
32216         the vector size.
32217         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
32218         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
32219         functions.
32220         (MAX_VECT_LEN): Delete.
32221         (expand_vec_perm_d): Add a vec_flags field.
32222         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
32223         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
32224         (aarch64_evpc_ext): Don't apply a big-endian lane correction
32225         for SVE modes.
32226         (aarch64_evpc_rev): Rename to...
32227         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
32228         (aarch64_evpc_rev_global): New function.
32229         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
32230         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
32231         MAX_VECT_LEN.
32232         (aarch64_evpc_sve_tbl): New function.
32233         (aarch64_expand_vec_perm_const_1): Update after rename of
32234         aarch64_evpc_rev.  Handle SVE permutes too, trying
32235         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
32236         than aarch64_evpc_tbl.
32237         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
32238         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
32239         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
32240         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
32241         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
32242         (aarch64_expand_sve_vcond): New functions.
32243         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
32244         of aarch64_vector_mode_p.
32245         (aarch64_dwarf_poly_indeterminate_value): New function.
32246         (aarch64_compute_pressure_classes): Likewise.
32247         (aarch64_can_change_mode_class): Likewise.
32248         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
32249         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
32250         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
32251         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
32252         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
32253         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
32254         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
32255         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
32256         constraints.
32257         (Dn, Dl, Dr): Accept const as well as const_vector.
32258         (Dz): Likewise.  Compare against CONST0_RTX.
32259         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
32260         of "vector" where appropriate.
32261         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
32262         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
32263         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
32264         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
32265         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
32266         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
32267         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
32268         (v_int_equiv): Extend to SVE modes.
32269         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
32270         mode attributes.
32271         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
32272         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
32273         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
32274         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
32275         (SVE_COND_FP_CMP): New int iterators.
32276         (perm_hilo): Handle the new unpack unspecs.
32277         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
32278         attributes.
32279         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
32280         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
32281         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
32282         (aarch64_equality_operator, aarch64_constant_vector_operand)
32283         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
32284         (aarch64_sve_nonimmediate_operand): Likewise.
32285         (aarch64_sve_general_operand): Likewise.
32286         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
32287         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
32288         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
32289         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
32290         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
32291         (aarch64_sve_float_arith_immediate): Likewise.
32292         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
32293         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
32294         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
32295         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
32296         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
32297         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
32298         (aarch64_sve_float_arith_operand): Likewise.
32299         (aarch64_sve_float_arith_with_sub_operand): Likewise.
32300         (aarch64_sve_float_mul_operand): Likewise.
32301         (aarch64_sve_vec_perm_operand): Likewise.
32302         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
32303         (aarch64_mov_operand): Accept const_poly_int and const_vector.
32304         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
32305         as well as const_vector.
32306         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
32307         in file.  Use CONST0_RTX and CONSTM1_RTX.
32308         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
32309         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
32310         Use aarch64_simd_imm_zero.
32311         * config/aarch64/aarch64-sve.md: New file.
32312         * config/aarch64/aarch64.md: Include it.
32313         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
32314         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
32315         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
32316         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
32317         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
32318         (sve): New attribute.
32319         (enabled): Disable instructions with the sve attribute unless
32320         TARGET_SVE.
32321         (movqi, movhi): Pass CONST_POLY_INT operaneds through
32322         aarch64_expand_mov_immediate.
32323         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
32324         CNT[BHSD] immediates.
32325         (movti): Split CONST_POLY_INT moves into two halves.
32326         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
32327         Split additions that need a temporary here if the destination
32328         is the stack pointer.
32329         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
32330         (*add<mode>3_poly_1): New instruction.
32331         (set_clobber_cc): New expander.
32333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32335         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
32336         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
32337         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
32338         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
32339         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
32340         Change innermode from fixed_mode_size to machine_mode.
32341         (simplify_subreg): Update call accordingly.  Handle a constant-sized
32342         subreg of a variable-length CONST_VECTOR.
32344 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
32345             Alan Hayward  <alan.hayward@arm.com>
32346             David Sherwood  <david.sherwood@arm.com>
32348         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
32349         (add_offset_to_base): New function, split out from...
32350         (create_mem_ref): ...here.  When handling a scale other than 1,
32351         check first whether the address is valid without the offset.
32352         Add it into the base if so, leaving the index and scale as-is.
32354 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
32356         PR c++/83778
32357         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
32358         fold_for_warn before checking if arg2 is INTEGER_CST.
32360 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
32362         * config/rs6000/predicates.md (load_multiple_operation): Delete.
32363         (store_multiple_operation): Delete.
32364         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
32365         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
32366         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
32367         guarded by TARGET_STRING.
32368         (rs6000_output_load_multiple): Delete.
32369         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
32370         OPTION_MASK_STRING / TARGET_STRING handling.
32371         (print_operand) <'N', 'O'>: Add comment that these are unused now.
32372         (const rs6000_opt_masks) <"string">: Change mask to 0.
32373         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
32374         (MASK_STRING): Delete.
32375         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
32376         parts.  Simplify.
32377         (load_multiple): Delete.
32378         (*ldmsi8): Delete.
32379         (*ldmsi7): Delete.
32380         (*ldmsi6): Delete.
32381         (*ldmsi5): Delete.
32382         (*ldmsi4): Delete.
32383         (*ldmsi3): Delete.
32384         (store_multiple): Delete.
32385         (*stmsi8): Delete.
32386         (*stmsi7): Delete.
32387         (*stmsi6): Delete.
32388         (*stmsi5): Delete.
32389         (*stmsi4): Delete.
32390         (*stmsi3): Delete.
32391         (movmemsi_8reg): Delete.
32392         (corresponding unnamed define_insn): Delete.
32393         (movmemsi_6reg): Delete.
32394         (corresponding unnamed define_insn): Delete.
32395         (movmemsi_4reg): Delete.
32396         (corresponding unnamed define_insn): Delete.
32397         (movmemsi_2reg): Delete.
32398         (corresponding unnamed define_insn): Delete.
32399         (movmemsi_1reg): Delete.
32400         (corresponding unnamed define_insn): Delete.
32401         * config/rs6000/rs6000.opt (mno-string): New.
32402         (mstring): Replace by deprecation warning stub.
32403         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
32405 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
32407         * regrename.c (regrename_do_replace): If replacing the same
32408         reg multiple times, try to reuse last created gen_raw_REG.
32410         PR debug/81155
32411         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
32412         main to workaround a bug in GDB.
32414 2018-01-12  Tom de Vries  <tom@codesourcery.com>
32416         PR target/83737
32417         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
32419 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
32421         PR rtl-optimization/80481
32422         * ira-color.c (get_cap_member): New function.
32423         (allocnos_conflict_by_live_ranges_p): Use it.
32424         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
32425         (setup_slot_coalesced_allocno_live_ranges): Ditto.
32427 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
32429         PR target/83628
32430         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
32431         (*saddl_se_1): Ditto.
32432         (*ssubsi_1): Ditto.
32433         (*ssubl_se_1): Ditto.
32435 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
32437         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
32438         rather than wi::to_widest for DR_INITs.
32439         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
32440         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
32441         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
32442         INTEGER_CSTs.
32443         (vect_analyze_group_access_1): Note that here.
32445 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
32447         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
32448         polynomial type sizes.
32450 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
32452         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
32453         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
32454         (gimple_add_tmp_var): Likewise.
32456 2018-01-12  Martin Liska  <mliska@suse.cz>
32458         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
32459         (gimple_alloc_sizes): Likewise.
32460         (dump_gimple_statistics): Use PRIu64 in printf format.
32461         * gimple.h: Change uint64_t to int.
32463 2018-01-12  Martin Liska  <mliska@suse.cz>
32465         * tree-core.h: Use uint64_t instead of int.
32466         * tree.c (tree_node_counts): Likewise.
32467         (tree_node_sizes): Likewise.
32468         (dump_tree_statistics): Use PRIu64 in printf format.
32470 2018-01-12  Martin Liska  <mliska@suse.cz>
32472         * Makefile.in: As qsort_chk is implemented in vec.c, add
32473         vec.o to linkage of gencfn-macros.
32474         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
32475         passing the info to record_node_allocation_statistics.
32476         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
32477         and pass the info.
32478         * ggc-common.c (struct ggc_usage): Add operator== and use
32479         it in operator< and compare function.
32480         * mem-stats.h (struct mem_usage): Likewise.
32481         * vec.c (struct vec_usage): Remove operator< and compare
32482         function. Can be simply inherited.
32484 2018-01-12  Martin Jambor  <mjambor@suse.cz>
32486         PR target/81616
32487         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
32488         * tree-ssa-math-opts.c: Include domwalk.h.
32489         (convert_mult_to_fma_1): New function.
32490         (fma_transformation_info): New type.
32491         (fma_deferring_state): Likewise.
32492         (cancel_fma_deferring): New function.
32493         (result_of_phi): Likewise.
32494         (last_fma_candidate_feeds_initial_phi): Likewise.
32495         (convert_mult_to_fma): Added deferring logic, split actual
32496         transformation to convert_mult_to_fma_1.
32497         (math_opts_dom_walker): New type.
32498         (math_opts_dom_walker::after_dom_children): New method, body moved
32499         here from pass_optimize_widening_mul::execute, added deferring logic
32500         bits.
32501         (pass_optimize_widening_mul::execute): Moved most of code to
32502         math_opts_dom_walker::after_dom_children.
32503         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
32504         * config/i386/i386.c (ix86_option_override_internal): Added
32505         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
32507 2018-01-12  Richard Biener  <rguenther@suse.de>
32509         PR debug/83157
32510         * dwarf2out.c (gen_variable_die): Do not reset old_die for
32511         inline instance vars.
32513 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
32515         PR target/81819
32516         * config/rx/rx.c (rx_is_restricted_memory_address):
32517         Handle SUBREG case.
32519 2018-01-12  Richard Biener  <rguenther@suse.de>
32521         PR tree-optimization/80846
32522         * target.def (split_reduction): New target hook.
32523         * targhooks.c (default_split_reduction): New function.
32524         * targhooks.h (default_split_reduction): Declare.
32525         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
32526         target requests first reduce vectors by combining low and high
32527         parts.
32528         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
32529         (get_vectype_for_scalar_type_and_size): Export.
32530         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
32531         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
32532         * doc/tm.texi: Regenerate.
32533         * config/i386/i386.c (ix86_split_reduction): Implement
32534         TARGET_VECTORIZE_SPLIT_REDUCTION.
32536 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
32538         PR target/83368
32539         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
32540         in PIC mode except for TARGET_VXWORKS_RTP.
32541         * config/sparc/sparc.c: Include cfgrtl.h.
32542         (TARGET_INIT_PIC_REG): Define.
32543         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
32544         (sparc_pic_register_p): New predicate.
32545         (sparc_legitimate_address_p): Use it.
32546         (sparc_legitimize_pic_address): Likewise.
32547         (sparc_delegitimize_address): Likewise.
32548         (sparc_mode_dependent_address_p): Likewise.
32549         (gen_load_pcrel_sym): Remove 4th parameter.
32550         (load_got_register): Adjust call to above.  Remove obsolete stuff.
32551         (sparc_expand_prologue): Do not call load_got_register here.
32552         (sparc_flat_expand_prologue): Likewise.
32553         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
32554         (sparc_use_pseudo_pic_reg): New function.
32555         (sparc_init_pic_reg): Likewise.
32556         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
32557         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
32559 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
32561         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
32562         Add item for branch_cost.
32564 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
32566         PR rtl-optimization/83565
32567         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
32568         not extend the result to a larger mode for rotate operations.
32569         (num_sign_bit_copies1): Likewise.
32571 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
32573         PR target/40411
32574         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
32575         -symbolic.
32576         Use values-Xc.o for -pedantic.
32577         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
32579 2018-01-12  Martin Liska  <mliska@suse.cz>
32581         PR ipa/83054
32582         * ipa-devirt.c (final_warning_record::grow_type_warnings):
32583         New function.
32584         (possible_polymorphic_call_targets): Use it.
32585         (ipa_devirt): Likewise.
32587 2018-01-12  Martin Liska  <mliska@suse.cz>
32589         * profile-count.h (enum profile_quality): Use 0 as invalid
32590         enum value of profile_quality.
32592 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
32594         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
32595         -mext-string options.
32597 2018-01-12  Richard Biener  <rguenther@suse.de>
32599         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
32600         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
32601         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
32602         Likewise.
32603         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
32605 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32607         * configure.ac (--with-long-double-format): Add support for the
32608         configuration option to change the default long double format on
32609         PowerPC systems.
32610         * config.gcc (powerpc*-linux*-*): Likewise.
32611         * configure: Regenerate.
32612         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
32613         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
32614         used without modification.
32616 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32618         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
32619         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
32620         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
32621         MISC_BUILTIN_SPEC_BARRIER.
32622         (rs6000_init_builtins): Likewise.
32623         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
32624         enum value.
32625         (speculation_barrier): New define_insn.
32626         * doc/extend.texi: Document __builtin_speculation_barrier.
32628 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
32630         PR target/83203
32631         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
32632         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
32633         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
32634         iterators.
32635         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
32636         integral modes instead of "ss" and "sd".
32637         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
32638         vectors with 32-bit and 64-bit elements.
32639         (vecdupssescalarmodesuffix): New mode attribute.
32640         (vec_dup<mode>): Use it.
32642 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
32644         PR target/83330
32645         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
32646         frame if argument is passed on stack.
32648 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
32650         PR target/82682
32651         * ree.c (combine_reaching_defs): Optimize also
32652         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
32653         reg2=any_extend(exp); reg1=reg2;, formatting fix.
32655 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
32657         PR middle-end/83189
32658         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
32660 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
32662         PR middle-end/83718
32663         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
32664         after they are computed.
32666 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
32668         PR tree-optimization/83695
32669         * gimple-loop-linterchange.cc
32670         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
32671         reset cached scev information after interchange.
32672         (pass_linterchange::execute): Remove call to scev_reset_htab.
32674 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32676         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
32677         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
32678         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
32679         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
32680         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
32681         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
32682         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
32683         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
32684         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
32685         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
32686         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
32687         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
32688         (V_lane_reg): Likewise.
32689         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
32690         New define_expand.
32691         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
32692         (vfmal_lane_low<mode>_intrinsic,
32693         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
32694         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
32695         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
32696         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
32697         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
32698         vfmsl_lane_high<mode>_intrinsic): New define_insns.
32700 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32702         * config/arm/arm-cpus.in (fp16fml): New feature.
32703         (ALL_SIMD): Add fp16fml.
32704         (armv8.2-a): Add fp16fml as an option.
32705         (armv8.3-a): Likewise.
32706         (armv8.4-a): Add fp16fml as part of fp16.
32707         * config/arm/arm.h (TARGET_FP16FML): Define.
32708         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
32709         when appropriate.
32710         * config/arm/arm-modes.def (V2HF): Define.
32711         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
32712         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
32713         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
32714         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
32715         vfmsl_low, vfmsl_high): New set of builtins.
32716         * config/arm/iterators.md (PLUSMINUS): New code iterator.
32717         (vfml_op): New code attribute.
32718         (VFMLHALVES): New int iterator.
32719         (VFML, VFMLSEL): New mode attributes.
32720         (V_reg): Define mapping for V2HF.
32721         (V_hi, V_lo): New mode attributes.
32722         (VF_constraint): Likewise.
32723         (vfml_half, vfml_half_selector): New int attributes.
32724         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
32725         define_expand.
32726         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
32727         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
32728         New define_insn.
32729         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
32730         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
32731         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
32732         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
32733         documentation.
32734         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
32735         Document new effective target and option set.
32737 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32739         * config/arm/arm-cpus.in (armv8_4): New feature.
32740         (ARMv8_4a): New fgroup.
32741         (armv8.4-a): New arch.
32742         * config/arm/arm-tables.opt: Regenerate.
32743         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
32744         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
32745         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
32746         Add matching rules for -march=armv8.4-a and extensions.
32747         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
32749 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
32751         PR target/81821
32752         * config/rx/rx.md (BW): New mode attribute.
32753         (sync_lock_test_and_setsi): Add mode suffix to insn output.
32755 2018-01-11  Richard Biener  <rguenther@suse.de>
32757         PR tree-optimization/83435
32758         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
32759         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
32760         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
32762 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32763             Alan Hayward  <alan.hayward@arm.com>
32764             David Sherwood  <david.sherwood@arm.com>
32766         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
32767         field.
32768         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
32769         (aarch64_print_address_internal): Use it to check for a zero offset.
32771 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32772             Alan Hayward  <alan.hayward@arm.com>
32773             David Sherwood  <david.sherwood@arm.com>
32775         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
32776         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
32777         Return a poly_int64 rather than a HOST_WIDE_INT.
32778         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
32779         rather than a HOST_WIDE_INT.
32780         * config/aarch64/aarch64.h (aarch64_frame): Protect with
32781         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
32782         hard_fp_offset, frame_size, initial_adjust, callee_offset and
32783         final_offset from HOST_WIDE_INT to poly_int64.
32784         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
32785         to_constant when getting the number of units in an Advanced SIMD
32786         mode.
32787         (aarch64_builtin_vectorized_function): Check for a constant number
32788         of units.
32789         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
32790         GET_MODE_SIZE.
32791         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
32792         attribute instead of GET_MODE_NUNITS.
32793         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
32794         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
32795         GET_MODE_SIZE for fixed-size registers.
32796         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
32797         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
32798         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
32799         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
32800         (aarch64_print_operand, aarch64_print_address_internal)
32801         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
32802         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
32803         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
32804         Handle polynomial GET_MODE_SIZE.
32805         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
32806         wider than SImode without modification.
32807         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
32808         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
32809         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
32810         passing and returning SVE modes.
32811         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
32812         rather than GEN_INT.
32813         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
32814         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
32815         (aarch64_allocate_and_probe_stack_space): Likewise.
32816         (aarch64_layout_frame): Cope with polynomial offsets.
32817         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
32818         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
32819         polynomial offsets.
32820         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
32821         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
32822         poly_int64 rather than a HOST_WIDE_INT.
32823         (aarch64_get_separate_components, aarch64_process_components)
32824         (aarch64_expand_prologue, aarch64_expand_epilogue)
32825         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
32826         (aarch64_anchor_offset): New function, split out from...
32827         (aarch64_legitimize_address): ...here.
32828         (aarch64_builtin_vectorization_cost): Handle polynomial
32829         TYPE_VECTOR_SUBPARTS.
32830         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
32831         GET_MODE_NUNITS.
32832         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
32833         number of elements from the PARALLEL rather than the mode.
32834         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
32835         rather than GET_MODE_BITSIZE.
32836         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
32837         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
32838         (aarch64_expand_vec_perm_const_1): Handle polynomial
32839         d->perm.length () and d->perm elements.
32840         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
32841         Apply to_constant to d->perm elements.
32842         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
32843         polynomial CONST_VECTOR_NUNITS.
32844         (aarch64_move_pointer): Take amount as a poly_int64 rather
32845         than an int.
32846         (aarch64_progress_pointer): Avoid temporary variable.
32847         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
32848         the mode attribute instead of GET_MODE.
32850 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32851             Alan Hayward  <alan.hayward@arm.com>
32852             David Sherwood  <david.sherwood@arm.com>
32854         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
32855         x exists before using it.
32856         (aarch64_add_constant_internal): Rename to...
32857         (aarch64_add_offset_1): ...this.  Replace regnum with separate
32858         src and dest rtxes.  Handle the case in which they're different,
32859         including when the offset is zero.  Replace scratchreg with an rtx.
32860         Use 2 additions if there is no spare register into which we can
32861         move a 16-bit constant.
32862         (aarch64_add_constant): Delete.
32863         (aarch64_add_offset): Replace reg with separate src and dest
32864         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
32865         Use aarch64_add_offset_1.
32866         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
32867         an rtx rather than an int.  Take the delta as a poly_int64
32868         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
32869         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
32870         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
32871         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
32872         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
32873         and aarch64_add_sp.
32874         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
32875         aarch64_add_constant.
32877 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32879         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
32880         Use scalar_float_mode.
32882 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32884         * config/aarch64/aarch64-simd.md
32885         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
32886         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
32887         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
32888         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
32889         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
32890         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
32891         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
32892         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
32893         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
32894         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
32896 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32898         PR target/83514
32899         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
32900         targ_options->x_arm_arch_string is non NULL.
32902 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
32904         * config/aarch64/aarch64.h
32905         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
32907 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
32909         PR target/82096
32910         * expmed.c (emit_store_flag_force): Swap if const op0
32911         and change VOIDmode to mode of op0.
32913 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32915         PR rtl-optimization/83761
32916         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
32917         than bytes to mode_for_size.
32919 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32921         PR middle-end/83189
32922         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
32923         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
32924         profile.
32926 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32928         PR middle-end/83575
32929         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
32930         when in layout mode.
32931         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
32932         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
32933         partition fixup.
32935 2018-01-10  Michael Collison  <michael.collison@arm.com>
32937         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
32938         * config/aarch64/aarch64-option-extension.def: Add
32939         AARCH64_OPT_EXTENSION of 'fp16fml'.
32940         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32941         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
32942         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
32943         * config/aarch64/constraints.md (Ui7): New constraint.
32944         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
32945         (VFMLA_SEL_W): Ditto.
32946         (f16quad): Ditto.
32947         (f16mac1): Ditto.
32948         (VFMLA16_LOW): New int iterator.
32949         (VFMLA16_HIGH): Ditto.
32950         (UNSPEC_FMLAL): New unspec.
32951         (UNSPEC_FMLSL): Ditto.
32952         (UNSPEC_FMLAL2): Ditto.
32953         (UNSPEC_FMLSL2): Ditto.
32954         (f16mac): New code attribute.
32955         * config/aarch64/aarch64-simd-builtins.def
32956         (aarch64_fmlal_lowv2sf): Ditto.
32957         (aarch64_fmlsl_lowv2sf): Ditto.
32958         (aarch64_fmlalq_lowv4sf): Ditto.
32959         (aarch64_fmlslq_lowv4sf): Ditto.
32960         (aarch64_fmlal_highv2sf): Ditto.
32961         (aarch64_fmlsl_highv2sf): Ditto.
32962         (aarch64_fmlalq_highv4sf): Ditto.
32963         (aarch64_fmlslq_highv4sf): Ditto.
32964         (aarch64_fmlal_lane_lowv2sf): Ditto.
32965         (aarch64_fmlsl_lane_lowv2sf): Ditto.
32966         (aarch64_fmlal_laneq_lowv2sf): Ditto.
32967         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
32968         (aarch64_fmlalq_lane_lowv4sf): Ditto.
32969         (aarch64_fmlsl_lane_lowv4sf): Ditto.
32970         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
32971         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
32972         (aarch64_fmlal_lane_highv2sf): Ditto.
32973         (aarch64_fmlsl_lane_highv2sf): Ditto.
32974         (aarch64_fmlal_laneq_highv2sf): Ditto.
32975         (aarch64_fmlsl_laneq_highv2sf): Ditto.
32976         (aarch64_fmlalq_lane_highv4sf): Ditto.
32977         (aarch64_fmlsl_lane_highv4sf): Ditto.
32978         (aarch64_fmlalq_laneq_highv4sf): Ditto.
32979         (aarch64_fmlsl_laneq_highv4sf): Ditto.
32980         * config/aarch64/aarch64-simd.md:
32981         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
32982         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32983         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
32984         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32985         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
32986         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
32987         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
32988         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
32989         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
32990         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
32991         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
32992         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
32993         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
32994         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
32995         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
32996         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
32997         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
32998         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
32999         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
33000         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
33001         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
33002         (vfmlsl_low_u32): Ditto.
33003         (vfmlalq_low_u32): Ditto.
33004         (vfmlslq_low_u32): Ditto.
33005         (vfmlal_high_u32): Ditto.
33006         (vfmlsl_high_u32): Ditto.
33007         (vfmlalq_high_u32): Ditto.
33008         (vfmlslq_high_u32): Ditto.
33009         (vfmlal_lane_low_u32): Ditto.
33010         (vfmlsl_lane_low_u32): Ditto.
33011         (vfmlal_laneq_low_u32): Ditto.
33012         (vfmlsl_laneq_low_u32): Ditto.
33013         (vfmlalq_lane_low_u32): Ditto.
33014         (vfmlslq_lane_low_u32): Ditto.
33015         (vfmlalq_laneq_low_u32): Ditto.
33016         (vfmlslq_laneq_low_u32): Ditto.
33017         (vfmlal_lane_high_u32): Ditto.
33018         (vfmlsl_lane_high_u32): Ditto.
33019         (vfmlal_laneq_high_u32): Ditto.
33020         (vfmlsl_laneq_high_u32): Ditto.
33021         (vfmlalq_lane_high_u32): Ditto.
33022         (vfmlslq_lane_high_u32): Ditto.
33023         (vfmlalq_laneq_high_u32): Ditto.
33024         (vfmlslq_laneq_high_u32): Ditto.
33025         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
33026         (AARCH64_FL_FOR_ARCH8_4): New.
33027         (AARCH64_ISA_F16FML): New ISA flag.
33028         (TARGET_F16FML): New feature flag for fp16fml.
33029         (doc/invoke.texi): Document new fp16fml option.
33031 2018-01-10  Michael Collison  <michael.collison@arm.com>
33033         * config/aarch64/aarch64-builtins.c:
33034         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
33035         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33036         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
33037         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
33038         (AARCH64_ISA_SHA3): New ISA flag.
33039         (TARGET_SHA3): New feature flag for sha3.
33040         * config/aarch64/iterators.md (sha512_op): New int attribute.
33041         (CRYPTO_SHA512): New int iterator.
33042         (UNSPEC_SHA512H): New unspec.
33043         (UNSPEC_SHA512H2): Ditto.
33044         (UNSPEC_SHA512SU0): Ditto.
33045         (UNSPEC_SHA512SU1): Ditto.
33046         * config/aarch64/aarch64-simd-builtins.def
33047         (aarch64_crypto_sha512hqv2di): New builtin.
33048         (aarch64_crypto_sha512h2qv2di): Ditto.
33049         (aarch64_crypto_sha512su0qv2di): Ditto.
33050         (aarch64_crypto_sha512su1qv2di): Ditto.
33051         (aarch64_eor3qv8hi): Ditto.
33052         (aarch64_rax1qv2di): Ditto.
33053         (aarch64_xarqv2di): Ditto.
33054         (aarch64_bcaxqv8hi): Ditto.
33055         * config/aarch64/aarch64-simd.md:
33056         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
33057         (aarch64_crypto_sha512su0qv2di): Ditto.
33058         (aarch64_crypto_sha512su1qv2di): Ditto.
33059         (aarch64_eor3qv8hi): Ditto.
33060         (aarch64_rax1qv2di): Ditto.
33061         (aarch64_xarqv2di): Ditto.
33062         (aarch64_bcaxqv8hi): Ditto.
33063         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
33064         (vsha512h2q_u64): Ditto.
33065         (vsha512su0q_u64): Ditto.
33066         (vsha512su1q_u64): Ditto.
33067         (veor3q_u16): Ditto.
33068         (vrax1q_u64): Ditto.
33069         (vxarq_u64): Ditto.
33070         (vbcaxq_u16): Ditto.
33071         * config/arm/types.md (crypto_sha512): New type attribute.
33072         (crypto_sha3): Ditto.
33073         (doc/invoke.texi): Document new sha3 option.
33075 2018-01-10  Michael Collison  <michael.collison@arm.com>
33077         * config/aarch64/aarch64-builtins.c:
33078         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
33079         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33080         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
33081         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
33082         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
33083         (AARCH64_ISA_SM4): New ISA flag.
33084         (TARGET_SM4): New feature flag for sm4.
33085         * config/aarch64/aarch64-simd-builtins.def
33086         (aarch64_sm3ss1qv4si): Ditto.
33087         (aarch64_sm3tt1aq4si): Ditto.
33088         (aarch64_sm3tt1bq4si): Ditto.
33089         (aarch64_sm3tt2aq4si): Ditto.
33090         (aarch64_sm3tt2bq4si): Ditto.
33091         (aarch64_sm3partw1qv4si): Ditto.
33092         (aarch64_sm3partw2qv4si): Ditto.
33093         (aarch64_sm4eqv4si): Ditto.
33094         (aarch64_sm4ekeyqv4si): Ditto.
33095         * config/aarch64/aarch64-simd.md:
33096         (aarch64_sm3ss1qv4si): Ditto.
33097         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
33098         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
33099         (aarch64_sm4eqv4si): Ditto.
33100         (aarch64_sm4ekeyqv4si): Ditto.
33101         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
33102         (sm3part_op): Ditto.
33103         (CRYPTO_SM3TT): Ditto.
33104         (CRYPTO_SM3PART): Ditto.
33105         (UNSPEC_SM3SS1): New unspec.
33106         (UNSPEC_SM3TT1A): Ditto.
33107         (UNSPEC_SM3TT1B): Ditto.
33108         (UNSPEC_SM3TT2A): Ditto.
33109         (UNSPEC_SM3TT2B): Ditto.
33110         (UNSPEC_SM3PARTW1): Ditto.
33111         (UNSPEC_SM3PARTW2): Ditto.
33112         (UNSPEC_SM4E): Ditto.
33113         (UNSPEC_SM4EKEY): Ditto.
33114         * config/aarch64/constraints.md (Ui2): New constraint.
33115         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
33116         * config/arm/types.md (crypto_sm3): New type attribute.
33117         (crypto_sm4): Ditto.
33118         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
33119         (vsm3tt1aq_u32): Ditto.
33120         (vsm3tt1bq_u32): Ditto.
33121         (vsm3tt2aq_u32): Ditto.
33122         (vsm3tt2bq_u32): Ditto.
33123         (vsm3partw1q_u32): Ditto.
33124         (vsm3partw2q_u32): Ditto.
33125         (vsm4eq_u32): Ditto.
33126         (vsm4ekeyq_u32): Ditto.
33127         (doc/invoke.texi): Document new sm4 option.
33129 2018-01-10  Michael Collison  <michael.collison@arm.com>
33131         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
33132         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
33133         (AARCH64_FL_FOR_ARCH8_4): New.
33134         (AARCH64_FL_V8_4): New flag.
33135         (doc/invoke.texi): Document new armv8.4-a option.
33137 2018-01-10  Michael Collison  <michael.collison@arm.com>
33139         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33140         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
33141         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
33142         * config/aarch64/aarch64-option-extension.def: Add
33143         AARCH64_OPT_EXTENSION of 'sha2'.
33144         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
33145         (crypto): Disable sha2 and aes if crypto disabled.
33146         (crypto): Enable aes and sha2 if enabled.
33147         (simd): Disable sha2 and aes if simd disabled.
33148         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
33149         New flags.
33150         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
33151         (TARGET_SHA2): New feature flag for sha2.
33152         (TARGET_AES): New feature flag for aes.
33153         * config/aarch64/aarch64-simd.md:
33154         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
33155         conditional on TARGET_AES.
33156         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
33157         (aarch64_crypto_sha1hsi): Make pattern conditional
33158         on TARGET_SHA2.
33159         (aarch64_crypto_sha1hv4si): Ditto.
33160         (aarch64_be_crypto_sha1hv4si): Ditto.
33161         (aarch64_crypto_sha1su1v4si): Ditto.
33162         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
33163         (aarch64_crypto_sha1su0v4si): Ditto.
33164         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
33165         (aarch64_crypto_sha256su0v4si): Ditto.
33166         (aarch64_crypto_sha256su1v4si): Ditto.
33167         (doc/invoke.texi): Document new aes and sha2 options.
33169 2018-01-10  Martin Sebor  <msebor@redhat.com>
33171         PR tree-optimization/83781
33172         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
33173         as string arrays.
33175 2018-01-11  Martin Sebor  <msebor@gmail.com>
33176             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33178         PR tree-optimization/83501
33179         PR tree-optimization/81703
33181         * tree-ssa-strlen.c (get_string_cst): Rename...
33182         (get_string_len): ...to this.  Handle global constants.
33183         (handle_char_store): Adjust.
33185 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
33186             Jim Wilson  <jimw@sifive.com>
33188         * config/riscv/riscv-protos.h (riscv_output_return): New.
33189         * config/riscv/riscv.c (struct machine_function): New naked_p field.
33190         (riscv_attribute_table, riscv_output_return),
33191         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
33192         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
33193         (riscv_compute_frame_info): Only compute frame->mask if not a naked
33194         function.
33195         (riscv_expand_prologue): Add early return for naked function.
33196         (riscv_expand_epilogue): Likewise.
33197         (riscv_function_ok_for_sibcall): Return false for naked function.
33198         (riscv_set_current_function): New.
33199         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
33200         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
33201         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
33202         * doc/extend.texi (RISC-V Function Attributes): New.
33204 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
33206         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
33207         check for 128-bit long double before checking TCmode.
33208         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
33209         128-bit long doubles before checking TFmode or TCmode.
33210         (FLOAT128_IBM_P): Likewise.
33212 2018-01-10  Martin Sebor  <msebor@redhat.com>
33214         PR tree-optimization/83671
33215         * builtins.c (c_strlen): Unconditionally return zero for the empty
33216         string.
33217         Use -Warray-bounds for warnings.
33218         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
33219         for non-constant array indices with COMPONENT_REF, arrays of
33220         arrays, and pointers to arrays.
33221         (gimple_fold_builtin_strlen): Determine and set length range for
33222         non-constant character arrays.
33224 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
33226         PR middle-end/81897
33227         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
33228         empty blocks.
33230 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
33232         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
33234 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
33236         PR target/83399
33237         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
33238         VECTOR_MEM_ALTIVEC_OR_VSX_P.
33239         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
33240         indexed_or_indirect_operand predicate.
33241         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
33242         (*vsx_le_perm_load_v8hi): Likewise.
33243         (*vsx_le_perm_load_v16qi): Likewise.
33244         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
33245         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
33246         (*vsx_le_perm_store_v8hi): Likewise.
33247         (*vsx_le_perm_store_v16qi): Likewise.
33248         (eight unnamed splitters): Likewise.
33250 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
33252         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
33253         * config/rs6000/emmintrin.h: Likewise.
33254         * config/rs6000/mmintrin.h: Likewise.
33255         * config/rs6000/xmmintrin.h: Likewise.
33257 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
33259         PR c++/43486
33260         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
33261         "public_flag".
33262         * tree.c (tree_nop_conversion): Return true for location wrapper
33263         nodes.
33264         (maybe_wrap_with_location): New function.
33265         (selftest::check_strip_nops): New function.
33266         (selftest::test_location_wrappers): New function.
33267         (selftest::tree_c_tests): Call it.
33268         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
33269         (maybe_wrap_with_location): New decl.
33270         (EXPR_LOCATION_WRAPPER_P): New macro.
33271         (location_wrapper_p): New inline function.
33272         (tree_strip_any_location_wrapper): New inline function.
33274 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
33276         PR target/83735
33277         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
33278         stack_realign_offset for the largest alignment of stack slot
33279         actually used.
33280         (ix86_find_max_used_stack_alignment): New function.
33281         (ix86_finalize_stack_frame_flags): Use it.  Set
33282         max_used_stack_alignment if we don't realign stack.
33283         * config/i386/i386.h (machine_function): Add
33284         max_used_stack_alignment.
33286 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
33288         * config/arm/arm.opt (-mbranch-cost): New option.
33289         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
33290         account.
33292 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
33294         PR target/83629
33295         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
33296         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
33298 2018-01-10  Richard Biener  <rguenther@suse.de>
33300         PR debug/83765
33301         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
33302         early out so it also covers the case where we have a non-NULL
33303         origin.
33305 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
33307         PR tree-optimization/83753
33308         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
33309         for non-strided grouped accesses if the number of elements is 1.
33311 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
33313         PR target/81616
33314         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
33315         * i386.h (TARGET_USE_GATHER): Define.
33316         * x86-tune.def (X86_TUNE_USE_GATHER): New.
33318 2018-01-10  Martin Liska  <mliska@suse.cz>
33320         PR bootstrap/82831
33321         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
33322         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
33323         partitioning.
33324         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
33325         CLEANUP_NO_PARTITIONING is not set.
33327 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
33329         * doc/rtl.texi: Remove documentation of (const ...) wrappers
33330         for vectors, as a partial revert of r254296.
33331         * rtl.h (const_vec_p): Delete.
33332         (const_vec_duplicate_p): Don't test for vector CONSTs.
33333         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
33334         * expmed.c (make_tree): Likewise.
33336         Revert:
33337         * common.md (E, F): Use CONSTANT_P instead of checking for
33338         CONST_VECTOR.
33339         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
33340         checking for CONST_VECTOR.
33342 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
33344         PR middle-end/83575
33345         * predict.c (force_edge_cold): Handle in more sane way edges
33346         with no prediction.
33348 2018-01-09  Carl Love  <cel@us.ibm.com>
33350         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
33351         V4SI, V4SF types.
33352         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
33353         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
33354         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
33355         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
33356         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
33357         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
33358         * config/rs6000/rs6000-protos.h: Add extern defition for
33359         rs6000_generate_float2_double_code.
33360         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
33361         function.
33362         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
33363         (float2_v2df): Add define_expand.
33365 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
33367         PR target/83628
33368         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
33369         op_mode in the force_to_mode call.
33371 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
33373         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
33374         instead of checking each element individually.
33375         (aarch64_evpc_uzp): Likewise.
33376         (aarch64_evpc_zip): Likewise.
33377         (aarch64_evpc_ext): Likewise.
33378         (aarch64_evpc_rev): Likewise.
33379         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
33380         instead of checking each element individually.  Return true without
33381         generating rtl if
33382         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
33383         whether all selected elements come from the same input, instead of
33384         checking each element individually.  Remove calls to gen_rtx_REG,
33385         start_sequence and end_sequence and instead assert that no rtl is
33386         generated.
33388 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
33390         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
33391         order of HIGH and CONST checks.
33393 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
33395         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
33396         if the destination isn't an SSA_NAME.
33398 2018-01-09  Richard Biener  <rguenther@suse.de>
33400         PR tree-optimization/83668
33401         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
33402         move prologue...
33403         (canonicalize_loop_form): ... here, renamed from ...
33404         (canonicalize_loop_closed_ssa_form): ... this and amended to
33405         swap successor edges for loop exit blocks to make us use
33406         the RPO order we need for initial schedule generation.
33408 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
33410         PR tree-optimization/64811
33411         * match.pd: When optimizing comparisons with Inf, avoid
33412         introducing or losing exceptions from comparisons with NaN.
33414 2018-01-09  Martin Liska  <mliska@suse.cz>
33416         PR sanitizer/82517
33417         * asan.c (shadow_mem_size): Add gcc_assert.
33419 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
33421         Don't save registers in main().
33423         PR target/83738
33424         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
33425         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
33426         * config/avr/avr.c (avr_set_current_function): Don't error if
33427         naked, OS_task or OS_main are specified at the same time.
33428         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
33429         OS_main.
33430         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
33431         attribute.
33432         * common/config/avr/avr-common.c (avr_option_optimization_table):
33433         Switch on -mmain-is-OS_task for optimizing compilations.
33435 2018-01-09  Richard Biener  <rguenther@suse.de>
33437         PR tree-optimization/83572
33438         * graphite.c: Include cfganal.h.
33439         (graphite_transform_loops): Connect infinite loops to exit
33440         and remove fake edges at the end.
33442 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
33444         * ipa-inline.c (edge_badness): Revert accidental checkin.
33446 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
33448         PR ipa/80763
33449         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
33450         symbols; not inline clones.
33452 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
33454         PR target/83507
33455         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
33456         hard registers.  Formatting fixes.
33458         PR preprocessor/83722
33459         * gcc.c (try_generate_repro): Pass
33460         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
33461         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
33462         do_report_bug.
33464 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
33465             Kito Cheng  <kito.cheng@gmail.com>
33467         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
33468         (riscv_leaf_function_p): Delete.
33469         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
33471 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
33473         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
33474         function.
33475         (do_ifelse): New function.
33476         (do_isel): New function.
33477         (do_sub3): New function.
33478         (do_add3): New function.
33479         (do_load_mask_compare): New function.
33480         (do_overlap_load_compare): New function.
33481         (expand_compare_loop): New function.
33482         (expand_block_compare): Call expand_compare_loop() when appropriate.
33483         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
33484         option description.
33485         (-mblock-compare-inline-loop-limit): New option.
33487 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
33489         PR target/83677
33490         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
33491         Reverse order of second and third operands in first alternative.
33492         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
33493         of first and second elements in UNSPEC_VPERMR vector.
33494         (altivec_expand_vec_perm_le): Likewise.
33496 2018-01-08  Jeff Law  <law@redhat.com>
33498         PR rtl-optimizatin/81308
33499         * tree-switch-conversion.c (cfg_altered): New file scoped static.
33500         (process_switch): If group_case_labels makes a change, then set
33501         cfg_altered.
33502         (pass_convert_switch::execute): If a switch is converted, then
33503         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
33505         PR rtl-optimization/81308
33506         * recog.c (split_all_insns): Conditionally cleanup the CFG after
33507         splitting insns.
33509 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
33511         PR target/83663 - Revert r255946
33512         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
33513         generation for cases where splatting a value is not useful.
33514         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
33515         across a vec_duplicate and a paradoxical subreg forming a vector
33516         mode to a vec_concat.
33518 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33520         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
33521         -march=armv8.3-a variants.
33522         * config/arm/t-multilib: Likewise.
33523         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
33525 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
33527         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
33528         to generate rtl.
33529         (cceq_ior_compare_complement): Give it a name so I can use it, and
33530         change boolean_or_operator predicate to boolean_operator so it can
33531         be used to generate a crand.
33532         (eqne): New code iterator.
33533         (bd/bd_neg): New code_attrs.
33534         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
33535         a single define_insn.
33536         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
33537         decrement (bdnzt/bdnzf/bdzt/bdzf).
33538         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
33539         with the new names of the branch decrement patterns, and added the
33540         names of the branch decrement conditional patterns.
33542 2018-01-08  Richard Biener  <rguenther@suse.de>
33544         PR tree-optimization/83563
33545         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
33546         cache.
33548 2018-01-08  Richard Biener  <rguenther@suse.de>
33550         PR middle-end/83713
33551         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
33553 2018-01-08  Richard Biener  <rguenther@suse.de>
33555         PR tree-optimization/83685
33556         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
33557         references to abnormals.
33559 2018-01-08  Richard Biener  <rguenther@suse.de>
33561         PR lto/83719
33562         * dwarf2out.c (output_indirect_strings): Handle empty
33563         skeleton_debug_str_hash.
33564         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
33566 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
33568         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
33569         (emit_store_direct): Likewise.
33570         (arc_trampoline_adjust_address): Likewise.
33571         (arc_asm_trampoline_template): New function.
33572         (arc_initialize_trampoline): Use asm_trampoline_template.
33573         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
33574         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
33575         * config/arc/arc.md (flush_icache): Delete pattern.
33577 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
33579         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
33580         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
33581         munaligned-access.
33583 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
33585         PR target/83681
33586         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
33587         by not USED_FOR_TARGET.
33588         (make_pass_resolve_sw_modes): Likewise.
33590 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
33592         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
33593         USED_FOR_TARGET.
33595 2018-01-08  Richard Biener  <rguenther@suse.de>
33597         PR middle-end/83580
33598         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
33600 2018-01-08  Richard Biener  <rguenther@suse.de>
33602         PR middle-end/83517
33603         * match.pd ((t * 2) / 2) -> t): Add missing :c.
33605 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
33607         PR middle-end/81897
33608         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
33609         basic blocks with a small number of successors.
33610         (convert_control_dep_chain_into_preds): Improve handling of
33611         forwarder blocks.
33612         (dump_predicates): Split apart into...
33613         (dump_pred_chain): ...here...
33614         (dump_pred_info): ...and here.
33615         (can_one_predicate_be_invalidated_p): Add debugging printfs.
33616         (can_chain_union_be_invalidated_p): Improve check for invalidation
33617         of paths.
33618         (uninit_uses_cannot_happen): Avoid unnecessary if
33619         convert_control_dep_chain_into_preds yielded nothing.
33621 2018-01-06  Martin Sebor  <msebor@redhat.com>
33623         PR tree-optimization/83640
33624         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
33625         subtracting negative offset from size.
33626         (builtin_access::overlap): Adjust offset bounds of the access to fall
33627         within the size of the object if possible.
33629 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
33631         PR rtl-optimization/83699
33632         * expmed.c (extract_bit_field_1): Restrict the vector usage of
33633         extract_bit_field_as_subreg to cases in which the extracted
33634         value is also a vector.
33636         * lra-constraints.c (process_alt_operands): Test for the equivalence
33637         substitutions when detecting a possible reload cycle.
33639 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
33641         PR debug/83480
33642         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
33643         by default if flag_selective_schedling{,2}.  Formatting fixes.
33645         PR rtl-optimization/83682
33646         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
33647         if it has non-VECTOR_MODE element mode.
33648         (vec_duplicate_p): Likewise.
33650         PR middle-end/83694
33651         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
33652         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
33654 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
33656         PR target/83604
33657         * config/i386/i386-builtin.def
33658         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
33659         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
33660         Require also OPTION_MASK_ISA_AVX512F in addition to
33661         OPTION_MASK_ISA_GFNI.
33662         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
33663         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
33664         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
33665         to OPTION_MASK_ISA_GFNI.
33666         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
33667         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
33668         OPTION_MASK_ISA_AVX512BW.
33669         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
33670         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
33671         addition to OPTION_MASK_ISA_GFNI.
33672         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
33673         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
33674         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
33675         to OPTION_MASK_ISA_GFNI.
33676         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
33677         a requirement for all ISAs rather than any of them with a few
33678         exceptions.
33679         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
33680         processing.
33681         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
33682         bitmasks to be enabled with 3 exceptions, instead of requiring any
33683         enabled ISA with lots of exceptions.
33684         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
33685         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
33686         Change avx512bw in isa attribute to avx512f.
33687         * config/i386/sgxintrin.h: Add license boilerplate.
33688         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
33689         to __AVX512F__ and __AVX512VL to __AVX512VL__.
33690         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
33691         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
33692         defined.
33693         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
33694         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
33695         temporarily sse2 rather than sse if not enabled already.
33697         PR target/83604
33698         * config/i386/sse.md (VI248_VLBW): Rename to ...
33699         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
33700         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
33701         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
33702         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
33703         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
33704         mode iterator instead of VI248_VLBW.
33706 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
33708         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
33709         (record_modified): Skip clobbers; add debug output.
33710         (param_change_prob): Use sreal frequencies.
33712 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33714         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
33715         punt for user-aligned variables.
33717 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33719         * tree-chrec.c (chrec_contains_symbols): Return true for
33720         POLY_INT_CST.
33722 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
33724         PR target/82439
33725         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
33726         of (x|y) == x for BICS pattern.
33728 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
33730         PR tree-optimization/83605
33731         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
33732         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
33733         can throw.
33735 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
33737         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
33738         * config/epiphany/rtems.h: New file.
33740 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33741             Uros Bizjak  <ubizjak@gmail.com>
33743         PR target/83554
33744         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
33745         QIreg_operand instead of register_operand predicate.
33746         * config/i386/i386.c (ix86_rop_should_change_byte_p,
33747         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
33748         comments instead of -fmitigate[-_]rop.
33750 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33752         PR bootstrap/81926
33753         * cgraphunit.c (symbol_table::compile): Switch to text_section
33754         before calling assembly_start debug hook.
33755         * run-rtl-passes.c (run_rtl_passes): Likewise.
33756         Include output.h.
33758 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33760         * tree-vrp.c (extract_range_from_binary_expr_1): Check
33761         range_int_cst_p rather than !symbolic_range_p before calling
33762         extract_range_from_multiplicative_op_1.
33764 2018-01-04  Jeff Law  <law@redhat.com>
33766         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
33767         redundant test in assertion.
33769 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33771         * doc/rtl.texi: Document machine_mode wrapper classes.
33773 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33775         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
33776         using tree_to_uhwi.
33778 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33780         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
33781         the VEC_PERM_EXPR fold to fail.
33783 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33785         PR debug/83585
33786         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
33787         to switched_sections.
33789 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33791         PR target/83680
33792         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
33793         test for d.testing.
33795 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
33797         PR target/83387
33798         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
33799         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
33801 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33803         PR debug/83666
33804         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
33805         is BLKmode and bitpos not zero or mode change is needed.
33807 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33809         PR target/83675
33810         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
33811         TARGET_VIS2.
33813 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
33815         PR target/83628
33816         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
33817         instead of MULT rtx.  Update all corresponding splitters.
33818         (*saddl_se): Ditto.
33819         (*ssub<modesuffix>): Ditto.
33820         (*ssubl_se): Ditto.
33821         (*cmp_sadd_di): Update split patterns.
33822         (*cmp_sadd_si): Ditto.
33823         (*cmp_sadd_sidi): Ditto.
33824         (*cmp_ssub_di): Ditto.
33825         (*cmp_ssub_si): Ditto.
33826         (*cmp_ssub_sidi): Ditto.
33827         * config/alpha/predicates.md (const23_operand): New predicate.
33828         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
33829         Look for ASHIFT, not MULT inner operand.
33830         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
33832 2018-01-04  Martin Liska  <mliska@suse.cz>
33834         PR gcov-profile/83669
33835         * gcov.c (output_intermediate_file): Add version to intermediate
33836         gcov file.
33837         * doc/gcov.texi: Document new field 'version' in intermediate
33838         file format. Fix location of '-k' option of gcov command.
33840 2018-01-04  Martin Liska  <mliska@suse.cz>
33842         PR ipa/82352
33843         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
33845 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33847         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
33849 2018-01-03  Martin Sebor  <msebor@redhat.com>
33851         PR tree-optimization/83655
33852         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
33853         checking calls with invalid arguments.
33855 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33857         * tree-vect-stmts.c (vect_get_store_rhs): New function.
33858         (vectorizable_mask_load_store): Delete.
33859         (vectorizable_call): Return false for masked loads and stores.
33860         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
33861         instead of gimple_assign_rhs1.
33862         (vectorizable_load): Handle IFN_MASK_LOAD.
33863         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
33865 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33867         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
33868         split out from..,
33869         (vectorizable_mask_load_store): ...here.
33870         (vectorizable_load): ...and here.
33872 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33874         * tree-vect-stmts.c (vect_build_all_ones_mask)
33875         (vect_build_zero_merge_argument): New functions, split out from...
33876         (vectorizable_load): ...here.
33878 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33880         * tree-vect-stmts.c (vect_check_store_rhs): New function,
33881         split out from...
33882         (vectorizable_mask_load_store): ...here.
33883         (vectorizable_store): ...and here.
33885 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33887         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
33888         split out from...
33889         (vectorizable_mask_load_store): ...here.
33891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33893         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
33894         (vect_model_store_cost): Take a vec_load_store_type instead of a
33895         vect_def_type.
33896         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
33897         (vect_model_store_cost): Take a vec_load_store_type instead of a
33898         vect_def_type.
33899         (vectorizable_mask_load_store): Update accordingly.
33900         (vectorizable_store): Likewise.
33901         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
33903 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33905         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
33906         IFN_MASK_LOAD calls here rather than...
33907         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
33909 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33910             Alan Hayward  <alan.hayward@arm.com>
33911             David Sherwood  <david.sherwood@arm.com>
33913         * expmed.c (extract_bit_field_1): For vector extracts,
33914         fall back to extract_bit_field_as_subreg if vec_extract
33915         isn't available.
33917 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33918             Alan Hayward  <alan.hayward@arm.com>
33919             David Sherwood  <david.sherwood@arm.com>
33921         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
33922         they are variable or constant sized.
33923         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
33924         slots for constant-sized data.
33926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33927             Alan Hayward  <alan.hayward@arm.com>
33928             David Sherwood  <david.sherwood@arm.com>
33930         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
33931         handling COND_EXPRs with boolean comparisons, try to find a better
33932         basis for the mask type than the boolean itself.
33934 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33936         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
33937         is calculated and how it can be overridden.
33938         * genmodes.c (max_bitsize_mode_any_mode): New variable.
33939         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
33940         if defined.
33941         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
33942         if nonzero.
33944 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33945             Alan Hayward  <alan.hayward@arm.com>
33946             David Sherwood  <david.sherwood@arm.com>
33948         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
33949         Remove the mode argument.
33950         (aarch64_simd_valid_immediate): Remove the mode and inverse
33951         arguments.
33952         * config/aarch64/iterators.md (bitsize): New iterator.
33953         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
33954         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
33955         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
33956         aarch64_simd_valid_immediate.
33957         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
33958         (aarch64_reg_or_bic_imm): Likewise.
33959         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
33960         with an insn_type enum and msl with a modifier_type enum.
33961         Replace element_width with a scalar_mode.  Change the shift
33962         to unsigned int.  Add constructors for scalar_float_mode and
33963         scalar_int_mode elements.
33964         (aarch64_vect_float_const_representable_p): Delete.
33965         (aarch64_can_const_movi_rtx_p)
33966         (aarch64_simd_scalar_immediate_valid_for_move)
33967         (aarch64_simd_make_constant): Update call to
33968         aarch64_simd_valid_immediate.
33969         (aarch64_advsimd_valid_immediate_hs): New function.
33970         (aarch64_advsimd_valid_immediate): Likewise.
33971         (aarch64_simd_valid_immediate): Remove mode and inverse
33972         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
33973         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
33974         and aarch64_float_const_representable_p on the result.
33975         (aarch64_output_simd_mov_immediate): Remove mode argument.
33976         Update call to aarch64_simd_valid_immediate and use of
33977         simd_immediate_info.
33978         (aarch64_output_scalar_simd_mov_immediate): Update call
33979         accordingly.
33981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33982             Alan Hayward  <alan.hayward@arm.com>
33983             David Sherwood  <david.sherwood@arm.com>
33985         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
33986         (mode_nunits): Likewise CONST_MODE_NUNITS.
33987         * machmode.def (ADJUST_NUNITS): Document.
33988         * genmodes.c (mode_data::need_nunits_adj): New field.
33989         (blank_mode): Update accordingly.
33990         (adj_nunits): New variable.
33991         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
33992         parameter.
33993         (emit_mode_size_inline): Set need_bytesize_adj for all modes
33994         listed in adj_nunits.
33995         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
33996         listed in adj_nunits.  Don't emit case statements for such modes.
33997         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
33998         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
33999         nothing if adj_nunits is nonnull.
34000         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
34001         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
34002         (emit_mode_fbit): Update use of print_maybe_const_decl.
34003         (emit_move_size): Likewise.  Treat the array as non-const
34004         if adj_nunits.
34005         (emit_mode_adjustments): Handle adj_nunits.
34007 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34009         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
34010         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
34011         (VECTOR_MODES): Use it.
34012         (make_vector_modes): Take the prefix as an argument.
34014 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34015             Alan Hayward  <alan.hayward@arm.com>
34016             David Sherwood  <david.sherwood@arm.com>
34018         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
34019         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
34020         for MODE_VECTOR_BOOL.
34021         * machmode.def (VECTOR_BOOL_MODE): Document.
34022         * genmodes.c (VECTOR_BOOL_MODE): New macro.
34023         (make_vector_bool_mode): New function.
34024         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
34025         MODE_VECTOR_BOOL.
34026         * lto-streamer-in.c (lto_input_mode_table): Likewise.
34027         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
34028         Likewise.
34029         * stor-layout.c (int_mode_for_mode): Likewise.
34030         * tree.c (build_vector_type_for_mode): Likewise.
34031         * varasm.c (output_constant_pool_2): Likewise.
34032         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
34033         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
34034         for MODE_VECTOR_BOOL.
34035         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
34036         of mode class checks.
34037         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
34038         instead of a list of mode class checks.
34039         (expand_vector_scalar_condition): Likewise.
34040         (type_for_widest_vector_mode): Handle BImode as an inner mode.
34042 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34043             Alan Hayward  <alan.hayward@arm.com>
34044             David Sherwood  <david.sherwood@arm.com>
34046         * machmode.h (mode_size): Change from unsigned short to
34047         poly_uint16_pod.
34048         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
34049         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34050         or if measurement_type is not polynomial.
34051         (fixed_size_mode::includes_p): Check for constant-sized modes.
34052         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
34053         return a poly_uint16 rather than an unsigned short.
34054         (emit_mode_size): Change the type of mode_size from unsigned short
34055         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
34056         (emit_mode_adjustments): Cope with polynomial vector sizes.
34057         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34058         for GET_MODE_SIZE.
34059         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34060         for GET_MODE_SIZE.
34061         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
34062         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
34063         * caller-save.c (setup_save_areas): Likewise.
34064         (replace_reg_with_saved_mem): Likewise.
34065         * calls.c (emit_library_call_value_1): Likewise.
34066         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
34067         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
34068         (gen_lowpart_for_combine): Likewise.
34069         * convert.c (convert_to_integer_1): Likewise.
34070         * cse.c (equiv_constant, cse_insn): Likewise.
34071         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
34072         (cselib_subst_to_values): Likewise.
34073         * dce.c (word_dce_process_block): Likewise.
34074         * df-problems.c (df_word_lr_mark_ref): Likewise.
34075         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
34076         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
34077         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
34078         (rtl_for_decl_location): Likewise.
34079         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
34080         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
34081         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
34082         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
34083         (expand_expr_real_1): Likewise.
34084         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
34085         (pad_below): Likewise.
34086         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34087         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
34088         * ira.c (get_subreg_tracking_sizes): Likewise.
34089         * ira-build.c (ira_create_allocno_objects): Likewise.
34090         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
34091         (ira_sort_regnos_for_alter_reg): Likewise.
34092         * ira-costs.c (record_operand_costs): Likewise.
34093         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
34094         (resolve_simple_move): Likewise.
34095         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
34096         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
34097         (lra_constraints): Likewise.
34098         (CONST_POOL_OK_P): Reject variable-sized modes.
34099         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
34100         (add_pseudo_to_slot, lra_spill): Likewise.
34101         * omp-low.c (omp_clause_aligned_alignment): Likewise.
34102         * optabs-query.c (get_best_extraction_insn): Likewise.
34103         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34104         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
34105         (expand_mult_highpart, valid_multiword_target_p): Likewise.
34106         * recog.c (offsettable_address_addr_space_p): Likewise.
34107         * regcprop.c (maybe_mode_change): Likewise.
34108         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
34109         * regrename.c (build_def_use): Likewise.
34110         * regstat.c (dump_reg_info): Likewise.
34111         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
34112         (find_reloads, find_reloads_subreg_address): Likewise.
34113         * reload1.c (eliminate_regs_1): Likewise.
34114         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
34115         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
34116         (simplify_binary_operation_1, simplify_subreg): Likewise.
34117         * targhooks.c (default_function_arg_padding): Likewise.
34118         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
34119         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
34120         (verify_gimple_assign_ternary): Likewise.
34121         * tree-inline.c (estimate_move_cost): Likewise.
34122         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34123         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
34124         (get_address_cost_ainc): Likewise.
34125         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
34126         (vect_supportable_dr_alignment): Likewise.
34127         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34128         (vectorizable_reduction): Likewise.
34129         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
34130         (vectorizable_operation, vectorizable_load): Likewise.
34131         * tree.c (build_same_sized_truth_vector_type): Likewise.
34132         * valtrack.c (cleanup_auto_inc_dec): Likewise.
34133         * var-tracking.c (emit_note_insn_var_location): Likewise.
34134         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
34135         (ADDR_VEC_ALIGN): Likewise.
34137 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34138             Alan Hayward  <alan.hayward@arm.com>
34139             David Sherwood  <david.sherwood@arm.com>
34141         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
34142         unsigned short.
34143         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34144         or if measurement_type is polynomial.
34145         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
34146         * combine.c (make_extraction): Likewise.
34147         * dse.c (find_shift_sequence): Likewise.
34148         * dwarf2out.c (mem_loc_descriptor): Likewise.
34149         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
34150         (extract_bit_field, extract_low_bits): Likewise.
34151         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
34152         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
34153         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
34154         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
34155         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34156         * reload.c (find_reloads): Likewise.
34157         * reload1.c (alter_reg): Likewise.
34158         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
34159         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
34160         * tree-if-conv.c (predicate_mem_writes): Likewise.
34161         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
34162         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
34163         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
34164         * valtrack.c (dead_debug_insert_temp): Likewise.
34165         * varasm.c (mergeable_constant_section): Likewise.
34166         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
34168 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34169             Alan Hayward  <alan.hayward@arm.com>
34170             David Sherwood  <david.sherwood@arm.com>
34172         * expr.c (expand_assignment): Cope with polynomial mode sizes
34173         when assigning to a CONCAT.
34175 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34176             Alan Hayward  <alan.hayward@arm.com>
34177             David Sherwood  <david.sherwood@arm.com>
34179         * machmode.h (mode_precision): Change from unsigned short to
34180         poly_uint16_pod.
34181         (mode_to_precision): Return a poly_uint16 rather than an unsigned
34182         short.
34183         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
34184         or if measurement_type is not polynomial.
34185         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
34186         in which the mode is already known to be a scalar_int_mode.
34187         * genmodes.c (emit_mode_precision): Change the type of mode_precision
34188         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
34189         initializer.
34190         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34191         for GET_MODE_PRECISION.
34192         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34193         for GET_MODE_PRECISION.
34194         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
34195         as polynomial.
34196         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
34197         (expand_field_assignment, make_extraction): Likewise.
34198         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
34199         (get_last_value): Likewise.
34200         * convert.c (convert_to_integer_1): Likewise.
34201         * cse.c (cse_insn): Likewise.
34202         * expr.c (expand_expr_real_1): Likewise.
34203         * lra-constraints.c (simplify_operand_subreg): Likewise.
34204         * optabs-query.c (can_atomic_load_p): Likewise.
34205         * optabs.c (expand_atomic_load): Likewise.
34206         (expand_atomic_store): Likewise.
34207         * ree.c (combine_reaching_defs): Likewise.
34208         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
34209         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
34210         * tree.h (type_has_mode_precision_p): Likewise.
34211         * ubsan.c (instrument_si_overflow): Likewise.
34213 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34214             Alan Hayward  <alan.hayward@arm.com>
34215             David Sherwood  <david.sherwood@arm.com>
34217         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
34218         polynomial numbers of units.
34219         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
34220         (valid_vector_subparts_p): New function.
34221         (build_vector_type): Remove temporary shim and take the number
34222         of units as a poly_uint64 rather than an int.
34223         (build_opaque_vector_type): Take the number of units as a
34224         poly_uint64 rather than an int.
34225         * tree.c (build_vector_from_ctor): Handle polynomial
34226         TYPE_VECTOR_SUBPARTS.
34227         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
34228         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
34229         (build_vector_from_val): If the number of units is variable,
34230         use build_vec_duplicate_cst for constant operands and
34231         VEC_DUPLICATE_EXPR otherwise.
34232         (make_vector_type): Remove temporary is_constant ().
34233         (build_vector_type, build_opaque_vector_type): Take the number of
34234         units as a poly_uint64 rather than an int.
34235         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
34236         VECTOR_CST_NELTS.
34237         * cfgexpand.c (expand_debug_expr): Likewise.
34238         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
34239         (store_constructor, expand_expr_real_1): Likewise.
34240         (const_scalar_mask_from_tree): Likewise.
34241         * fold-const-call.c (fold_const_reduction): Likewise.
34242         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
34243         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
34244         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
34245         (fold_relational_const): Likewise.
34246         (native_interpret_vector): Likewise.  Change the size from an
34247         int to an unsigned int.
34248         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
34249         TYPE_VECTOR_SUBPARTS.
34250         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
34251         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
34252         duplicating a non-constant operand into a variable-length vector.
34253         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
34254         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
34255         * ipa-icf.c (sem_variable::equals): Likewise.
34256         * match.pd: Likewise.
34257         * omp-simd-clone.c (simd_clone_subparts): Likewise.
34258         * print-tree.c (print_node): Likewise.
34259         * stor-layout.c (layout_type): Likewise.
34260         * targhooks.c (default_builtin_vectorization_cost): Likewise.
34261         * tree-cfg.c (verify_gimple_comparison): Likewise.
34262         (verify_gimple_assign_binary): Likewise.
34263         (verify_gimple_assign_ternary): Likewise.
34264         (verify_gimple_assign_single): Likewise.
34265         * tree-pretty-print.c (dump_generic_node): Likewise.
34266         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34267         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
34268         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
34269         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
34270         (vect_shift_permute_load_chain): Likewise.
34271         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
34272         (expand_vector_condition, optimize_vector_constructor): Likewise.
34273         (lower_vec_perm, get_compute_type): Likewise.
34274         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34275         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
34276         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
34277         (vect_recog_mask_conversion_pattern): Likewise.
34278         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
34279         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
34280         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34281         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
34282         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
34283         (vectorizable_shift, vectorizable_operation, vectorizable_store)
34284         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
34285         (supportable_widening_operation): Likewise.
34286         (supportable_narrowing_operation): Likewise.
34287         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
34288         Likewise.
34289         * varasm.c (output_constant): Likewise.
34291 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34292             Alan Hayward  <alan.hayward@arm.com>
34293             David Sherwood  <david.sherwood@arm.com>
34295         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
34296         so that both the length == 3 and length != 3 cases set up their
34297         own permute vectors.  Add comments explaining why we know the
34298         number of elements is constant.
34299         (vect_permute_load_chain): Likewise.
34301 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34302             Alan Hayward  <alan.hayward@arm.com>
34303             David Sherwood  <david.sherwood@arm.com>
34305         * machmode.h (mode_nunits): Change from unsigned char to
34306         poly_uint16_pod.
34307         (ONLY_FIXED_SIZE_MODES): New macro.
34308         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
34309         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
34310         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
34311         New typedefs.
34312         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
34313         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
34314         or if measurement_type is not polynomial.
34315         * genmodes.c (ZERO_COEFFS): New macro.
34316         (emit_mode_nunits_inline): Make mode_nunits_inline return a
34317         poly_uint16.
34318         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
34319         Use ZERO_COEFFS when emitting initializers.
34320         * data-streamer.h (bp_pack_poly_value): New function.
34321         (bp_unpack_poly_value): Likewise.
34322         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34323         for GET_MODE_NUNITS.
34324         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34325         for GET_MODE_NUNITS.
34326         * tree.c (make_vector_type): Remove temporary shim and make
34327         the real function take the number of units as a poly_uint64
34328         rather than an int.
34329         (build_vector_type_for_mode): Handle polynomial nunits.
34330         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
34331         * emit-rtl.c (const_vec_series_p_1): Likewise.
34332         (gen_rtx_CONST_VECTOR): Likewise.
34333         * fold-const.c (test_vec_duplicate_folding): Likewise.
34334         * genrecog.c (validate_pattern): Likewise.
34335         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
34336         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34337         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
34338         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
34339         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
34340         * rtlanal.c (subreg_get_info): Likewise.
34341         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34342         (vect_grouped_load_supported): Likewise.
34343         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
34344         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
34345         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
34346         (simplify_const_unary_operation, simplify_binary_operation_1)
34347         (simplify_const_binary_operation, simplify_ternary_operation)
34348         (test_vector_ops_duplicate, test_vector_ops): Likewise.
34349         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
34350         instead of CONST_VECTOR_NUNITS.
34351         * varasm.c (output_constant_pool_2): Likewise.
34352         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
34353         explicit-encoded elements in the XVEC for variable-length vectors.
34355 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34357         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
34359 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34360             Alan Hayward  <alan.hayward@arm.com>
34361             David Sherwood  <david.sherwood@arm.com>
34363         * coretypes.h (fixed_size_mode): Declare.
34364         (fixed_size_mode_pod): New typedef.
34365         * builtins.h (target_builtins::x_apply_args_mode)
34366         (target_builtins::x_apply_result_mode): Change type to
34367         fixed_size_mode_pod.
34368         * builtins.c (apply_args_size, apply_result_size, result_vector)
34369         (expand_builtin_apply_args_1, expand_builtin_apply)
34370         (expand_builtin_return): Update accordingly.
34372 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34374         * cse.c (hash_rtx_cb): Hash only the encoded elements.
34375         * cselib.c (cselib_hash_rtx): Likewise.
34376         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
34377         CONST_VECTOR encoding.
34379 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
34380             Jeff Law  <law@redhat.com>
34382         PR target/83641
34383         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
34384         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
34385         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
34386         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
34388         PR target/83641
34389         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
34390         explicitly probe *sp in a noreturn function if there were any callee
34391         register saves or frame pointer is needed.
34393 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
34395         PR debug/83621
34396         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
34397         BLKmode for ternary, binary or unary expressions.
34399         PR debug/83645
34400         * var-tracking.c (delete_vta_debug_insn): New inline function.
34401         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
34402         insns from get_insns () to NULL instead of each bb separately.
34403         Use delete_vta_debug_insn.  No longer static.
34404         (vt_debug_insns_local, variable_tracking_main_1): Adjust
34405         delete_vta_debug_insns callers.
34406         * rtl.h (delete_vta_debug_insns): Declare.
34407         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
34408         instead of variable_tracking_main.
34410 2018-01-03  Martin Sebor  <msebor@redhat.com>
34412         PR tree-optimization/83603
34413         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
34414         arguments past the endof the argument list in functions declared
34415         without a prototype.
34416         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
34417         Avoid checking when arguments are null.
34419 2018-01-03  Martin Sebor  <msebor@redhat.com>
34421         PR c/83559
34422         * doc/extend.texi (attribute const): Fix a typo.
34423         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
34424         issuing -Wsuggest-attribute for void functions.
34426 2018-01-03  Martin Sebor  <msebor@redhat.com>
34428         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
34429         offset_int::from instead of wide_int::to_shwi.
34430         (maybe_diag_overlap): Remove assertion.
34431         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
34432         * gimple-ssa-sprintf.c (format_directive): Same.
34433         (parse_directive): Same.
34434         (sprintf_dom_walker::compute_format_length): Same.
34435         (try_substitute_return_value): Same.
34437 2018-01-03  Jeff Law  <law@redhat.com>
34439         PR middle-end/83654
34440         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
34441         non-constant residual for zero at runtime and avoid probing in
34442         that case.  Reorganize code for trailing problem to mirror handling
34443         of the residual.
34445 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
34447         PR tree-optimization/83501
34448         * tree-ssa-strlen.c (get_string_cst): New.
34449         (handle_char_store): Call get_string_cst.
34451 2018-01-03  Martin Liska  <mliska@suse.cz>
34453         PR tree-optimization/83593
34454         * tree-ssa-strlen.c: Include tree-cfg.h.
34455         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
34456         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
34457         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
34458         to false.
34459         (strlen_dom_walker::before_dom_children): Call
34460         gimple_purge_dead_eh_edges. Dump tranformation with details
34461         dump flags.
34462         (strlen_dom_walker::before_dom_children): Update call by adding
34463         new argument cleanup_eh.
34464         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
34466 2018-01-03  Martin Liska  <mliska@suse.cz>
34468         PR ipa/83549
34469         * cif-code.def (VARIADIC_THUNK): New enum value.
34470         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
34471         thunks.
34473 2018-01-03  Jan Beulich  <jbeulich@suse.com>
34475         * sse.md (mov<mode>_internal): Tighten condition for when to use
34476         vmovdqu<ssescalarsize> for TI and OI modes.
34478 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
34480         Update copyright years.
34482 2018-01-03  Martin Liska  <mliska@suse.cz>
34484         PR ipa/83594
34485         * ipa-visibility.c (function_and_variable_visibility): Skip
34486         functions with noipa attribure.
34488 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
34490         * gcc.c (process_command): Update copyright notice dates.
34491         * gcov-dump.c (print_version): Ditto.
34492         * gcov.c (print_version): Ditto.
34493         * gcov-tool.c (print_version): Ditto.
34494         * gengtype.c (create_file): Ditto.
34495         * doc/cpp.texi: Bump @copying's copyright year.
34496         * doc/cppinternals.texi: Ditto.
34497         * doc/gcc.texi: Ditto.
34498         * doc/gccint.texi: Ditto.
34499         * doc/gcov.texi: Ditto.
34500         * doc/install.texi: Ditto.
34501         * doc/invoke.texi: Ditto.
34503 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34505         * vector-builder.h (vector_builder::m_full_nelts): Change from
34506         unsigned int to poly_uint64.
34507         (vector_builder::full_nelts): Update prototype accordingly.
34508         (vector_builder::new_vector): Likewise.
34509         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
34510         (vector_builder::operator ==): Likewise.
34511         (vector_builder::finalize): Likewise.
34512         * int-vector-builder.h (int_vector_builder::int_vector_builder):
34513         Take the number of elements as a poly_uint64 rather than an
34514         unsigned int.
34515         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
34516         from unsigned int to poly_uint64.
34517         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
34518         (vec_perm_indices::new_vector): Likewise.
34519         (vec_perm_indices::length): Likewise.
34520         (vec_perm_indices::nelts_per_input): Likewise.
34521         (vec_perm_indices::input_nelts): Likewise.
34522         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
34523         number of elements per input as a poly_uint64 rather than an
34524         unsigned int.  Use the original encoding for variable-length
34525         vectors, rather than clamping each individual element.
34526         For the second and subsequent elements in each pattern,
34527         clamp the step and base before clamping their sum.
34528         (vec_perm_indices::series_p): Handle polynomial element counts.
34529         (vec_perm_indices::all_in_range_p): Likewise.
34530         (vec_perm_indices_to_tree): Likewise.
34531         (vec_perm_indices_to_rtx): Likewise.
34532         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
34533         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
34534         (tree_vector_builder::new_binary_operation): Handle polynomial
34535         element counts.  Return false if we need to know the number
34536         of elements at compile time.
34537         * fold-const.c (fold_vec_perm): Punt if the number of elements
34538         isn't known at compile time.
34540 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34542         * vec-perm-indices.h (vec_perm_builder): Change element type
34543         from HOST_WIDE_INT to poly_int64.
34544         (vec_perm_indices::element_type): Update accordingly.
34545         (vec_perm_indices::clamp): Handle polynomial element_types.
34546         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
34547         (vec_perm_indices::all_in_range_p): Likewise.
34548         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
34549         than shwi trees.
34550         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
34551         polynomial vec_perm_indices element types.
34552         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
34553         * fold-const.c (fold_vec_perm): Likewise.
34554         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
34555         * tree-vect-generic.c (lower_vec_perm): Likewise.
34556         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
34557         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
34558         element type to HOST_WIDE_INT.
34560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34561             Alan Hayward  <alan.hayward@arm.com>
34562             David Sherwood  <david.sherwood@arm.com>
34564         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
34565         rather than an int.  Use plus_constant.
34566         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
34567         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
34569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34570             Alan Hayward  <alan.hayward@arm.com>
34571             David Sherwood  <david.sherwood@arm.com>
34573         * calls.c (emit_call_1, expand_call): Change struct_value_size from
34574         a HOST_WIDE_INT to a poly_int64.
34576 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34577             Alan Hayward  <alan.hayward@arm.com>
34578             David Sherwood  <david.sherwood@arm.com>
34580         * calls.c (load_register_parameters): Cope with polynomial
34581         mode sizes.  Require a constant size for BLKmode parameters
34582         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
34583         forces a parameter to be padded at the lsb end in order to
34584         fill a complete number of words, require the parameter size
34585         to be ordered wrt UNITS_PER_WORD.
34587 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34588             Alan Hayward  <alan.hayward@arm.com>
34589             David Sherwood  <david.sherwood@arm.com>
34591         * reload1.c (spill_stack_slot_width): Change element type
34592         from unsigned int to poly_uint64_pod.
34593         (alter_reg): Treat mode sizes as polynomial.
34595 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34596             Alan Hayward  <alan.hayward@arm.com>
34597             David Sherwood  <david.sherwood@arm.com>
34599         * reload.c (complex_word_subreg_p): New function.
34600         (reload_inner_reg_of_subreg, push_reload): Use it.
34602 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34603             Alan Hayward  <alan.hayward@arm.com>
34604             David Sherwood  <david.sherwood@arm.com>
34606         * lra-constraints.c (process_alt_operands): Reject matched
34607         operands whose sizes aren't ordered.
34608         (match_reload): Refer to this check here.
34610 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34611             Alan Hayward  <alan.hayward@arm.com>
34612             David Sherwood  <david.sherwood@arm.com>
34614         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
34615         that the mode size is in the set {1, 2, 4, 8, 16}.
34617 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34618             Alan Hayward  <alan.hayward@arm.com>
34619             David Sherwood  <david.sherwood@arm.com>
34621         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
34622         Use plus_constant instead of gen_rtx_PLUS.
34624 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34625             Alan Hayward  <alan.hayward@arm.com>
34626             David Sherwood  <david.sherwood@arm.com>
34628         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
34629         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
34630         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
34631         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
34632         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
34633         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
34634         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
34635         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
34636         * config/i386/i386.c (ix86_push_rounding): ...this new function.
34637         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
34638         a poly_int64.
34639         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
34640         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
34641         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
34642         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
34643         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
34644         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
34645         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
34646         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
34647         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
34648         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
34649         function.
34650         * expr.c (emit_move_resolve_push): Treat the input and result
34651         of PUSH_ROUNDING as a poly_int64.
34652         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
34653         (emit_push_insn): Likewise.
34654         * lra-eliminations.c (mark_not_eliminable): Likewise.
34655         * recog.c (push_operand): Likewise.
34656         * reload1.c (elimination_effects): Likewise.
34657         * rtlanal.c (nonzero_bits1): Likewise.
34658         * calls.c (store_one_arg): Likewise.  Require the padding to be
34659         known at compile time.
34661 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34662             Alan Hayward  <alan.hayward@arm.com>
34663             David Sherwood  <david.sherwood@arm.com>
34665         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
34666         Use plus_constant instead of gen_rtx_PLUS.
34668 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34669             Alan Hayward  <alan.hayward@arm.com>
34670             David Sherwood  <david.sherwood@arm.com>
34672         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
34673         rather than an int.
34675 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34676             Alan Hayward  <alan.hayward@arm.com>
34677             David Sherwood  <david.sherwood@arm.com>
34679         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
34680         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
34681         via stack temporaries.  Treat the mode size as polynomial too.
34683 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34684             Alan Hayward  <alan.hayward@arm.com>
34685             David Sherwood  <david.sherwood@arm.com>
34687         * expr.c (expand_expr_real_2): When handling conversions involving
34688         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
34689         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
34690         as a poly_uint64 too.
34692 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34693             Alan Hayward  <alan.hayward@arm.com>
34694             David Sherwood  <david.sherwood@arm.com>
34696         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
34698 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34699             Alan Hayward  <alan.hayward@arm.com>
34700             David Sherwood  <david.sherwood@arm.com>
34702         * combine.c (can_change_dest_mode): Handle polynomial
34703         REGMODE_NATURAL_SIZE.
34704         * expmed.c (store_bit_field_1): Likewise.
34705         * expr.c (store_constructor): Likewise.
34706         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
34707         and polynomial REGMODE_NATURAL_SIZE.
34708         (gen_lowpart_common): Likewise.
34709         * reginfo.c (record_subregs_of_mode): Likewise.
34710         * rtlanal.c (read_modify_subreg_p): Likewise.
34712 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34713             Alan Hayward  <alan.hayward@arm.com>
34714             David Sherwood  <david.sherwood@arm.com>
34716         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
34717         numbers of elements.
34719 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34720             Alan Hayward  <alan.hayward@arm.com>
34721             David Sherwood  <david.sherwood@arm.com>
34723         * match.pd: Cope with polynomial numbers of vector elements.
34725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34726             Alan Hayward  <alan.hayward@arm.com>
34727             David Sherwood  <david.sherwood@arm.com>
34729         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
34730         in a POINTER_PLUS_EXPR.
34732 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34733             Alan Hayward  <alan.hayward@arm.com>
34734             David Sherwood  <david.sherwood@arm.com>
34736         * omp-simd-clone.c (simd_clone_subparts): New function.
34737         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
34738         (ipa_simd_modify_function_body): Likewise.
34740 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34741             Alan Hayward  <alan.hayward@arm.com>
34742             David Sherwood  <david.sherwood@arm.com>
34744         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
34745         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
34746         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
34747         (expand_vector_condition, vector_element): Likewise.
34748         (subparts_gt): New function.
34749         (get_compute_type): Use subparts_gt.
34750         (count_type_subparts): Delete.
34751         (expand_vector_operations_1): Use subparts_gt instead of
34752         count_type_subparts.
34754 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34755             Alan Hayward  <alan.hayward@arm.com>
34756             David Sherwood  <david.sherwood@arm.com>
34758         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
34759         (vect_compile_time_alias): ...this new function.  Do the calculation
34760         on poly_ints rather than trees.
34761         (vect_prune_runtime_alias_test_list): Update call accordingly.
34763 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34764             Alan Hayward  <alan.hayward@arm.com>
34765             David Sherwood  <david.sherwood@arm.com>
34767         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
34768         numbers of units.
34769         (vect_schedule_slp_instance): Likewise.
34771 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34772             Alan Hayward  <alan.hayward@arm.com>
34773             David Sherwood  <david.sherwood@arm.com>
34775         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
34776         constant and extern definitions for variable-length vectors.
34777         (vect_get_constant_vectors): Note that the number of units
34778         is known to be constant.
34780 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34781             Alan Hayward  <alan.hayward@arm.com>
34782             David Sherwood  <david.sherwood@arm.com>
34784         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
34785         of units as polynomial.  Choose between WIDE and NARROW based
34786         on multiple_p.
34788 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34789             Alan Hayward  <alan.hayward@arm.com>
34790             David Sherwood  <david.sherwood@arm.com>
34792         * tree-vect-stmts.c (simd_clone_subparts): New function.
34793         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
34795 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34796             Alan Hayward  <alan.hayward@arm.com>
34797             David Sherwood  <david.sherwood@arm.com>
34799         * tree-vect-stmts.c (vectorizable_call): Treat the number of
34800         vectors as polynomial.  Use build_index_vector for
34801         IFN_GOMP_SIMD_LANE.
34803 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34804             Alan Hayward  <alan.hayward@arm.com>
34805             David Sherwood  <david.sherwood@arm.com>
34807         * tree-vect-stmts.c (get_load_store_type): Treat the number of
34808         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
34809         for variable-length vectors.
34810         (vectorizable_mask_load_store): Treat the number of units as
34811         polynomial, asserting that it is constant if the condition has
34812         already been enforced.
34813         (vectorizable_store, vectorizable_load): Likewise.
34815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34816             Alan Hayward  <alan.hayward@arm.com>
34817             David Sherwood  <david.sherwood@arm.com>
34819         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
34820         of units as polynomial.  Punt if we can't tell at compile time
34821         which vector contains the final result.
34823 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34824             Alan Hayward  <alan.hayward@arm.com>
34825             David Sherwood  <david.sherwood@arm.com>
34827         * tree-vect-loop.c (vectorizable_induction): Treat the number
34828         of units as polynomial.  Punt on SLP inductions.  Use an integer
34829         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
34830         cast of such a series for variable-length floating-point
34831         reductions.
34833 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34834             Alan Hayward  <alan.hayward@arm.com>
34835             David Sherwood  <david.sherwood@arm.com>
34837         * tree.h (build_index_vector): Declare.
34838         * tree.c (build_index_vector): New function.
34839         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
34840         of units as polynomial, forcibly converting it to a constant if
34841         vectorizable_reduction has already enforced the condition.
34842         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
34843         to create a {1,2,3,...} vector.
34844         (vectorizable_reduction): Treat the number of units as polynomial.
34845         Choose vectype_in based on the largest scalar element size rather
34846         than the smallest number of units.  Enforce the restrictions
34847         relied on above.
34849 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34850             Alan Hayward  <alan.hayward@arm.com>
34851             David Sherwood  <david.sherwood@arm.com>
34853         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
34854         number of units as polynomial.
34856 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34857             Alan Hayward  <alan.hayward@arm.com>
34858             David Sherwood  <david.sherwood@arm.com>
34860         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
34861         * target.def (autovectorize_vector_sizes): Return the vector sizes
34862         by pointer, using vector_sizes rather than a bitmask.
34863         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
34864         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
34865         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
34866         Likewise.
34867         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
34868         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
34869         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
34870         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
34871         * omp-general.c (omp_max_vf): Likewise.
34872         * omp-low.c (omp_clause_aligned_alignment): Likewise.
34873         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
34874         * tree-vect-loop.c (vect_analyze_loop): Likewise.
34875         * tree-vect-slp.c (vect_slp_bb): Likewise.
34876         * doc/tm.texi: Regenerate.
34877         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
34878         to a poly_uint64.
34879         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
34880         the vector size as a poly_uint64 rather than an unsigned int.
34881         (current_vector_size): Change from an unsigned int to a poly_uint64.
34882         (get_vectype_for_scalar_type): Update accordingly.
34883         * tree.h (build_truth_vector_type): Take the size and number of
34884         units as a poly_uint64 rather than an unsigned int.
34885         (build_vector_type): Add a temporary overload that takes
34886         the number of units as a poly_uint64 rather than an unsigned int.
34887         * tree.c (make_vector_type): Likewise.
34888         (build_truth_vector_type): Take the number of units as a poly_uint64
34889         rather than an unsigned int.
34891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34892             Alan Hayward  <alan.hayward@arm.com>
34893             David Sherwood  <david.sherwood@arm.com>
34895         * target.def (get_mask_mode): Take the number of units and length
34896         as poly_uint64s rather than unsigned ints.
34897         * targhooks.h (default_get_mask_mode): Update accordingly.
34898         * targhooks.c (default_get_mask_mode): Likewise.
34899         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
34900         * doc/tm.texi: Regenerate.
34902 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34903             Alan Hayward  <alan.hayward@arm.com>
34904             David Sherwood  <david.sherwood@arm.com>
34906         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
34907         * omp-general.c (omp_max_vf): Likewise.
34908         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
34909         (expand_omp_simd): Handle polynomial safelen.
34910         * omp-low.c (omplow_simd_context): Add a default constructor.
34911         (omplow_simd_context::max_vf): Change from int to poly_uint64.
34912         (lower_rec_simd_input_clauses): Update accordingly.
34913         (lower_rec_input_clauses): Likewise.
34915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34916             Alan Hayward  <alan.hayward@arm.com>
34917             David Sherwood  <david.sherwood@arm.com>
34919         * tree-vectorizer.h (vect_nunits_for_cost): New function.
34920         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
34921         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
34922         (vect_analyze_slp_cost): Likewise.
34923         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
34924         (vect_model_load_cost): Likewise.
34926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34927             Alan Hayward  <alan.hayward@arm.com>
34928             David Sherwood  <david.sherwood@arm.com>
34930         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
34931         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
34932         from an unsigned int * to a poly_uint64_pod *.
34933         (calculate_unrolling_factor): New function.
34934         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
34936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34937             Alan Hayward  <alan.hayward@arm.com>
34938             David Sherwood  <david.sherwood@arm.com>
34940         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
34941         from an unsigned int to a poly_uint64.
34942         (_loop_vec_info::slp_unrolling_factor): Likewise.
34943         (_loop_vec_info::vectorization_factor): Change from an int
34944         to a poly_uint64.
34945         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
34946         (vect_get_num_vectors): New function.
34947         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
34948         (vect_get_num_copies): Use vect_get_num_vectors.
34949         (vect_analyze_data_ref_dependences): Change max_vf from an int *
34950         to an unsigned int *.
34951         (vect_analyze_data_refs): Change min_vf from an int * to a
34952         poly_uint64 *.
34953         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34954         than an unsigned HOST_WIDE_INT.
34955         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
34956         (vect_analyze_data_ref_dependence): Change max_vf from an int *
34957         to an unsigned int *.
34958         (vect_analyze_data_ref_dependences): Likewise.
34959         (vect_compute_data_ref_alignment): Handle polynomial vf.
34960         (vect_enhance_data_refs_alignment): Likewise.
34961         (vect_prune_runtime_alias_test_list): Likewise.
34962         (vect_shift_permute_load_chain): Likewise.
34963         (vect_supportable_dr_alignment): Likewise.
34964         (dependence_distance_ge_vf): Take the vectorization factor as a
34965         poly_uint64 rather than an unsigned HOST_WIDE_INT.
34966         (vect_analyze_data_refs): Change min_vf from an int * to a
34967         poly_uint64 *.
34968         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
34969         vfm1 as a poly_uint64 rather than an int.  Make the same change
34970         for the returned bound_scalar.
34971         (vect_gen_vector_loop_niters): Handle polynomial vf.
34972         (vect_do_peeling): Likewise.  Update call to
34973         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
34974         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
34975         be constant.
34976         * tree-vect-loop.c (vect_determine_vectorization_factor)
34977         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
34978         (vect_get_known_peeling_cost): Likewise.
34979         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
34980         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
34981         (vect_transform_loop): Likewise.  Use the lowest possible VF when
34982         updating the upper bounds of the loop.
34983         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
34984         rather than an int.
34985         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
34986         polynomial unroll factors.
34987         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
34988         (vect_make_slp_decision): Likewise.
34989         (vect_supported_load_permutation_p): Likewise, and polynomial
34990         vf too.
34991         (vect_analyze_slp_cost): Handle polynomial vf.
34992         (vect_slp_analyze_node_operations): Likewise.
34993         (vect_slp_analyze_bb_1): Likewise.
34994         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34995         than an unsigned HOST_WIDE_INT.
34996         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
34997         (vectorizable_load): Handle polynomial vf.
34998         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
34999         a poly_uint64.
35000         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
35002 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35003             Alan Hayward  <alan.hayward@arm.com>
35004             David Sherwood  <david.sherwood@arm.com>
35006         * match.pd: Handle bit operations involving three constants
35007         and try to fold one pair.
35009 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35011         * tree-vect-loop-manip.c: Include gimple-fold.h.
35012         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
35013         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
35014         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
35015         Add a path that uses a step of VF instead of 1, but disable it
35016         for now.
35017         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
35018         and niters_no_overflow parameters.  Update calls to
35019         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
35020         Create a new SSA name if the latter choses to use a ste other
35021         than zero, and return it via niters_vector_mult_vf_var.
35022         * tree-vect-loop.c (vect_transform_loop): Update calls to
35023         vect_do_peeling, vect_gen_vector_loop_niters and
35024         slpeel_make_loop_iterate_ntimes.
35025         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
35026         (vect_gen_vector_loop_niters): Update declarations after above changes.
35028 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
35030         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
35031         128-bit round to integer instructions.
35032         (ceil<mode>2): Likewise.
35033         (btrunc<mode>2): Likewise.
35034         (round<mode>2): Likewise.
35036 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35038         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
35039         unaligned VSX load/store on P8/P9.
35040         (expand_block_clear): Allow the use of unaligned VSX
35041         load/store on P8/P9.
35043 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35045         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
35046         New function.
35047         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
35048         swap associated with both a load and a store.
35050 2018-01-02  Andrew Waterman  <andrew@sifive.com>
35052         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
35053         * config/riscv/riscv.md (clear_cache): Use it.
35055 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
35057         * web.c: Remove out-of-date comment.
35059 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35061         * expr.c (fixup_args_size_notes): Check that any existing
35062         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
35063         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
35064         (emit_single_push_insn): ...here.
35066 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35068         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
35069         (const_vector_encoded_nelts): New function.
35070         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
35071         (const_vector_int_elt, const_vector_elt): Declare.
35072         * emit-rtl.c (const_vector_int_elt_1): New function.
35073         (const_vector_elt): Likewise.
35074         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
35075         of CONST_VECTOR_ELT.
35077 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35079         * expr.c: Include rtx-vector-builder.h.
35080         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
35081         directly on the tree encoding.
35082         (const_vector_from_tree): Likewise.
35083         * optabs.c: Include rtx-vector-builder.h.
35084         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
35085         sequence of "u" values.
35086         * vec-perm-indices.c: Include rtx-vector-builder.h.
35087         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
35088         directly on the vec_perm_indices encoding.
35090 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35092         * doc/rtl.texi (const_vector): Describe new encoding scheme.
35093         * Makefile.in (OBJS): Add rtx-vector-builder.o.
35094         * rtx-vector-builder.h: New file.
35095         * rtx-vector-builder.c: Likewise.
35096         * rtl.h (rtx_def::u2): Add a const_vector field.
35097         (CONST_VECTOR_NPATTERNS): New macro.
35098         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
35099         (CONST_VECTOR_DUPLICATE_P): Likewise.
35100         (CONST_VECTOR_STEPPED_P): Likewise.
35101         (CONST_VECTOR_ENCODED_ELT): Likewise.
35102         (const_vec_duplicate_p): Check for a duplicated vector encoding.
35103         (unwrap_const_vec_duplicate): Likewise.
35104         (const_vec_series_p): Check for a non-duplicated vector encoding.
35105         Say that the function only returns true for integer vectors.
35106         * emit-rtl.c: Include rtx-vector-builder.h.
35107         (gen_const_vec_duplicate_1): Delete.
35108         (gen_const_vector): Call gen_const_vec_duplicate instead of
35109         gen_const_vec_duplicate_1.
35110         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
35111         (gen_const_vec_duplicate): Use rtx_vector_builder.
35112         (gen_const_vec_series): Likewise.
35113         (gen_rtx_CONST_VECTOR): Likewise.
35114         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
35115         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35116         Build a new vector rather than modifying a CONST_VECTOR in-place.
35117         (handle_special_swappables): Update call accordingly.
35118         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
35119         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35120         Build a new vector rather than modifying a CONST_VECTOR in-place.
35121         (handle_special_swappables): Update call accordingly.
35123 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35125         * simplify-rtx.c (simplify_const_binary_operation): Use
35126         CONST_VECTOR_ELT instead of XVECEXP.
35128 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35130         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
35131         the selector elements to be different from the data elements
35132         if the selector is a VECTOR_CST.
35133         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
35134         ssizetype for the selector.
35136 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35138         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
35139         before testing each element individually.
35140         * tree-vect-generic.c (lower_vec_perm): Likewise.
35142 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35144         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
35145         * selftest-run-tests.c (selftest::run_tests): Call it.
35146         * vector-builder.h (vector_builder::operator ==): New function.
35147         (vector_builder::operator !=): Likewise.
35148         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
35149         (vec_perm_indices::all_from_input_p): New function.
35150         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35151         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
35152         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
35153         instead of reading the VECTOR_CST directly.  Detect whether both
35154         vector inputs are the same before constructing the vec_perm_indices,
35155         and update the number of inputs argument accordingly.  Use the
35156         utility functions added above.  Only construct sel2 if we need to.
35158 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35160         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
35161         the broadcast of the low byte.
35162         (expand_mult_highpart): Use an explicit encoding for the permutes.
35163         * optabs-query.c (can_mult_highpart_p): Likewise.
35164         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
35165         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35166         (vectorizable_bswap): Likewise.
35167         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
35168         explicit encoding for the power-of-2 permutes.
35169         (vect_permute_store_chain): Likewise.
35170         (vect_grouped_load_supported): Likewise.
35171         (vect_permute_load_chain): Likewise.
35173 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35175         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
35176         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
35177         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
35178         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35179         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
35180         (vect_gen_perm_mask_any): Likewise.
35182 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35184         * int-vector-builder.h: New file.
35185         * vec-perm-indices.h: Include int-vector-builder.h.
35186         (vec_perm_indices): Redefine as an int_vector_builder.
35187         (auto_vec_perm_indices): Delete.
35188         (vec_perm_builder): Redefine as a stand-alone class.
35189         (vec_perm_indices::vec_perm_indices): New function.
35190         (vec_perm_indices::clamp): Likewise.
35191         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
35192         (vec_perm_indices::new_vector): New function.
35193         (vec_perm_indices::new_expanded_vector): Update for new
35194         vec_perm_indices class.
35195         (vec_perm_indices::rotate_inputs): New function.
35196         (vec_perm_indices::all_in_range_p): Operate directly on the
35197         encoded form, without computing elided elements.
35198         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
35199         encoding.  Update for new vec_perm_indices class.
35200         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
35201         the given vec_perm_builder.
35202         (expand_vec_perm_var): Update vec_perm_builder constructor.
35203         (expand_mult_highpart): Use vec_perm_builder instead of
35204         auto_vec_perm_indices.
35205         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
35206         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
35207         or double series encoding as appropriate.
35208         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
35209         vec_perm_indices instead of auto_vec_perm_indices.
35210         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35211         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35212         (vect_permute_store_chain): Likewise.
35213         (vect_grouped_load_supported): Likewise.
35214         (vect_permute_load_chain): Likewise.
35215         (vect_shift_permute_load_chain): Likewise.
35216         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35217         (vect_transform_slp_perm_load): Likewise.
35218         (vect_schedule_slp_instance): Likewise.
35219         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35220         (vectorizable_mask_load_store): Likewise.
35221         (vectorizable_bswap): Likewise.
35222         (vectorizable_store): Likewise.
35223         (vectorizable_load): Likewise.
35224         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
35225         vec_perm_indices instead of auto_vec_perm_indices.  Use
35226         tree_to_vec_perm_builder to read the vector from a tree.
35227         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
35228         vec_perm_builder instead of a vec_perm_indices.
35229         (have_whole_vector_shift): Use vec_perm_builder and
35230         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
35231         truncation to calc_vec_perm_mask_for_shift.
35232         (vect_create_epilog_for_reduction): Likewise.
35233         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
35234         from auto_vec_perm_indices to vec_perm_indices.
35235         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35236         instead of changing individual elements.
35237         (aarch64_vectorize_vec_perm_const): Use new_vector to install
35238         the vector in d.perm.
35239         * config/arm/arm.c (expand_vec_perm_d::perm): Change
35240         from auto_vec_perm_indices to vec_perm_indices.
35241         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35242         instead of changing individual elements.
35243         (arm_vectorize_vec_perm_const): Use new_vector to install
35244         the vector in d.perm.
35245         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
35246         Update vec_perm_builder constructor.
35247         (rs6000_expand_interleave): Likewise.
35248         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
35249         (rs6000_expand_interleave): Likewise.
35251 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35253         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
35254         to qimode could truncate the indices.
35255         * optabs.c (expand_vec_perm_var): Likewise.
35257 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35259         * Makefile.in (OBJS): Add vec-perm-indices.o.
35260         * vec-perm-indices.h: New file.
35261         * vec-perm-indices.c: Likewise.
35262         * target.h (vec_perm_indices): Replace with a forward class
35263         declaration.
35264         (auto_vec_perm_indices): Move to vec-perm-indices.h.
35265         * optabs.h: Include vec-perm-indices.h.
35266         (expand_vec_perm): Delete.
35267         (selector_fits_mode_p, expand_vec_perm_var): Declare.
35268         (expand_vec_perm_const): Declare.
35269         * target.def (vec_perm_const_ok): Replace with...
35270         (vec_perm_const): ...this new hook.
35271         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
35272         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
35273         * doc/tm.texi: Regenerate.
35274         * optabs.def (vec_perm_const): Delete.
35275         * doc/md.texi (vec_perm_const): Likewise.
35276         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
35277         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
35278         expand_vec_perm for constant permutation vectors.  Assert that
35279         the mode of variable permutation vectors is the integer equivalent
35280         of the mode that is being permuted.
35281         * optabs-query.h (selector_fits_mode_p): Declare.
35282         * optabs-query.c: Include vec-perm-indices.h.
35283         (selector_fits_mode_p): New function.
35284         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
35285         is defined, instead of checking whether the vec_perm_const_optab
35286         exists.  Use targetm.vectorize.vec_perm_const instead of
35287         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
35288         fit in the vector mode before using a variable permute.
35289         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
35290         vec_perm_indices instead of an rtx.
35291         (expand_vec_perm): Replace with...
35292         (expand_vec_perm_const): ...this new function.  Take the selector
35293         as a vec_perm_indices rather than an rtx.  Also take the mode of
35294         the selector.  Update call to shift_amt_for_vec_perm_mask.
35295         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
35296         Use vec_perm_indices::new_expanded_vector to expand the original
35297         selector into bytes.  Check whether the indices fit in the vector
35298         mode before using a variable permute.
35299         (expand_vec_perm_var): Make global.
35300         (expand_mult_highpart): Use expand_vec_perm_const.
35301         * fold-const.c: Includes vec-perm-indices.h.
35302         * tree-ssa-forwprop.c: Likewise.
35303         * tree-vect-data-refs.c: Likewise.
35304         * tree-vect-generic.c: Likewise.
35305         * tree-vect-loop.c: Likewise.
35306         * tree-vect-slp.c: Likewise.
35307         * tree-vect-stmts.c: Likewise.
35308         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
35309         Delete.
35310         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
35311         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
35312         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
35313         (aarch64_vectorize_vec_perm_const): ...this new function.
35314         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35315         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35316         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
35317         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
35318         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35319         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35320         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
35321         into...
35322         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
35323         check for NEON modes.
35324         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
35325         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
35326         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
35327         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
35328         into...
35329         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
35330         the old VEC_PERM_CONST conditions.
35331         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
35332         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
35333         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
35334         (ia64_vectorize_vec_perm_const_ok): Merge into...
35335         (ia64_vectorize_vec_perm_const): ...this new function.
35336         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
35337         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
35338         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
35339         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
35340         * config/mips/mips.c (mips_expand_vec_perm_const)
35341         (mips_vectorize_vec_perm_const_ok): Merge into...
35342         (mips_vectorize_vec_perm_const): ...this new function.
35343         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
35344         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
35345         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
35346         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
35347         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
35348         (rs6000_expand_vec_perm_const): Delete.
35349         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
35350         Delete.
35351         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35352         (altivec_expand_vec_perm_const_le): Take each operand individually.
35353         Operate on constant selectors rather than rtxes.
35354         (altivec_expand_vec_perm_const): Likewise.  Update call to
35355         altivec_expand_vec_perm_const_le.
35356         (rs6000_expand_vec_perm_const): Delete.
35357         (rs6000_vectorize_vec_perm_const_ok): Delete.
35358         (rs6000_vectorize_vec_perm_const): New function.
35359         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35360         an element count and rtx array.
35361         (rs6000_expand_extract_even): Update call accordingly.
35362         (rs6000_expand_interleave): Likewise.
35363         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
35364         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
35365         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
35366         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
35367         (rs6000_expand_vec_perm_const): Delete.
35368         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35369         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35370         (altivec_expand_vec_perm_const_le): Take each operand individually.
35371         Operate on constant selectors rather than rtxes.
35372         (altivec_expand_vec_perm_const): Likewise.  Update call to
35373         altivec_expand_vec_perm_const_le.
35374         (rs6000_expand_vec_perm_const): Delete.
35375         (rs6000_vectorize_vec_perm_const_ok): Delete.
35376         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
35377         reference to the SPE evmerge intructions.
35378         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35379         an element count and rtx array.
35380         (rs6000_expand_extract_even): Update call accordingly.
35381         (rs6000_expand_interleave): Likewise.
35382         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
35383         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
35384         new function.
35385         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35387 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35389         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
35390         vector mode and that that mode matches the mode of the data
35391         being permuted.
35392         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
35393         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
35394         directly using expand_vec_perm_1 when forcing selectors into
35395         registers.
35396         (expand_vec_perm_var): New function, split out from expand_vec_perm.
35398 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35400         * optabs-query.h (can_vec_perm_p): Delete.
35401         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
35402         * optabs-query.c (can_vec_perm_p): Split into...
35403         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
35404         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
35405         particular selector is valid.
35406         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35407         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35408         (vect_grouped_load_supported): Likewise.
35409         (vect_shift_permute_load_chain): Likewise.
35410         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35411         (vect_transform_slp_perm_load): Likewise.
35412         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35413         (vectorizable_bswap): Likewise.
35414         (vect_gen_perm_mask_checked): Likewise.
35415         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
35416         implementations of variable permutation vectors into account
35417         when deciding which selector to use.
35418         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
35419         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
35420         with a false third argument.
35421         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
35422         to test whether the constant selector is valid and can_vec_perm_var_p
35423         to test whether a variable selector is valid.
35425 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35427         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
35428         * optabs-query.c (can_vec_perm_p): Likewise.
35429         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
35430         instead of vec_perm_indices.
35431         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
35432         (vect_gen_perm_mask_checked): Likewise,
35433         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
35434         (vect_gen_perm_mask_checked): Likewise,
35436 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
35438         * optabs-query.h (qimode_for_vec_perm): Declare.
35439         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
35440         (qimode_for_vec_perm): ...this new function.
35441         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
35443 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35445         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
35446         does not have a conditional at the top.
35448 2018-01-02  Richard Biener  <rguenther@suse.de>
35450         * ipa-inline.c (big_speedup_p): Fix expression.
35452 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
35454         PR target/81616
35455         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
35456         for generic 4->6.
35458 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
35460         PR target/81616
35461         Generic tuning.
35462         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
35463         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
35464         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
35465         cond_taken_branch_cost 3->4.
35467 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
35469         PR tree-optimization/83581
35470         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
35471         TODO_cleanup_cfg if any changes have been made.
35473         PR middle-end/83608
35474         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
35475         convert_modes if target mode has the right side, but different mode
35476         class.
35478         PR middle-end/83609
35479         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
35480         last argument when extracting from CONCAT.  If either from_real or
35481         from_imag is NULL, use expansion through memory.  If result is not
35482         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
35483         the parts directly to inner mode, if even that fails, use expansion
35484         through memory.
35486         PR middle-end/83623
35487         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
35488         check for bswap in mode rather than HImode and use that in expand_unop
35489         too.
35491 Copyright (C) 2018 Free Software Foundation, Inc.
35493 Copying and distribution of this file, with or without modification,
35494 are permitted in any medium without royalty provided the copyright
35495 notice and this notice are preserved.